Carsten Marohn Rainforestation Farming on Leyte Island, Philippines

Carsten Marohn

Rainforestation farming on Leyte island, Philippines – aspects of soil fertility and carbon sequestration potential

Institute for Plant Production and Agroecology in the Tropics and Subtropics Faculty Agricultural Sciences, University of Hohenheim, 2007

Institute for Plant Production and Agroecology in the Tropics and Subtropics

University of Hohenheim

Prof. Dr. J. Sauerborn

Rainforestation farming on Leyte island, Philippines –

aspects of soil fertility and carbon sequestration potential

Dissertation submitted in fulfilment of the requirements for the degree

'Doktor der Agrarwissenschaften' (Dr.sc.agr. / Ph.D. in Agricultural Sciences)

to the Faculty Agricultural Sciences

presented by

Carsten Marohn

Stuttgart

2007 This thesis was accepted as a doctoral dissertation in fulfilment of the requirements for the degree 'Doktor der Agrarwissenschaften' by the Faculty Agricultural Sciences at University of Hohenheim on 24/10/2007.

Date of oral examination: 02/11/2007

Examination Committee:

Supervisor and Review Prof. Dr. J. Sauerborn Institute for Plant Production and Agroecology in the Tropics and Subtropics, University of Hohenheim

Co-Reviewer Prof. Dr. R. Jahn Institute for Agricultural and Nutritional Sciences, Martin Luther-University Halle-Wittenberg

Additional Examiner Prof. Dr. J.N. Wünsche Institute for Special Crop Cultivation and Crop Physiology, University of Hohenheim

Vice-Dean and Head of the Prof. Dr. W. Bessei Committee Dean of the Faculty Agricultural Sciences, University of Hohenheim •

Contents 01 Introduction and scope...... 6 1.1 Study area...... 7 1.1.1 Geography...... 7 1.1.2 Geology, geomorphology and soils...... 8 1.1.3 Climate...... 13 1.1.4 Natural vegetation...... 17 1.1.5 Population, culture and economy...... 19 1.1.6 Land use, tenure and reforestation...... 22 1.2 Agroforestry systems...... 26 1.2.1 Rainforestation...... 27 1.3 Carbon sequestration...... 29 1.3.1 Climate change and carbon pools...... 29 1.3.2 Institutional and legal framework for funding of carbon sequestration through afforestation and reforestation...... 32 1.3.3 Potential of carbon sequestration through agroforestry...... 34 02 Sites, material and methods...... 36 2.1 Approach...... 36 2.2 Site selection for paired plots...... 36 2.3 Land-use history and plot installation in Cienda...... 38 2.3.1 Species and planting material...... 38 2.3.2 Plot design...... 43 2.4 Meteorological data...... 44 2.4.1 Weather data...... 44 2.4.2 PAR measurements...... 44 2.5 Soil analyses...... 45 2.5.1 Subplots and sampling scheme...... 45 2.5.2 Soil profiles...... 46 2.5.3 Soil sampling...... 46 2.5.4 pH...... 46 2.5.5 Bulk density and volumetric water contents...... 46 2.5.6 Gravimetric water contents and soil water potential...... 46 2.5.7 Particle size distribution...... 47 2.5.8 Total nitrogen...... 48 2.5.9 Phosphorus...... 48 2.5.10 CECeff, CECpot and base saturation...... 48 2.5.11 Exchangeable basic cations...... 49 2.5.12 Pedogenic oxides of Fe, Al and Mn...... 49 2.5.13 Soil organic carbon...... 49 2.5.14 Physical fractionation of soil organic matter...... 52 2.5.15 Substrate-Induced Respiration...... 53 2.5.16 Basal respiration...... 54 2.5.17 Soil respiration...... 54 2.5.18 Phosphatase activity...... 55 2.6 Biomass measurements...... 56 2.6.1 Mulched biomass...... 56 2.6.2 Undergrowth biomass and growth rates...... 56 2.6.3 Root length and weight density...... 56 2.6.4 Aboveground biomass growth of planted species...... 56

• 3 •

2.6.5 C, N and P contents of plant tissues...... 58 2.6.6 Litter production...... 59 2.6.7 Litter decomposition...... 59 2.7 Plant measurements required for modelling...... 60 2.7.1 Crop parametrisation...... 60 2.7.2 Tree parametrisation...... 60 2.8 Statistics...... 62 03 Characterisation of soils...... 63 3.1 Profile descriptions...... 63 3.1.1 Haplic Cambisol, Cienda PN1...... 64 3.1.2 Haplic Cambisol, Cienda PN2...... 66 3.1.3 Stagnic Luvisol, Cienda PN3...... 68 3.1.4 Dystric Nitisol, Cienda (Rainforestation demo plot)...... 70 3.1.5 Chromic Cambisol, LSU...... 72 3.1.5.1 Ferri-stagnic Luvisol, Marcos...... 74 3.1.6 Ferri-chromic Luvisol, Pangasugan...... 76 3.1.7 Hypereutric Cambisol, stagnic properties, Maitum...... 78 3.1.8 Stagnic Cambisol, Patag...... 80 3.1.8.1 Calcari-Mollic Leptosol, Punta...... 82 3.2 Synopsis and Discussion Soil Profiles...... 84 3.2.1 Parent Material...... 84 3.2.2 Formation of volcanic soils...... 85 3.2.3 Topography...... 85 3.2.4 Single parameters compared across all study sites...... 91 3.2.5 Water Balance...... 96 3.2.6 Ecological evaluation - summary...... 99 04 Effects of land use on soil rehabilitation – a paired plot approach...... 102 4.1 Land use history...... 102 4.2 Soil samples...... 103 4.2.1 Soil carbon, nitrogen and pH...... 103 4.2.2 Available Ca2+, Mg2+, K+ and Na+...... 104 4.2.3 Basal Respiration...... 105 4.2.4 Microbial carbon, Q10 and qCO2...... 107 4.2.5 Available PI and phosphatase activity...... 110 4.3 Leaf litter production and decomposition under different tree systems...... 112 4.3.1 Leaf litter production...... 112 4.3.2 Litter decomposition...... 112 4.4 Synopsis...... 114 05 Plant growth in an agroforestry system under different small-scale environments.....120 5.1 Site parameters...... 120 5.1.1 Soil organic carbon (Corg)...... 120 5.1.2 SOM pools derived by physical fractionation...... 122 5.1.3 Soil and basal respiration...... 127 5.1.4 Microbial carbon...... 130 5.1.5 Litter production...... 132 5.1.6 Litter decomposition...... 133 5.1.7 Root length and weight density...... 135 5.1.8 PAR measurements...... 136 5.1.9 Synopsis of environmental parameters...... 138 5.2 Plant performance...... 143

• 4 •

5.2.1 Planted abaca...... 143 5.2.2 Planted trees...... 147 5.3 Environmental conditions for abaca growth...... 152 5.3.1 Survival rates of abaca...... 152 5.3.2 Abaca growth...... 155 06 Modelling growth and carbon sequestration of agroforestry systems in Leyte...... 159 6.1 Parametrisation...... 159 6.1.1 Crop parametrisation...... 160 6.1.2 Tree parametrisation...... 161 6.1.3 Site data...... 162 6.1.4 Management and profitability...... 163 6.2 Model calibration...... 163 6.3 Validation...... 166 6.3.1 Abaca and tree seedlings in Cienda...... 166 6.3.2 Trees planted 1996 at Cienda site...... 169 6.3.3 Trees planted at LSU in 1991-3...... 170 6.4 Modelling land use scenarios...... 172 6.4.1 Scenarios...... 172 6.5 Modelling outputs...... 174 6.5.1 Total carbon balance compared across land uses...... 174 6.5.2 Biomass distribution between plants and agroforestry zones...... 175 6.5.3 Soil conservation...... 179 6.5.4 Nutrient supply – acquisition of subsoil Phosphorus...... 180 6.6 General discussion...... 181 6.6.1 Evaluation of modelling assumptions...... 181 6.6.2 Magnitudes of stocks...... 183 6.6.3 Lessons learnt from modelling with WaNuLCAS...... 186 07 Conclusions and outlook...... 189 08 Abstract...... 195 09 Kurzfassung...... 198 10 Acknowledgements...... 201 11 References...... 202 12 Appendices...... 218

• 5 • 1 Introduction and scope

1 Introduction and scope In recent years variety of successional forestry and agroforestry schemes have been promoted and implemented in tropical and subtropical regions in order to make better use of resources, increase system resilience, mitigate environmental impacts of resource use and/or increase farmers' income (e.g. HART 1980, EWEL 1999, COICAP 1999, MILZ 2001). In the Philippines as one of the most severely deforested countries worldwide (KUMMER & TURNER 1994), conservation of tree biodiversity is another important objective of such systems. In the mid-1990s, the official Philippine reforestation scheme based on few fast-growing exotic tree species was contrasted with a more diversified planting system. Reintroducing high diversity of indigenous trees in dense multi-storey structure, a so-called high-density closed canopy system was developed (MARGRAF & MILAN 1996). Focus of this approach was clearly on conservation related to indigenous timber trees, mainly of the Dipterocarpaceae family. Later, more importance was given to profitability of the system, especially during the critical first years after planting, and fruit trees were assigned higher priority. Another step towards rentability could be taken through the participation in Clean Development Mechanism (CDM) projects, which reward carbon sequestration through reforestation. For this study high-density closed canopy plots planted 1993-6 as well as a new plot (installed 2004) were surveyed in order to assess: 1. how site conditions (canopy closure, slope position and selected soil parameters) influence mortality and biomass production of trees and crops during the crucial first years after installation; for this purpose inventories, biomass measurements, PAR measurements and soil analyses were conducted on the 2004 plot;

2. how the trees influence selected soil parameters: For this research question a paired plot approach comparing the >10-year old high-density closed canopy plots to adjacent fallowed land and classical reforestation was chosen. Soil parameters were selected, that are supposed to indicate short- to mid-term changes in land-use;

3. if amounts of sequestered CO2 could make reforestation an option for CDM funding especially during the economically critical first years after planting. To assess this, biomass growth and carbon contents of the new plot as well as in fallowed plots were extrapolated to subsequent years using a computer model and validating results with existing inventories of high-density closed canopy plots.

• 6 • 1 Introduction and scope

1.1 Study area

1.1.1 Geography Among the more than 7.000 Philippine islands, Leyte is situated in the Centre- East Visayas archipelago. The Visayas are delimited by Luzon in the North, Mindanao in the South, Palawan to the West and the Pacific Ocean to the East (fig.1). They are divided into three administrative units: Western Visayas, comprising Panay and Negros Occidental, Central Visayas with Negros Oriental, Bohol and Cebu, and Eastern Visayas (or Region VIII) with Leyte, Samar and Biliran as main islands. Leyte island is situated at 9°55' to 11°48' N and 124°17' to 125°18' E. North-South extension is roughly 200km, distance from West to East ranges from 60km (between Tacloban and Palompon) to about 25km (Baybay to Abuyog). Total land area of Leyte is 762.178ha (NAMRIA, 2003). The Figure 1: Administrative regions of the island is divided into two administrative Philippines (CHOKKALINGAM ET AL. 2006) regions, Leyte with Tacloban, and Southern Leyte with Maasin as capital. Other commercial centres of Leyte are Ormoc City and the town of Baybay on the Western coast (fig.2). Baybay municipality, where most of the research sites are located, is situated in the Centre West of Leyte, characterised by its position between the Camotes Sea to the West and the steep slopes of the Cordillera Central to the East. Some of the highest peaks like Mt. Pangasugan and Mt. Emik rise up to more than 1200m asl within a distance of less than 10km from the coast. As an administrative unit, Baybay is bordered by the municipalities of Albuera and Burauen to the North, Javier, Abuyog and Mahaplag to the East and Inopacan to the South.

Figure 2: Leyte island, geography

• 7 • 1 Introduction and scope

1.1.2 Geology, geomorphology and soils

1.1.2.1 Tectonics and stratigraphy As a result of plate tectonic movements, the Eastern and Western margins of the Pacific were and still are subject to strong volcanic and seismic activities. Where the Eurasian and Pacific tectonic plates collide from Alaska through Kamchatka and Japan to the Philippines and Indonesia, intense volcanic activity has influenced the formation of terrestrial surfaces since the tertiary (SANTOS & RAMOS 1995). Within the Philippine archipelago, the Philippine Sea Plate from the East and the Eurasian Plate from the West subduct beneath the Philippine Mobile Belt, which is uplifted (ZANG & NING 2002). Along the subduction zones trenches and troughs were formed: During early Miocene the Manila Trench originated from a subduction of the Eurasian Plate under the Philippine Mobile Belt 1 (SANTOS & RAMOS 1995) . The Philippine Trench, extending along the Eastern margin of the Philippine territory from Luzon through Samar and Leyte to Mindanao, was formed during the Pliocene.

In the course of the tectonic processes which led to the formation of the Manila and (later) the Philippine Trench, a fault line broke up parallel to the latter. This Philippine Fault Line crosses the entire Philippine archipelago for approx. 1200km in a NNW-SSE course (fig. 3). Coming from the North, the fault line enters Leyte island near Biliran Strait. As its formation was closely related to upfolding processes (BARRIER 1991, after SANTOS & RAMOS 1995), the fault line coincides with or parallels the ridges of Leyte's central mountain range, the so-called Leyte Cordillera. North of Ormoc the fault line splits up into at least three parallel lines which form a pull- apart zone of 60km length. Within this tensional block the highest volcanic activity Figure 3: Tectonic plates and trenches in the occurred during the Miocene. Some of the Philippines (SANTOS & RAMOs 1995) formerly most active volcanoes (Alto Peak, Mt. Lobi, Mt. Mahagnao) are situated here and at present geothermal energy is extracted in the Greater Tongonan Area. Further south at least two fault lines cross each other North-east of Baybay town, before they finally separate into three and travel along both sides of Sogod Bay and Cabalian Bay (ANONYMOUS 1993). In contrast to the tectonic movement around the Manila Trench, which has almost come to a halt (SANTOS & RAMOS 1995), the Philippine Fault Line is still in motion. During an observation period from 1991-2002, BACOLCOL, BARRIER & DUQUESNOY (2004) measured an annual displacement rate of 2.3-3.6cm. Main structural processes in Central Leyte's fault zone, East-West compression and wrench faulting, are still going on (PANEM 1992). The

1 According to YUMUL ET AL. (2003) a collision of the Palawan microcontinental block with the Philippine Mobile Belt caused a rotation of Luzon, which then onramped the South China Sea Plate.

• 8 • 1 Introduction and scope resulting tensions are absorbed by slipping and seismic activities close to the fault zone (BACOLCOL, BARRIER & DUQUESNOY 2004). The Philippine Fault Line divides Leyte into two geologically different rock formations: To the West, the Central Philippine Arc Terrane (CPAT) consists of a sedimentary basement superimposed by volcanic layers, which are interrupted by marine sediments. To the East the East Luzon-Samar-Mindanao Disrupted Terrane (ELSMDT), based on metamorphites, is more heterogeneous and contains limestone, clastic sediments, volcanic and metamorphic elements. Both are divided by the Burauen Graben, but have been amalgamated. Historically, in Leyte province a pre-oligocene basement complex of amalgamated ultramafic and metamorphic rocks was overlaid by marine sediments until early to middle miocene (AQUINO ET AL. 1983). Volcanism then lead to folding, intrusions, extrusions and volcanic flows until the Pleistocene; thus porphyric and dacitic layers alternate with sedimentary sequences (BAYRANTE 1982). After the quarternary recession of the sea, mainly sedimentary deposits like siltstone and conglomerates (AQUINO ET AL. 1983) were left behind. In the following, calc-alkaline volcanics and intrusives dominated (SANTOS & RAMOS 1995); pyroclastic flows composed of a crystal-vitric tuff matrix containing pumice and andesitic material (BAYRANTE 1982) constitute the parent material of many volcanic soils in the area. Other important components are calcareous rocks and breccia. Once volcanism had ceased, weathering and erosion became the dominating factors for land formation and in recent times alluvial depositions accumulated in the lowlands and river deltas (AQUINO ET AL. 1983).

1.1.2.2 Surface geology Surface geology along Leyte Cordillera is determined by volcanic constituents of miocenic origin like basaltic and andesitic materials, partly covered by younger (pliocenic) conglomerates and pyroclastics and – in the outer zones – by quarternary volcanic ashes (JAHN & ASIO 1998). This also applies to the study area: The western foothills and mountain slopes of the cordillera developed from intermediate basaltic volcanoclastics (JAHN & ASIO 1998), which cover the miocenic Burauen volcanics (andesites, dacitic flows and basalt). The latter form the geological surface on the east side of the mountains. South of Baybay, coralline limestone and volcanic sediments alternate on a small scale. Further south, coralline limestone dominates (fig.4). The eastern and north-eastern regions as well as Ormoc Valley and the western coastal areas are characterised by holocene alluvial lowlands extending towards the footslopes of the Cordillera. On the very north- easterly tip of the island, bordering Samar, a cretaceous mountain complex represents the oldest geological formation in Leyte. Figure 4: Geology of Leyte (SANTOS & RAMOS 1995)

• 9 • 1 Introduction and scope

1.1.2.3 Present landforms The central cordillera traverses Leyte island along the Philippine Fault Line from NNW to SSE, its highest elevation, Alto Peak, reaching more than 1300m asl (JAHN & ASIO 1998). Some of the highest summits in Baybay municipal district like Mt. Pangasugan with its almost vertical slopes can be seen from all study sites on the West coast between Marcos and Patag. These mountains rise up to more than 1100m asl within a distance of merely 5-6km from the coastline (BUREAU OF COAST AND GEODETIC SERVICE 1982). Valleys are generally V- shaped in the upper parts of this central portion of the cordillera. The deep valleys and deeply weathered saprolites in the lower parts have been interpreted as indicators for long-lasting erosion processes (JAHN & ASIO 1998). As a consequence of the rugged relief and heavy rains, erosion and landslides are dominating natural processes in this landscape2. In front of the central mountain range, foothills and isolated plateaus spread half way to the coastline. Figure 5: Topography of Leyte Southwards from Pagbanganan River, which discharges in Baybay city, mountain elevations decrease gradually and landscape forms are more gentle due to their limestone origin. Maitum and Punta study sites are located in this area, where calcareous and volcanic materials coexist. Further south, coralline Karst dominates the landscape (see fig.5). To the east of the Cordillera, the mountain spurs bottom out into an extensive lowland plain that finally reaches the Pacific coast. The northeastern part of Leyte, too, is dominated by alluvial lowland plains with the exception of a mountain group at the very tip, where Leyte faces Samar island.

1.1.2.4 Soils Soil forming processes: Among the soil forming factors – parent material, organisms, topography, climate (BIRKELAND 1999) and time (LAVELLE & SPAIN 2005) – the first factor geologically subdivides Leyte soils into calcareous and volcanic regions. High temperatures and humid climate are strong driving factors for soil formation particularly in well-drained tropical soils, where they accelerate mineralisation, weathering of the parent material, leaching and loss of bases, resulting in acidification, desilification (leading to a relative accumulation of Fe- and Al-sesquioxides) and formation of hematite. For the formation of the different clay minerals as well as for the sesquioxides, according to ASIO (1996), drainage of the soil (in other words: residence time of the weathering solution) plays a key role. In his mineralogic research on an Andosol – Alisol catena in Western Leyte, ASIO found that both profiles developed from basalt composed of plagioklase feldspars and pyroxene as most important minerals with lower concentrations

2 In many cases exacerbated by kaingin, the local slash-and-burn land use practice.

• 10 • 1 Introduction and scope of magnetite. These main components are easily weatherable (due to their narrow SiO2/Al2O3-ratio, s. BIRKELAND 1999) under humid tropical conditions. Weathering of the Andosol's pyroxene and plagioklase finally led to the formation of halloysite and kaolinite, secondary two-layer lattice silicates (JAHN & ASIO 1998) in the present Alisol. Besides these low-activity clays, goethite, hematite and quartz make up for most of the mineral compounds. On the other hand, upland soil formation does oftenly not take place in situ and many soils are still genetically immature. This is due to the steep slopes in the geologically young landscape, which make erosion and landslides – at least on volcanic rock and intensified by human influence – important soil forming factors (ASIO 1996). As an example of ancient natural erosion, JAHN & ASIO (1998) mention the remaining deeply weathered saprolitic peneplains west of the cordillera with their deep valleys. On a smaller scale and partly man-made, effects of erosion can be observed in most upland areas including the research sites, with colluvial material superimposed to the original soil. Moreover, many lowland soils with gleyic properties are still immature, because soil formation is impeded by waterlogging (ASIO 1996). Relocation of material within the soil profile is also characteristic for Leyte soils and can be observed as clay accumulation in argillic horizons3. Yet, although most of the uphill soils in Leyte are well-drained (BARRERA ET AL. 1954), ferralitic soils can only be found on some plateaus and hills with little colluvial influence, where soil genesis has taken place in situ. In some places soil profiles are divided into a colluvial brownish upper and a lower reddish zone. Due to the young geological age, weathering of the parent material has not progressed profoundly, which mitigates the effects of leaching: In spite of progressing acidification, contents in basic cations are still relatively high in most Leyte soils due to reserves still present in the minerals (JAHN & ASIO 1998). Investigating an Andosol – Alisol catena on Leyte's West Coast, ASIO (1996) identified humus accumulation, loss of bases, acidification, braunification, clay formation and desilification/ferralitisation as main driving processes for soil genesis. Among this study's experimental sites, humus accumulation was observed mainly in calcareous soils; although these are biologically very active (DAUB 2002), drought as a consequence of excessive drainage might impede mineralisation of humus to some extent. Distribution of soils: Compared to other tropical regions, many South-east Asian soils are relatively young and there is still a considerable share of soils rich in basic cations (such as Luvisols). Yet many Philippine upland soils are already depleted in basic cations due to their advanced stage of development, if for example an Andosol – [Luvisol4 / ] Alisol – 5 [Acrisol] chronosequence is assumed (s. ASIO 1996). According to the FAO Soil Map of the World, most soils in the Philippines are Acrisols, followed by Cambisols and Luvisols (see table 1). For hilly soils, Acrisols cover the largest portion with more than 5 million ha, Luvisols, Cambisols and Andosols follow with less than 500.000ha each (FAO Gateway).

3 On the other hand, ASIO (1996) suggests that in some volcanic Baybay soils clay formation took place in situ. 4 meaning that they have CEC of >24cmolC /kg clay distinguishing them from Acrisols. Alisols also have high CEC, but base saturation of < 50% and additionally alic properties within the major part between 25 to 100cm depth. 5 ASIO (1996) concludes from rock and soil samples of a representative Baybay Alisol developed from basalt, that this soil has already lost 90% of its original content of basic cations and 85% of the initial P2O5.

• 11 • 1 Introduction and scope

Table 1: FAO soil types in the Philippines (FAO Gateway for Land and Water Information) Soil Types Area (ha) Originally, soils in Leyte were classified according to suitability for agriculture, the Acrisols 12,596,447 main criterion being drainage. The classical Cambisols 8,680,048 soil survey by BARRERA ET AL. (1954) Luvisols 3,816,680 distinguishes poorly, moderately and well- Fluvisols 599,450 drained flat lowlands on one hand and well- Andosols 559,114 drained rolling uplands on the other, Gleysols 401,409 independently of calcareous or non- Nitisols 300,439 calcareous underground. Each of these Regosols 224,404 classes was divided into subtypes describing Kastanozems 216,461 texture and locality (e.g. Maasin clay, Arenosols 209,748 Pawing fine sandy loam or simply rough Phaeozems 52,472 mountainous land).

For western Leyte, the FAO Soil Map shows Acrisols, Luvisols and Cambisols as main soil groups (Ultisols and Alfisols in fig.6). A typical toposequence of volcanic origin could consist of Ochric Andosols (highest elevations), Orthic Acrisols and Luvisols down the slopes of the Cordillera, and Gleyic or Eutric Cambisols in the alluvial lowlands. For the lower mountains and hills, which are relevant for reforestation, Alisols and Cambisols have also been described besides the dominant Acrisols and Luvisols; soils on calcareous rock have been classified as Humic Acrisols (FAO), Phaeozems (DAUB 2002) and Cambisols.

Ecological evaluation of upland soils in western Leyte: To date only few soils in western Leyte have been studied in depth. Volcanic soils in the area are generally acidic in reaction and low in bulk density. 6 ASIO's research (1996) on a Haplic Alisol near LSU showed, that soil physical parameters (rooting depth, rootability, drainage, water and air capacity) were favourable for plant growth but to some extent attenuated by the soil's supposedly high erodibility. Among soil chemical parameters the main constraints were P and Figure 6: Soils of Leyte (ASIO & MODINA 1994) to some extent available K, whereas availability of Ca and Mg was not limiting for plant growth (ASIO ET AL. 1998).

6 This soil has been classified as Alisol by ASIO, DAUB and ZÖFEL due to its argic B, CEC > 24cmolc/kg and base saturation < 50%. No analysis was conducted to determine, whether the soil meets FAO requirement for alic properties (> 60% Al3+ of CEC) in the major parts between 25 and 100cm depth.

• 12 • 1 Introduction and scope

Sesquioxides, humic substances and low activity clay minerals were the main reasons for the high phosphorus retention in the Alisol which has been observed to an even higher degree in Baybay Andosols by ZIKELI (1998). Other limiting factors in these Andosols were low base saturation and CEC. On the other hand, physical structure of Andosols was found to be excellent as long as rooting depth was sufficient (ASIO 1996; ZIKELI 1998). A soil on coralline limestone on the study site in Punta (approx. 5km south of Baybay) was classified as Calcaric Phaeozem by DAUB (2002). It is characterised by a Mollic A horizon and a shallow AH -BC profile and, according to FAO, must not contain secondary CaCO3 within the upper 100cm. Many upland soils in Leyte with exception only of the steepest mountainous parts are degraded as originally forested land has been converted to other land uses decades ago (s. 2.1.6). ASIO (1996) identified changes in soil colour and structure, reduced thickness of AH and AB horizons as consequence of erosion, increased bulk density, reduced pore volume and aggregate stability caused by compaction, lower contents in humus and reduced soil respiration as indicators for past land clearing. Strategies to rehabilitate these soils include soil and canopy cover to enhance interception, transpiration and evaporation and thus reduce leaching, and tight nutrient cycles through diversified planting design.

1.1.3 Climate Seasons in the Philippines are mainly determined by wind and rainfall patterns as temperatures do not vary strongly. Three phenomena exercise most influence on the archipelago during the transcourse of the year (after ARAKAWA 1969; NIEUWOLT 1977): • The humid north-east monsoon from Oct – Mar, which originates in polar regions and is deflected as it enters the Philippines in south Luzon and Samar, thus in Leyte the main wind direction is SSW-wards. • North Pacific trade winds from changing directions occurring Mar – May, which carry dry air to Leyte, even at their lower boundaries. • The southwest or summer monsoon Habagat during May – Sep, which is more humid than the NE monsoon at the time it originates in the Indian Ocean, but loses part of its moisture passing Palawan and the western Visayas before reaching Leyte. Exposition and orographic lifting modify these general principles on a smaller scale: North and east exposed areas like e.g. Aparri, Legaspi (NIEUWOLT 1977) and Tacloban (ASIO 1996) are subject to a typical east coast rainfall pattern with maxima coinciding with the NE monsoon, while in Manila (W exposure) and S or SE exposed regions rainfall distribution is inverse with a maximum during the SW monsoon. Despite this, entire Leyte (including the West coast) belongs to the first rainfall subtype. CORONAS' (1920, as cited by ARAKAWA 1969) classification is based on rainfall patterns, dividing Leyte into two climatic zones east and west of the Cordillera. Both have in common, that there is no dry7 month, and least amounts of rain fall in springtime. On the east side a pronounced rain maximum during NE monsoon by far exceeds precipitation of the cyclonic rains during summer. The climate of western Leyte (Ormoc, LSU and Maasin) can be classified as Af according to KÖPPEN (1931), a rainforest climate with long-term mean temperature of the coldest 8 month >18°C and precipitations of the driest month >60mm . A so-called PLO wind regime refers to regular-directional trade winds and summer monsoon. Climate data for LSU, Baybay, have been observed since the 1970s by the PAGASA network (see fig.7).

7 If monthly averages are considered 8 Subclassifications characterising minor dry periods - m or s'' – are not permitted for average monthly rainfall >60mm.

• 13 • 1 Introduction and scope

Figure 7: Climate chart (rainfall, pan evaporation, minimum, mean and maximum air temperatures) for LSU, Baybay, based on PAGASA data 1976 – 2005

Mean average temperature 1976-2005 was 27.5°C with monthly means varying only about 2°C, far below daily amplitudes as e.g. 10.9°C observed by BALZER (1994). Average total annual rainfall was 2748mm. In many publications Baybay region is considered a humid area without any dry periods throughout the year. BALZER (1994) first questioned this standpoint: During Mar – May, monthly potential evapotranspiration can easily exceed rainfall indicating potential drought stress for plants. In addition, the monthly resolution usually depicted in climate charts does not show erratic rainfall with frequent dry periods of up to two weeks which, in a balance of monthly resolution, are compensated for by two or three heavy rains (see fig.9). During the period depicted in fig.8, for 63% of all days evaporation exceeded rainfall. In addition, during El Niño years precipitation would clearly fall below average values. As can be seen from fig. 7, 2004 was an exceptionally dry year, that received only 2317mm of rainfall compared to 3327mm in 2003 and 3159mm in 2005. In spite of favourable physical properties of many soils in western Leyte (JAHN & ASIO 1998), those with very good drainage or shallow soils (especially on limestone), would under these conditions not hold enough water to supply plants9. Thus, dry seasons do occur from a plant physiological point of view. Phenological observations also show, that many native trees shed their leaves during the drier period in Mar – May.

9 Although, according to KÖPPEN (1931), at annual precipitation >2000mm, a dry season of up to 4 months would not have any long-term impact on natural vegetation.

• 14 • 1 Introduction and scope

Figure 8: Precipitation and ETP for LSU Jan 2004- Apr 2006

In the perception of locals, seasons are more often linked to different wind regimes, like the wet Habagat monsoon. Typhoon season peaks from Aug to Sep, the Philippines being one of the most typhoon-prone regions in the world with an average of around 20 events per year (ARAKAWA 1969). These tropical storms can reach velocities of 10m/s as they reach the Philippines from the East. Some are deflected towards the North and streak the pacific side of the archipelago from Samar to NE Luzon. Others pass straight westward mainly through Samar and SW Luzon. Although Leyte's west coast is relatively sheltered by the mountain range and Baybay municipality is considered a non-typhoon-prone region, the area is affected by torrential rainfall brought along by typhoons which can cause flash floods and landslides10. Due to the extreme changes in landscape, weather data collected at PAGASA LSU (7m asl) need to be used with some caution for comparisons to the experimental site in Cienda, located 5km southeast from LSU and >100m asl. As an example, rainfall from end of May until beginning of July 2004 (onset of Westerly Habagat) was clearly higher at LSU and often did not even reach Cienda (fig. 9). Also during the dry season (Mar – May) Cienda is likely to receive less rain11. On the other hand, strong rainfall events are often caused by uprising air masses at the luff side of the Cordillera and tend to be more extreme in Cienda (examples May 14-16; June 8-9, June 14-16, July 26).

10 In 1991, heavy rain led to a landslide burying more than 5.000 inhabitants of Ormoc city, less than 50km north of the research area. In 2006, a similar disaster occurred in Guinsaugon, Southern Leyte. 11 which is not evident in fig. 8 due to a break-down of the rain gauge at Cienda from May 18-Jun 5

• 15 • 1 Introduction and scope

160 Rainfall at Cienda and LSU Rainfall Cienda [mm] 140 Rainfall PAGASA [mm] May 1 - Sep 30, 2004

120

100

40 Rainfall [mm]Rainfall

1-Mai-04 6-Jun-04 3-Jul-04 10-Mai-04 19-Mai-04 28-Mai-04 15-Jun-04 24-Jun-04 12-Jul-04 21-Jul-04 30-Jul-04 8-Aug-04 17-Aug-04 26-Aug-04 4-Sep-04 13-Sep-04 22-Sep-04

150 Rainfall Cienda [mm] 130 Rainfall at Cienda and LSU Rainfall PAGASA [mm]

110 Oct 1, 2004 - Apr 10, 2005

50 Rainfall [mm] Rainfall

-10

1-Okt-04 5-Dez-04 8-Feb-05 6-Mrz-05 1-Apr-05 14-Okt-04 27-Okt-04 9-Nov-04 22-Nov-04 18-Dez-04 31-Dez-04 13-Jan-05 26-Jan-05 21-Feb-05 19-Mrz-05 Figure 9: Rainfall at LSU and Cienda sites

For air temperatures, differences typical for continental versus maritime climates could be observed on a small scale: The land inward site Cienda showed higher daily amplitudes for minima as well as for maxima compared to coastal LSU in 2004 - 5 (fig.10). Coincidence of the lowest maxima values at both sites can be taken as a quality control of measurements. Many days with low maxima at Cienda could be explained as clouds stopped by the mountains. Annual means for LSU, however, exceeded the ones at Cienda clearly with respect to maxima and slightly for minima.

• 16 • 1 Introduction and scope

Minimum and maximum air temperatures at Cienda and LSU Min. Cienda Min. LSU Daily means, May 1, 2004 - April 10, 2005 Max. Cienda Max. LSU 35

25 Air temperatureAir [°C]

06-Jul-04 28-Jul-04 01-Mai-04 23-Mai-04 14-Jun-04 19-Aug-04 10-Sep-04 02-Okt-04 24-Okt-04 15-Nov-04 07-Dez-04 29-Dez-04 20-Jan-05 11-Feb-05 05-Mrz-05 27-Mrz-05 Figure 10: Minimum and maximum air temperatures in Cienda and LSU from May 1st, 2004 until April 10th, 2005. Note the stretched scale, which may overemphasise some trends Soil climate has been classified as isohyperthermic, i.e. above 22°C throughout the year with amplitude of less than 6°C in 50cm depth (ASIO 1996).

1.1.4 Natural vegetation The Visayas are part of the paleotropic Malesian Floristic Region, Philippinean Province. The latter includes the entire Philippine archipelago to the exception of Palawan and the Calamian Islands. After the most recent glacial period and the subsequent rise of the sea level the Philippinean Province was separated from Borneo, Sulawesi and New Guinea (RANGIN ET AL. 1989, as cited by LANGENBERGER 2003), to which it had been linked before. Australian and mainland Asian species, which occur naturally in the Philippines, indicate that there were land bridges between today's Mindanao and the Australian Region and also between Luzon and Taiwan as well as the Asian continent. Still, the Philippines host many endemic species, which is characteristic for islands. DENR/UNEP (1997) report, that 5% of the World's flora, more than 13500 species, and 22.5% of the Malesian vascular flora grow in Philippine forests. DAVIS ET AL. (1995 as cited by LANGENBERGER 2003) found that 39% out of 8900 vascular plants were endemic, and LANGENBERGER (2003) identified 52% endemic tree species in his survey conducted on the slopes of Mt. Pangasugan, Leyte. There have been various tree species inventories in the Philippines, starting from the classical study by WHITFORD (1911), who established the classification of Philippine forest types, which is still used today. According to Whitford, ten main forest types can be distinguished in the Philippines. All of these except the Pine Type Forest exist in Leyte and are described in more detail as an idealised toposequence from shore to summit with special reference to Mt. Pangasugan (table 2).

• 17 • 1 Introduction and scope

Table 2: Forest types of Leyte after WHITFORD (1911). Explanations on habitats by Langenberger (2003 and pers. comm.) 1 Mangrove type forests 2 Beach type, dominated by Terminalia spp. and Calophyllum spp. 3 Lauan-Hagakhak type prevails from sea level up to 150m asl and in higher elevations along riverbeds. Temporary waterlogging and even flooding is tolerated. Species are adapted to a short or no dry season and are almost evergreen. Apart from Shorea contorta and Dipterocarpus validus, Toona kalantas, Dracontomelon dao, D. edule, Terminalia microcarpa and T. nitens are typical trees of Lauan- Hagakhak forests. Palms, lianas and smaller trees are typical understorey species. 4 Yakal-Lauan type forests grow at sea level in climates without or with short dry seasons and are slightly deciduous. On the slopes of Mt. Pangasugan this type was found on steep and/or dry sites as well as on old landslides together with long-lived pioneer species. 5 Molave type: The dominant and eponymous species is Molave, Vitex parviflora. This forest type is bound to lower elevations from 0-150m asl, occurs in climates with no to distinct dry season, and is deciduous. Often it is restricted to dry sites by competing Dipterocarp forests. 6 Lauan type: Different Dipterocarp species occur jointly and dominate this forest type. They form a homogeneous evergreen canopy of up to 50m height. Elevation range is from 0-400m asl, with short or no dry season. In Baybay this type can still be found on moderately steep sites with comparatively deep rooting space around 400m asl. Typical places for Lauan type forests are deeply weathered peneplains like the Cienda demo farm (s. 3.1.4). 7 Lauan-Apitong type forest is found between 0 and 400m asl, but in areas with a pronounced dry season. As a consequence species are deciduous. 8 Tangile-Oak type forest, named after Shorea polysperma (Tangile), can be found mainly on ridges of Mt. Pangasugan from 400-900m asl. Typical species like Tristania decorticata, Hopea acuminata or Cinnamonum mercadoi are adapted to a short or no dry season and evergreen. 9 Mossy type forests are restricted to higher elevations (above 900m asl). Trees are evergreen and often stunted in growth.

According to WHITFORD (1911), numbers 3 to 7 represent the Dipterocarp forest dominated by approx. 75% dipterocarp trees. Other important families growing below the Dipterocarpaceae canopy are Ebenaceae as well as Rubiaceae, Euphorbiaceae and Myrsinaceae in the understorey. Although the craggy and fragmented terrain on Mt. Pangasugan's slopes makes it difficult to directly apply WHITFORD's scheme, some of the main categories are supposed to dominate the area west of Mt. Pangasugan. These are Beach Forest and Lauan- Hagakhak in lower elevations and Lauan Type on plateaus. Above 450m (lower montane forests), species belonging to Mossy Forest can be found. Apart from elevations commonly used for categorisation, LANGENBERGER (2003) found that to an even larger extent species composition is determined by relief position. In a vertical zonation of dipterocarp forests, he found the highest species diversity in the undergrowth, a somewhat lower influence on plant biodiversity was attributed to relief position.

• 18 • 1 Introduction and scope

Another important impact factor on plant biodiversity is, according to CONNELL (1978), non- equilibrium, which is typical along Leyte central cordillera with its relatively young soils and frequent landslides. An important aspect with regard to the establishment of plantations and for reforestation is availability of seeds. Mature mother trees have to be observed regularly and visited for seed harvest. As these trees are scarce and often situated in remote areas, age of first flowering and fruiting has been monitored by some authors to assess the feasibility of seed nursery installation. Estimating the life cycle of dipterocarp trees between 300 and 1400 years (and maturity for extraction about 70-140 years), NG (1966) notes the first flowering age of 50 dipterocarp species planted in an arboretum in Kepong, Malaysia, between the 20th and 30th year, mentioning also high variability between individuals of the same species. FOXWORTHY (1932, as cited by NG) observed first flowers for some of the same species at 6-20 years, which corresponds with findings of LANGENBERGER (2005) and own observations.

1.1.5 Population, culture and economy12

1.1.5.1 Population and housing Among the 76.5mn inhabitants of the Republic of the Philippines in 2000 (according to the census of 2000; in 2004 there were 83.75mn), the Eastern Visayan or Region VIII holds a share of 4.72% (see table 3). Out of these, 54% lived on Leyte island in 2000. In relationship to land area, 2.74% of the Philippine population reside on 2.85% of the national territory, namely Leyte island, including provinces Leyte, Southern Leyte (split from Leyte province in 1959) and Biliran (split off 1992). The ratio of land area / population almost doubles, when the less densely populated Samar provinces, also part of Region VIII, are included in the calculations.

Table 3: Population of Leyte Island in (calculated after NATIONAL STATISTICS OFFICE 2000)

Population Population [%] Area km2 Area [%] Philippines 76.498.700 100.00 300.000 100.00 Eastern Visayas 3.610.355 4.72 21.432 7.14 Leyte 1.592.336 2.08 6.268 2.09 Southern Leyte 360.160 0.47 1.735 0.58 Biliran 140.274 0.18 555 0.19

The Philippines are often referred to as the country with the highest population growth in entire Asia. This has been linked to the strong influence of the Roman Catholic church on the causes13 and to the rapid decline of natural resources (which were once the base for one of the richest East Asian nations) and increase of poverty (DARGANTES & KOCH 1994) on the effects side. Still, KUMMER & TURNER (1994) found, that a statistical correlation of deforestation could be established rather to legal and infrastructural than population factors. Compared to nation-wide 2.36%, population growth rate in Leyte province was only 1.13% (during the reference period 1995-2000), having dropped from 1.89% during the previous quintennial. In Leyte province, about 21% of the population lived in the two major cities of Tacloban 12 Official data from the National Statistics Office, census 2000, or own calculations based on these figures, if not stated otherwise. 13 E.g. in 2005 the catholic church campaigned massively against governmental family planning initiatives.

• 19 • 1 Introduction and scope

(Province Capital) and Ormoc and another 6% in Baybay, Leyte's biggest village. The remaining residential areas consist of 40 municipalities, which are subdivided into Barangays of usually 1.000 to 2.000 inhabitants. Household size in Leyte province in 2000 was 4.92 persons, who shared less than 10m2 floor area in 25% and less than 20m2 in (cumulative) 51% of the households. For the Eastern Visayas, 51% of households were supplied with drinking water from a community grid, 18% out of these having an individual faucet. 48% used electricity as energy for lighting (others kerosene, liquefied gas), 65% of households had radio and 28% television.

1.1.5.2 Culture and human development In the 2000 census interviews, 40% of Leyte's inhabitants categorised themselves as Bisayan, 38% as Waray, third most important ethnic group was Cebuanos with 20%. Correspondingly, main native dialects are Bisayan, Waray-Waray in the NE of the island and Cebuano; the official Philippino dialect Tagalog is understood in Leyte; a minority speak Kankanai. English has become official teaching language for colleges and universities and is widely spoken by the younger generation and professionals. Education at elementary level had been achieved by 52% of persons older than 5 years, of secondary level by 23%. In a Human Development Index survey carried out by the National Statistical Coordination Board in 2000, Leyte ranks 49th and Southern Leyte 31st among the 77 Philippine provinces. Both improved 7 and 6 ranks, respectively, since the last survey in 1997. HDI is the product of weighed factors describing life expectancy, education (enrolment and functional literacy) and real per capita income. Dominating religion as in most regions of the Philippines is catholicism (93%), being competed by an increasing number of other christian groups.

1.1.5.3 Economy and income generation Gross Regional Domestic Product of Region VIII grew 6.9% in 2004 as compared to the previous year. Shares of the three main sectors were evenly distributed (see fig.11). The agricultural, fisheries and forestry sector is clearly dominated by its first subsectors which contributed 99.9% of GRDP and since 2000 maintained a relatively stable annual growth around 5%. Forestry sector is slowly recovering after shrinking to 0.6% of its year 2000 size within only three years. According to the National Statistics Office's Special Review on Agriculture (2004), about 30% of the households in Region VIII (for Leyte 20%) owned agricultural land, the average farm size for Leyte being 1.9ha. Main agricultural product of the Eastern Visayas was rice before tubers / roots / bulbs, corn and sugarcane among the annual crops (in Leyte corn ranked second). Irrigated land amounts to almost 19% of agricultural areas and is mainly used for rice Figure 11: Gross Regional Product Region production. VIII (source NATIONAL STATISTICAL COORDINATION Coconut dominated the permanent crops BOARD)

• 20 • 1 Introduction and scope followed by abaca (mainly from Southern Leyte) and banana (data based on individuals, not hectares). Livestock production in Region VIII is concentrated in Leyte with approx. 200.000 heads of hog. Other important species are carabao (water buffalo) and chicken, which have doubled in numbers from 1991-2002. Leyte and to a larger extent Samar are among the poorest Philippine islands, which can be concluded from income-related indices. Real per capita income in 2000 was 13.267PHP for Leyte, less than two thirds of the philippine average (21.104PHP). Where available income is not sufficient to cover the costs of food and basic needs (determined on an average basis), the household would drop below a theoretical poverty line. Although costs of living are lower in the country-side, the situation for rural Leyte became more severe between 1997 and 2000, whereas in urban areas improvements could be observed (see fig. 12).

For four barangays in the research area, a sociological survey carried out by DAGOY ET AL. (1994) showed, that 70% of households were squats and most depended on land areas smaller than one hectare. One of the most urgent shortcomings mentioned in >200 interviews was land tenure (no official titulation) apart from income and job problems. A large proportion of the dwellers were migrants. ASIO (1994) mentions violence (World War II and political conflicts during the 1970s to 80s) as a main cause for migration apart from government resettlement programs, that issue stewardship contracts on state-owned land in remote areas. DARGANTES & KOCH (1994) investigated motives and habits of migrants, who as a consequence of lacking official land title Figure 12: Number of households below poverty and job opportunities often see forest line (after NATIONAL STATISTICAL COORDINATION BOARD) resources as a source of monetary income. These so-called forest farmers represent the lowest socio-economic class in society. They perceive forests as common goods. Some practice a merely extractive form of income generation, that frequently extends onto private land. Others take a piece of forest land under cultivation with or without informal consent of the present owner or community. Often these informal claims are recognised among villagers and remain valid over generations (pers. comm. of a forest farmer from Guadalupe, Baybay). This kind of customary law can be based on the fact that a person was the first to clear an area of natural forest and cemented by successive planting of coconut trees. Even officials of social forestry and land reform projects are reported to consider these factors once the area is declared Alienable & Disposable (DARGANTES & KOCH 1994).

• 21 • 1 Introduction and scope

1.1.6 Land use, tenure and reforestation Originally the Philippines were mostly covered by forests, agriculture being practised only along the coasts. Among the plant species of most commercial interest were the hardwood timbers of the Dipterocarpaceae family, which are known under trade names such as Red and White Meranti, Merawan or Balau (SOERIANEGARA ET AL. 1993) or Philippine Mahogany, among others. The economic value of these timbers led to drastic deforestation, which started during colonial times under the Spaniards. LASCO ET AL. (2001) estimated, that until 1521, still 27mn ha (90% of the area) of the Philippines were covered by primary forests. Under the Spanish regime, deforestation at a larger scale began. When the Philippines were 'handed over' to the USA in 1898, still 70% of the land area were primary forest (KUMMER 1992). Most parts of the Philippines suffered severe deforestation especially during the second half of the 20th century. Due to the high profitability of Dipterocarp forests (high density of premium timber per land area), and rising demand for Southeast Asian timber in Europe, the USA and Japan, more than half of the Philippine rainforest were logged over between 1945 and 1987 (KUMMER & TURNER 1994). PULHIN ET AL. (in CHOKKALINGAM ET AL. 2006) compiled historical data from different sources, demonstrating, that forest area has steadily decreased from 17.2 to 7.2 mn ha (equivalent to 57 and 24% of the country's area) between 1934 and 2003. Deforestation -1 rates reached an all-time high of 300,000ha a from 1977-1980 (GUIANG in DURST ET AL. 2001). According to the DENR Forest Management Bureau (2001; as cited in LASCO ET AL. 2004) less than 1mn ha of old growth forests exist to date and the Philippines have become a net importer of wood (DURST ET AL. 2001) . Some Visayan islands like Negros and Cebu have been completely clear-cut. For Leyte, 60% of the original forest had been clear-cut by the end of World War II; by the 1990s despite a nationwide logging ban implemented stepwise starting in the 1980s14, only 10% of primary rainforest were left (ASIO 1996, quoting BARRERA 1954 and DEPT. OF AGRICULTURE 1992). The declaration of logging bans without providing the necessary means for control had shifted most of logging operations into illegality and exacerbated the situation (GUIANG in DURST ET AL. 2001). Remaining patches of primary forest are mainly located in the steep and inaccessible parts along the Cordillera Central. During the 1990s the average annual deforestation rate amounted to 89,000ha or 1.4% of the forested areas15. NAMRIA (2003, based on an SSC-SPOT survey 1987-88) statistics tell that forested areas comprise 37% in Leyte and 49% in Southern Leyte. It is noteworthy, however, that forest from this point of view is a tenurial rather than biological category (s. below) and out of the 301,290ha of forested land in Leyte, only 7,570ha are old growth and 56,677ha are residual forests, the remaining majority being reproduction and brush area. Also according to NAMRIA forest cover statistics, less than 5% of the so-called forest area represents closed vegetation. DARGANTES & KOCH (1994) give similar figures: Out of the 309,000ha officially proclaimed as forest area (per definition all land >18% slope), 170,000ha underlie different land-uses, and only 93,000ha or <12% of Leyte's land area are still under forest. A more recent survey (FMB, Dec 31st, 200316) lists 65,977ha of forests for Leyte province, 57,332ha of open and 3,962ha of closed forests. After conversion of part of Leyte's original forest vegetation to secondary forest due to logging operations, large parts of the upland areas were colonised by farmers, who often

14 For Leyte in 1982 by MNR Administrative Order No. 31. LASCO & PULHIN (2003) state: The other forest types have been protected since 1992 under the National Integrated Protected Area System (NIPAS). However, most of these areas are protected only on paper and remain open access. 15 FAO 2000: www.fao.org/forestry/fo/fra/index.jsp 16 Download from http://forestry.denr.gov.ph/landuse8.htm Sep 25th, 2006. Plantation forests are not included.

• 22 • 1 Introduction and scope followed the logging roads. According to FDC-UPLB and FAO (as cited by DURST ET AL. 2001), only 10% of forest losses originated directly from logging operations, while slash- and-burn in logged-over areas (60%) and agricultural expansion (30%) accounting for the majority. Generally, the secondary forest was then cleared and planted to annual crops (traditional kaingin slash-and-burn system) or coconut. Besides the dominating coconut areas, nowadays scrublands and grassland cover large parts of the land. According to NAMRIA, 2003, non-forested areas are distributed as follows: 176,198ha are under coconut, 93,707 under to grassland and 77,024 under agricultural use (fig.13).

Land-Uses Leyte Island

Forest Mangrov es Coconut Grassland Agriculture Bare/Rocky Land Not Determined Built-Up

Figure 13: Shares of different land-uses in Leyte (compiled after NAMRIA 2003)

Productive lowland areas along the densely populated coast are generally planted to paddy rice. Around Ormoc commercial use as pasture or sugarcane plantations is common; these lands are relatively scarce (< 1ha/family) or unevenly distributed, so that most small farmers additionally, some exclusively, depend on remote upland areas' products for home consumption or income generation (DAGOY ET AL. 1994). Among the different forms of upland cultivation the following are most common: • Perennial systems like coconut in combination with pasture or secondary forest with abaca; these seem to be linked to titled land; • traditional slash-and-burn practice (kaingin) followed by planting of annuals like tubers; such lands are mainly (state-owned) forest areas under customary law; • typical products harvested by forest farmers are abaca, banana, tubers, pineapple and coconut. Hunting and extraction of NTFP like honey or rattan may also contribute to forest farmers' income. For Leyte island land uses are shown in fig. 14, a GIS-map based on SPOT satellite imagery. A typical zonation of land areas in a village in Baybay district is shown in fig. 15-16).

• 23 • 1 Introduction and scope

Figure 14: Land cover map issued by FMB-DENR. Dark green signatures indicate natural old growth forest, light green secondary forests, pink coconut and bush fallow, yellow rice or sugarcane

Figure 15: Barangay data board of Pangasugan showing typical zonation of different land uses. Right half LSU, left half, from bottom to top: Seashore, rice fields and village area, coconut area, forested mountain range

• 24 • 1 Introduction and scope

Figure 16: Typical land use in Baybay: Rice paddies on levelled lands, followed by old coconut plantations, secondary forest, natural landslides on the steep slopes and old growth forest around the summit of Mt. Pangasugan

Different kinds of land tenurial status coexist, from official A&D (alienable and disposable) and titled lands to squats or the traditional hereditary right to land use conceded by the community to the person who first cleared the forest. These customary rights coexist with state law, which establishes, that all areas >18% slope are by definition forest lands and as such excluded from agricultural use; many of the traditional kaingin areas are in fact found on slopes of >100% inclination. DURST ET AL. (2001) observe, that more than 5 mn ha of public forestlands in the timberland category are not covered by any form of tenure, and are considered 'open-access' areas. Even the official legislation reflects the conflict between agricultural use and protected areas, as the designation of land to either land use was frequently reversed depending on the change of political authorities (DARGANTES & KOCH 1994).

The Philippines were one of the first Asian countries to initiate reforestation programmes CHOKKALINGAM ET AL. (2006) distinguish three periods of reforestation efforts, from 1910 until the US colonial regime, from 1946 until the mid-1970s (which coincides with the highest deforestation rates) as government initiatives and until today as multi-sectoral approaches. It is remarkable, that during the first and second phase, the Philippine authorities used mostly indigenous species for reforestation. Phase 3 extended activities to the private sector and social forestry projects carried out by NGO; international funds were also involved to a high degree. The background of these efforts was an imminent timber shortage and consequently 80% of the planted trees were fast-growing exotic trees (mainly Swietenia macrophylla, Acacia mangium, A. auriculiformis, Eucalyptus spp. and Gmelina arborea). On the other hand, government institutions actively contributed to land conversion through settlement projects. Planting of coconut was frequently considered by authorities an argument pro legalisation of squatted areas (DARGANTES 1994). KUMMER & TURNER (1994) as

• 25 • 1 Introduction and scope well as DARGANTES (1994) see government colonisation projects in conjunction with institutional corruption and export-oriented policies as a main driving force of deforestation (see also GUIANG in DURST ET AL. 2001). Recently, the classical colonisation projects including reforestation with fast-growing tree species (s. 1.2.1) have been complemented with CBFM (community-based forestry management) and similar approaches. Such projects are based on a zoning into protected areas, buffer zones with limited, controlled, traditional and multiple use, the latter being not restricted. Despite the participative background of these programmes, it remains unclear, how farmers' communities with their limited resources can control the appropriate use of these areas. In addition, progressive legal initiatives are often not executed due to lack of personnel (GROETSCHEL ET AL. 2001) or political intrigues. FAO classifies almost 70% of the Philippine terrain as too steep, eroded or shallow for agriculture (Land Capability Classes M, N, Y. FAO Gateway for Water and Land Information, online source). Obviously, transformation of these steep forest sites into agricultural use lead to serious erosion problems even for relatively diversified smallholder farms (e.g. abaca, tubers, banana) under canopy. For wide-spaced extensive coconut plantations, degradation would be even more severe. Loss of forest cover in combination with excessive rainfall causes erosion with subsequent loss of soil fertility and stability, declining yields, danger of catastrophic man-made landslides and siltation of rivers and marine ecosystems.

1.2 Agroforestry systems

In this study agroforestry systems are understood as trees grown simultaneously with annual or perennial crops in the same parcel17. Among the numerous agroforestry systems developed worldwide through centuries, some common aspects are, that different species benefit or at least complement each other or that farmers expect higher outcomes, in the form of yields or revenues, environmental services or risk and labour minimisation. Some assets attributed to such systems are: • Soil protection through a multi-storey canopy, which reduces erosivity of rainfall (WISCHMEIER & SMITH 1978) • Reduction of microclimatic extremes under the closed canopy and maintained soil moisture favour growing conditions for shade-tolerant understorey and late- successional plants. Soil microbial biomass and biological activity can also be enhanced under such conditions (YAN ET AL. 2003; MAO ET AL. 1992; MARTIUS ET AL. 2004). • Effective resource use is facilitated through multi-storey canopies; belowground, the safety net and nutrient pump function of deep-rooting trees can prevent mobile ions from being leached (SCHROTH ET AL. 2001); nutrient cycles are kept tight. Minerals are transferred from deeper soil layers via leaf litter to the topsoil, where shallow-rooting plants can make use of them (CANNELL ET AL. 1996). Diversity of trees can foster diversity of mycorrhizae and increase strategies of nutrient utilisation; overall productivity can be increased through better resource use (HE 2005). • Diversification in plant species as a strategy to increase structural and organismic diversity can increase resilience of the system. It is expected, that self-regulation can to some extent control pest populations (SCHROTH ET AL. 2000). • Economically, peaks of labour demand can be flattened as planting, management and harvesting activities will be more spread with an increasing number of crops grown. The

17 Some authors include sequential systems; those are not considered here.

• 26 • 1 Introduction and scope

monetary risk when loosing crops or of dropping producer prices can be reduced by diversification of marketable products. (ANDERSON & DOMSCH 1985; PADOCH ET AL. 1985) • Diversification of products for home consumption can improve nutrition and health of farmer families. • Recently the potential of agroforestry for CO2 sequestration has been assessed 18 (MONTAGNINI & NAIR 2004). Under the principle of additionality , calculations can be made only in comparison to a referential land-use system (SCHROEDER 1994). Although well- managed pasture has been reported to contribute comparable amounts of organic (V. NOORDWIJK ET AL. 1997) and microbial (IZQUIERDO ET AL. 2003) carbon to the soil, aboveground biomass C accumulation will certainly be higher in agroforestry systems. Differences in agroforestry systems exist with respect to species richness and resemblance of natural ecosystems (ASHTON & DUCEY 2000). Traditional systems were often of minimal impact and close to nature. Examples are fruit trees deliberately planted along gathering and hunting trails in the forest (HECHT & POSEY 1989); enrichment planting in official forestry programmes might be adoptions of these ancient experiences. Hundreds of years ago rainforest dwellers made use of detailed knowledge of soils, species and natural succession (SALICK 1989), selected promising varieties of fruit trees and integrated them together with annual crops into natural ecosystems under consideration of small-scale soil characteristics (BALÉE 1989). A more technological approach was chosen from the 1970s onwards, when alley- and hedgerow-cropping, silvo-pastoral systems and timber plantations of fast-growing species were introduced. Especially the latter were criticised to impoverish soils (in the case of Eucalyptus plantations in South America) or erode the genetic diversity of native species through introduction of high-yielding varieties (RAO, SINGH & DAY 2000). On the other hand some authors highlight, that diversification of species should not be overemphasised as long as a set of ecological functions is maintained (RUSSELL 2002; LANGI ET AL. in GEROLD ET AL. 2004). During the last decades the traditional techniques and knowledge were rediscovered and documented (BRECKLING & BIRKENMEIER 2000) and new systems were developed and implemented (ASSOCIACIÓN DE AGRICULTURA ECOLÓGICA 1998; MILZ 2001). Many of these were based on ancient experiences of indigenous tribes. Special attention was given to natural succession (EWEL 1999; MONGELI 1999), functional groups of plants (guilds, MOLLISON 1988) and imitation of structural and species diversity of natural forests (COICAP 1999).

1.2.1 Rainforestation The Closed Canopy & High Diversity Forest Farming System was initiated by an international organisation in a development, not a research context. Consequently the term rainforestation was coined as a marketing instrument rather than a strict scientific definition. The theoretical framework of rainforestation is based on the assumption, that a system imitating the natural climax vegetation (here: dipterocarp forest) in physical structure and species composition should be the most resilient possible land-use. Basic principles are the 3-storey structure and focus on native species as well as the four guilds of lumber, fruit trees, climbers and shade-tolerant tuber crops, which are to be planted (MARGRAF & MILAN 1996). On the other hand promotion of a standardised planting pattern has been deliberately omitted; concerning planting distances and choice of species, the scheme may be modified with respect to farmers' preferences, site characteristics and availability of seedlings. Departing from the era of fast-growing 'miracle trees', Gmelina spp., Acacia mangium, Swietenia macrophylla, Eucalyptus spp. and other exotic species

18 A prerequisite for CDM projects to be accepted.

• 27 • 1 Introduction and scope were initially part of the rainforestation pattern. These were complemented with Citrus spp. for more rapid income generation (POSAS, personal communication). By the time, focus shifted more and more towards native – especially high-value Dipterocarpaceae – species and resulted in a total ban of exotic trees (MARGRAF & MILAN 1996) as these had been discovered to be less resistant to extreme climatic events (KOLB 2003) and more susceptible to numerous pests and diseases (CHOKKALINGAM ET AL. 2006). The approach was tested in various planting patterns, from loose random position to high-density straight lines of 2x2 and even 1x2m. One lesson from the early days of rainforestation was to distinguish pioneer and shade-loving trees, the latter ones being planted after the establishment phase of the first (usually in two subsequent years). Yet, even shade-loving timber trees exposed to full sunlight can attain high survival rates (QUIMIO 1998), as long as proper maintenance is guaranteed. Although starting from a forester's perspective, other guilds as fruit trees, climbers and shade tolerant tuber crops like Dioscorea sp. or Colocasia sp. were also included. From the Philippines, a variety of traditional agroforestry systems have been reported. An early scientific description of such systems was the Hanunóo example by CONKLIN (1957), a system of swidden agriculture, which integrates Cocos and Areca palms, cocoa, malay apple, jackfruit and mango trees, bamboo and abaca after a first phase of annuals and banana. The Naalad system in Cebu, based on bundles of Leucaena branches placed in the field, has helped people to reduce fallowing periods, although slow depletion of the soil cannot be totally avoided (LASCO & SUSON, unpublished). In Banaue, a more than 200 year- old farming system based on rice terraces includes forest patches of 0,05 – 5ha in size (LASCO unpublished). Cocos palm was intercropped since the 1930s with annual crops and banana and later on with fruit trees, Leucaena sp. and introduced timber species like Swietenia sp. and Gmelina sp. (BULLECER & STARK 2004). The official reforestation strategy focussed on the aforementioned as well as Eucalyptus spp. and Acacia mangium, which were sometimes alternated with fruit trees and adapted to local conditions (BUGAYONG 2004). However, constraints of this approach such as poor wood quality and retarded growth of the 'miracle trees' after the first years (own interviews with farmers) as well as poor resistance to typhoons (KOLB 2003) are frequently mentioned. Thus, rainforestation aimed at replacing the kaingin system on former fallows and releasing pressure from primary and still close-to-natural secondary forests. The annual component was intended to guarantee subsistence or even a small income, so that clearing of forest areas would not be necessary anymore. More than ten years after its initiation, rainforestation approach still has not passed the prototype stage and a critical mass has not been reached. Possible reasons for the lacking adoption through farmers are missing ownership (seedlings were given for free and weeding was carried out by project workers in many cases) and low short-term rentability or, as GROETSCHEL ET AL. (2001) put it, people cannot address environmental needs due to their short-term economic needs. Annual crops are demanding in terms of management and rainforestation terrain is usually distant from villages. Another cause for discouragement mentioned by farmers and consultants is the bureaucratic process of tree registration and obtaining a logging permit for planted trees from forestry officials. Lately, there has been a tendency to integrate abaca (Musa textilis, a fibre producer) into the rainforestation design and merge traditional abaca (under secondary forest) and rainforestation systems, thus broadening the approach from a mainly biodiversity-based to a more market-oriented one. Both components are suited for the same tenurial, land-use and ecological zones, namely slopes situated between lowland rice fields and secondary forests, which are mostly under fallow with coconut. Moreover, abaca is not a labour- intensive crop until harvest. Carbon sequestration as an additional opportunity to generate

• 28 • 1 Introduction and scope income in the short run could be another asset of abaca planting due to rapid biomass build-up and longevity of the fibres, so that carbon once it has been sequestered would not be released quickly.

1.3 Carbon sequestration

Alarming news about increasing frequency of climate extremes and underestimation of global warming can almost every day be read in the press19. Facing an increase of atmospheric CO2 from 280 to 358ppm since the pre-industrial era as well as a general trend of global warming (HOUGHTON ET AL. 2001; SCHNEIDER V. DEIMLING ET AL. 2006), concrete initiatives to reduce Greenhouse Gases (GHG) emissions have been started. Since the Kyoto Protocol entered into force in February 2005 and emission accounting is becoming a reality in many countries, binding commitments have to be met by the undersigned parties and the respective emittents of (GHG) in these countries. A pacific insular country like the Philippines is likely to be most severely hit by increasing temperatures through the consequently rising sea levels and higher rainfall variability of the Asian summer monsoon, although some models extrapolate an under average warming for the area. (HOUGHTON ET AL. 2001). Even if a small not industrialised country cannot have a large impact on the global GHG balance, climate-related compensation projects could at least mitigate some of its consequences.

1.3.1 Climate change and carbon pools

An increase of concentrations of atmospheric CO2 due to industrialisation has been 20 predicted as soon as 1895 by S. ARRHENIUS . At the current state of knowledge, it is still difficult to quantitatively separate and attribute shares of the greenhouse effect to certain anthropogenic drivers like transport or land use change; at least there is now a broad consensus, that the sum of effects is too high for a natural phenomenon (IPCC 2003). Many of the mechanisms21 caused by elevated atmospheric GHG concentrations are understood, but orders of magnitude, sometimes even directions of change, are still difficult to predict. The Intergovernmental Panel on Climatic Change (IPCC) listed the factors contributing to climate change and their warming / cooling impact on the atmosphere relative to the pre- industrial era (fig.17).

19 Examples: 1990s were the warmest decade of the millenium and 1998 was the warmest year for the Northern Hemisphere (HOUGHTON ET AL. 2001). The six warmest summers during history of measurements in Germany occurred in the last 20 years (DEUTSCHER WETTERDIENST). Greenland Ice is melting faster than expected (VELICOGNA & WAHR 2006) 20 Der Anstieg des CO2 wird zukünftigen Menschen erlauben, unter einem wärmeren Himmel zu leben. [In the future, increase of CO2 will allow man to live under warmer skies.] 21 Like warming, increased plant growth and decomposition of SOM, rise of sea level (expansion and melting), changes in the thermohaline circulation (Atlantic).

• 29 • 1 Introduction and scope

Figure 17: Relative influence of different factors on climate change and scientific understanding of mechanisms. Error bars indicate range given in the compiled studies. Source IPCC (HOUGHTON ET AL. 2001) An important share of radiative forcing has been attributed to carbon dioxide, methane and nitrous oxide, which are released from natural as well as anthropogenic processes, and from a range of man-made halogenated hydrocarbons. Carbon dioxide is released from respiration and burning processes, while CH4 is set free under anaerobic conditions e.g. in soils or rumina and N2O through denitrification in soils, under anaerobic conditions and depending on N supply and temperatures. BARETH (2000) gives an overview of the relative shares of the main GHG and man-made drivers with respect to climate change (fig.18-20).

Figure 18: Shares of natural GHG to Figure 19: Shares of anthropogenic GHG warming (after data from BARETH 2000) to warming (after data from BARETH 2000)

• 30 • 1 Introduction and scope

As lifetimes and Global Warming Potential (GWP) differ between GHG (table 4), these are usually expressed as CO2 equivalents for calculations. Although CO2 has the lowest GWP of the three gases, it exerts the strongest impact on climate due to its relatively high concentrations in the atmosphere. Among the various pools of CO2, terrestrial biomass has been a source (LLOYD 1999) of CO2 or at least neutral (SCHIMEL ET AL. 2001) for a long time, but is currently a sink22. This change has been attributed to higher biomass productivity through CO2 fertilisation, reducing stomatal conductance and thus enhancing water, light and nitrogen use Figure 20: Anthropogenic causes of global warming efficiency, as well as anthropogenic (after data from BARETH 2000) nitrogen depositions (GIFFORD ET AL. 2000; HOUGHTON ET AL. 2001). Expansion of Northern Hemisphere forests and large-scale land use change in the Tropics like natural regeneration of abandoned lands and fire prevention have also been mentioned as explanations for the increased absorption of CO2 through terrestrial biomass (SCHIMEL ET AL. 2001). The feedback of elevated temperatures on decomposition rates and thus release of CO2 from litter and soils into the atmosphere has been highlighted by POWLSON (2005).

Table 4: Lifetime and Global Warming Potential (GWP) of GHG. GWP projected to 20a residence time in the atmosphere

GHG Lifetime GWP [a] during 20a

CO2 5-200 1

CH4 12 62

N2O 114 275

An overview of the different carbon pools (based on data presented by WATSON at IPCC COP 6 2001) is given in fig. 21, where red arrows indicate the main anthropogenic CO2 sources, namely combustion of fossil fuels and tropical deforestation, which lead to an atmospheric build-up of CO2 in the atmosphere. Oceans are also sinks, but will not be discussed here in detail.

22 Sources release GHG into the atmosphere, while sinks absorb them.

• 31 • 1 Introduction and scope

Figure 21: Global carbon pools during the 1990s (data from WATSON, presentation at IPCC CoP 6, 2001)

-1 The existence of a residual terrestrial sink of 2-4 PgC a (SCHIMEL ET AL. 2001) has been concluded from the overall carbon balance. The eminent role of land use change (LUC) in the tropics follows from the aforementioned change of the terrestrial biomass from source to sink. Natural and regenerated forests are seen as key factors to climate stabilisation. Higher productivity of forests through elevated CO2 levels could lead to increasing carbon storage even at a new productivity equilibrium (CHAMBERS ET AL. 2001). Of similar importance is the storage of carbon in soils (SCHWENDENMANN 2002), which account for 30% of C in tropical forest ecosystems (MOURA- COSTA 1996 as cited by LASCO ET AL. 2004). Especially the passive carbon pool, stabilised by minerals, and also charcoal-C are practically immobilised from the cycle, but transferring carbon from the atmosphere into the passive pool takes long times (CHAMBERS ET AL. 2001). Although the importance of carbon in soils has been demonstrated and widely accepted (v. NOORDWIJK ET AL. 1997; GUO & GIFFORD 2002; MURTY ET AL. 2002), carbon budgets of soils are often neglected or assumed to remain stable for accounting purposes (e.g. UNFCCC procedures). Besides soil carbon balances, CH4 and N2O pools and fluxes in agroforestry (VERCHOT ET AL. 2004), permanence of carbon in pools, saturation of carbon pools, separability and attribution of fluxes (IPCC 2001) are often mentioned as future research needs.

1.3.2 Institutional and legal framework for funding of carbon sequestration through afforestation and reforestation Most countries worldwide have joined the UN Framework Convention on Climatic Change (UNFCCC). Since February 2005 the Kyoto Protocol as legally binding treaty took effect once more than 55 Parties to the Convention, responsible for more than 55% of Annex I23 countries' CO2 emissions in 1990, had acceded. The Kyoto Protocol obliges all undersigned parties to meet fixed targets reducing their 23 This group includes most industrialised nations, their share of global GHG emissions amounts to 61.6% (UNFCCC 2006 ).

• 32 • 1 Introduction and scope greenhouse gas (GHG) emissions. The overall objective is to cut global GHG emissions to a level of at least 5% below those of the baseline year 1990 within the commitment period 2008-2012. Developing countries and such in transition to market economy are (partially) exempted, but as a principle the right to pollute is intended to be equally distributed and to be traded. Three instruments were designed to implement polluter-pays-principle, market mechanisms and a theoretical overall control of total GHG emissions: Joint Implementation rewards GHG reduction or removal projects implemented by Annex I parties with Emission Reduction Units. Emission Trading allows participants of Annex I countries with Assigned Amount Units (AAU) to buy CO2 equivalents in order to meet their emission commitments or to sell them and benefit from reduced GHG emissions. Finally, Clean Development Mechanism (CDM) involves partnership between developed and developing countries in such a way, that participants from the first group can implement GHG reduction or removal projects in developing countries. CDM credits can be acquired and saved before the beginning of the first commitment period in 2008. For this reason initiatives like the Prototype Carbon Fund (initiated as a PPP project by the World Bank with the participation of several government and companies) have already shown interest to realise CDM projects. Potentials of CDM measures in context with rentability have been investigated by BEERBAUM (2001). According to Article 12 of the Kyoto Protocol, purpose of CDM is to assist non-Annex I parties in sustainable development and Annex I countries to meet their emission reduction targets. CDM projects have to generate real, measurable and long-term benefits related to the mitigation of climate change; these benefits must be additional to a business as usual- baseline without CDM activities. The Philippines are a Party to the UNFCCC and have nominated a Designated National Authority (DNA) for CDM issues, the Dept. of Environment and Natural resources (DENR). Thus Philippine projects are eligible for CDM. Briefly, after receiving a Project Design Document including a socio-economic and environmental impact assessment, Designated Operational Entities (DOE) will validate the plan and, if successful, propose it to the CDM executive board for registration. Project financing can be obtained from any entity interested to buy Certificates of Emission Reduction (CER). During the project cycle, emission reductions are monitored and emission reductions are verified. Finally, the project is certified and CER are issued (UNFCCC CoP9 2003). Project lifecycles can last 30 or three times 20 years. To assure permanence of the GHG reductions, two kinds of CER are issued: temporary or tCER expire at the end of the commitment period following the one of issuance and long-term or lCER are valid during the entire project lifecycle (UNFCCC CoP9, 2003). The principal activities of GHG reduction are conservation or expansion of carbon stocks and substitution of techniques and processes based on fossil fuels by renewables. Most CDM project activities are currently concentrated in the substitution category, namely in the energy industries (50%, including renewables) and waste management (24%) sectors. Agriculture covers a relatively small share (16%), many projects in this sector are related to biomass power plants and CH4 reduction. Within the category of expanded sinks, agricultural projects are excluded, only afforestation and reforestation24 are supported. Currently, CER of 1bn tCO2 equivalents are expected to be generated on the basis of existing CDM projects until 2012 (press release UNFCCC secretariat 09/06/06). 306 projects were registered Sep 24, 2006, about half of them in the Asian and Pacific region and one in the Philippines (another one has been submitted for registration25). Most of this CO2 reduction will be accredited in China (45%), Brazil (16%), South Korea (13%) and 24 A&D

• 33 • 1 Introduction and scope

India (12%). The most important Annex I countries participating in CDM projects are the Netherlands (68 projects), UK (52), Japan (24), Spain (11) and Canada (11).

1.3.3 Potential of carbon sequestration through agroforestry There are various agricultural and forestry practices suitable to reduce carbon emissions, like low-impact logging techniques, soil carbon management (reduced tillage, among others), fire prevention or fertilising / liming of forests, which shall not be discussed in depth here. Agroforestry is generally eligible for CDM projects if the respective meets the national requirements of forests (mainly coverage and height, s. below). Respective feasibility and rentability studies have been carried out for several tropical countries (e.g. GINOGA ET AL. 2005, SHIVELY ET AL. 2004). -1 SCHROEDER (1994) calculated potential stocks of 50MgC ha and sequestration rates of 10MgC ha-1 a-1 during a cycle of 30 years for agroforestry in humid ecosystems. Assuming 160mn ha of land to be available for agroforestry in the tropics, he concludes, that 1.5 to 8 Pg C could be sequestered annually. DIXON (1995) gives even larger areas of 585-1275mn ha worldwide that could be planted to agroforestry. ALBRECHT & KANDJI (2003) estimate that agroforestry systems worldwide could sequester 10-15% of globally emitted CO2. They give a first approximation of 12-228MgC ha-126 sequestration potential for Southeast Asian Agrosilvicultural systems. Finally, IPCC sees a potential of C sequestration by agroforestry -1 systems of almost 0.4PgC a (R. WATSON, IPCC, presentation during CoP 6 meeting). Agroforestry has a long history in the Philippines and even been part of the state's land rehabilitation programmes (CHOKKALINGAM ET AL. 2006), so that there should be sufficient knowledge and experience to establish agroforestry projects under the umbrella of CDM. Potentials of such projects have been evaluated (LASCO ET AL. 2005B) and proposals are under way. As mentioned, agroforestry projects fall under the category of Afforestation or Reforestation (A/R). However, the definition of forest, on which the applicability of A/R projects depends, leaves some flexibility to the host countries: A forest presumes a minimum area of 0.05 to 1ha, at least 10-30% of canopy coverage and 2-5m potential height of the trees27. This might lead to difficulties e.g. in the case of existing coconut stands: If the crown cover exceeded the national forest threshold, the stand would be considered a forest and as such could not be reforested anymore. Afforestation and reforestation according to CDM regulations imply, that the respective area has been deforested at least since Dec 31st, 1989. This regulation may pose a problem to countries, where a large share of deforestation occurred since the 1990s, which is not the case for the Philippines. For A/R in rural areas, small projects would probably best serve development strategies and couple livelihood and environmental goals. So-called small-scale (SSC) projects have -1 been included to CDM and defined as those sequestering less than 8000t CO2 a and involving low-income groups (UNFCCC CoP9 2003). Efforts have been undertaken to simplify procedures for SSC, e.g. through eligibility of bundled projects. The Intergovernmental Panel on Climatic Change (IPCC), a consultative body to the UNFCCC, has elaborated guidelines for validation and monitoring of CDM projects, the so-called Good Practice Guidelines – Land-Use, Land-Use Change and Forestry (GPG-LULUCF). These guidelines give detailed procedures and methodologies to determine carbon pools

25 Reductions of 96,000t CO2 equ. are expected annually for a waste water treatment plant, 57,000t CO2 for the NorthWind Bangui Bay Project. 26 Over a lifecycle of 50 years. 27 Detailed prerequisites on eligibility of land for A/R projects can be found in Annex 16 of the CDM – Executive Board's report 22

• 34 • 1 Introduction and scope in ecosystems and rates of change caused by LUC28. CDM regulations are based on these guidelines. To apply for SSC-AR projects, the following criteria are validated: The project boundary must be clearly delimited and a stratification, i.e. mapping of land-use, soil and vegetation, has to be available. Baseline scenarios per stratum have been determined and baseline carbon stocks evaluated. An ex-ante GHG removal calculation per stratum has been performed (only trees are taken into account, soil and undergrowth remain constant) and additionality29 proven. Reductions for elevated transport, fertiliser, burning, tillage or other 30 CO2-releasing management (leakage ) have been discounted and ex-ante actual net GHG removal calculated. Project implementation, stratification, ex-post GHG removals and leakage will be monitored during the project cycle as described under 1.3.2. According to PULHIN ET AL. (in CHOKKALINGAM ET AL. 2006), the Forest Management Bureau identified a rehabilitation target of 5.5mn ha for the Philippines. LASCO ET AL. (2004) estimate that there is a potential for CDM-LUCF projects of 2-10mn ha degraded grasslands in the Philippines. On the basis of data published by NAMRIA (2003; s. section 1.1.6) for Leyte island, roughly 270,000ha of non-forested forest land could be theoretically available for CDM projects. Subtracting the area under coconut (a conservative approach in case these would not match A/R requirements), 94,000ha of grasslands would still be potentially available. These areas could additionally be restricted by the national legislation, which generally does not permit agriculture on forest land. On the other hand Agroforestry Farm License Agreements with a validity of 25 years are issued by the DENR-FMB. Legal aspects on this matter (like Certificates of Stewardship for 25a, if the respective land has been cultivated for at least 25a) have been discussed by BUGAYONG (2004). At least two reforestation projects in the Philippines are being proposed to the CDM. One of them is the LLDA31–Tanay Streambank Rehabilitation Project in Luzon. The LLDA and the municipality of Tanay will sell certificates to the World Bank's BioCarbon Fund. The certificates are planned to originate from streambank rehabilitation, 70ha of reforestation of denuded lands, and 25ha of agroforestry implemented by local farmers. 1400 to 3200t C are expected to be fixed within the project cycle from 2004-14 (LASCO ET AL. 2005A). One of the most important uncertainty when assessing prospects of agroforestry for CDM projects is still carbon sequestration itself. As LASCO & PULHIN (2004) state, there are still limited data on carbon sequestration compared to carbon stocks [...] because carbon stocks can be easily calculated using allometric equations [...], but [...] biomass change and carbon sequestration requires long-term monitoring. This study will use data for carbon sequestration acquired from modelling and calibrated with data from field experiments with different timber and fruit tree species on a typical site in Leyte and relate the findings to those of other authors. Opportunities and obstacles shall be assessed from the perspective of AR-SSC applicants to the CDM.

28 e.g. Crop → crop, Grassland → grassland, Crop → Forest and Grassland → Forest 29 Benefits with regard to carbon sequestration, that would not have occurred without the project (baseline scenario). 30 Defined as decrease or increase in greenhouse gas benefits outside the project’s accounting boundary as a result of project activities, leakage is commonly understood as decrease, i.e. CO2 offsets into the atmosphere. 31 Laguna Lake Development Authority

• 35 • 2 Sites, material and methods

2 Sites, material and methods

2.1 Approach

Variability was a key challenge to this study: Apart from relief and soil variability on a small scale, the existing rainforestation plots generally were highly diversified not only in tree species, but also in planting distances and schemes, plot size and management. For this reason, two stages of plots were monitored to describe growth and carbon sequestration potential of rainforestation systems on the one hand and effects of the trees on soils on the other. Firstly, a 1ha rainforestation plot with a regular planting design was established in 2004 to document the departure point of the system, also as a baseline for future studies. This model plot focused on early yielding species and served to monitor growth and mortality during the first years, which are critical for farmers' motivation, weeding costs and survival of plants. Heterogeneity of the plot with respect to canopy closure and soil was assessed during the first field phase to explain spatial differences in plant performance. Biomass data were later used to validate modelling results. Secondly, in 2005 soil from different rainforestation plots planted 1992-1996 was sampled as paired plot comparisons each with an adjacent reference area under traditional coconut-bush fallow or coconut-grassland use. For two sites, reforestation with fast growing exotic Gmelina arborea was available as a third land-use alternative representing the official reforestation strategy. The paired-plot approach was chosen in order to detect possible effects of rainforestation systems on soils. Aside from a general characterisation and classification of soils (FAO), parameters were selected that respond to land-use changes within a short- to mid-term period. Besides SOM-related parameters, analyses reflecting microbial activity and litter dynamics were of most interest. Third, a modelling approach was chosen to extrapolate future biomass growth and carbon sequestration potential from the new plots and also to compare C-sequestration under traditional fallow, Gmelina and native trees. For tree parametrisation, which requires mature trees, measurements were taken on individuals from the >10 year-old rainforestation sites, forest sites and well established orchards. For validation an inventory conducted by KOLB (2003) in 2002 was used as a reference to modelling predictions. For abaca parametrisation, KELLMAN'S (1970) allometric equation and V. NOORDWIJK ET AL. (2002) gave some trends, yet destructive measurements were carried out on the plots established in 2004.

2.2 Site selection for paired plots

Sites for chronosequence comparisons were selected out of the 28 existing Rainforestation Farms in Leyte, about half of which fitted the minimum criteria of an area ≥0.25ha, successful in a broad sense (i.e. not burnt, cleared or abandoned) and existence of an adjacent reference area (grassland or fallow, usually with coconut). Out of these, seven sites were selected. The selected rainforestation sites are shown in fig. 22; most of them are situated on volcanic soils (formed from volcanic ashes or basaltic parent material; ASIO, personal communication) with the exception of a limestone soil in Punta. All plots had been installed between 1993-96 on medium- to high-gradient hills, which were used mainly for coconut production or fallowed before. Profile pits and sampling were mostly on the middle slope, with one exception (plateau, see table 5).

• 36 • 2 Sites, material and methods

Information on land use history of the plantations and reference areas was obtained from interviews with the landowners. Tenurial status, land-use history, agricultural practices and particular information on plants and soil where recorded during interviews with farmers at the time of the first field visit and completed through semi-structured interviews using questionnaires read and filled by the interviewer. Oral history normally goes back until adults' grandfathers' generation, which coincides with Japanese occupation during the Second World War and, sometimes, first clearing of the land. Important incidents such as land reform under Marcos (1972), the Comprehensive Agrarian Reform under C. Aquino (1988) or the Ormoc flashflood disaster (1991) served as reference years for oral history reports.

Figure 22: Locations of research sites

General information on the sites is given in table 5. Details are shown under results in section 3. Table 5: General information on research sites

# Location Planted Exposition Slope Size Parent Soil unit (FAO) [year] [%] [ha] material 1 Cienda Demo 1996 S 3-5 1 Volcanic Dystric Nitisol 2 Patag 1993-4 WNW 75 1 Volcanic Stagnic Cambisol 3 Marcos 1995 W 60 0.3 Volcanic Ferri-stagnic Luvisol 4 LSU 1993-5 SW 70 2 Volcanic Chromic Cambisol 5 Pangasugan 1996-97 WSW 40 0.6 Volcanic Ferri-chromic Luvisol 6 Maitum 'Early 1990s' NW 15-75 0.5 Volcanic Hypereutric Cambisol 7 Punta 'Early 1990s' WNW 50-150 5.4 Limestone Calcari-mollic Leptosol

• 37 • 2 Sites, material and methods

2.3 Land-use history and plot installation in Cienda

Site selection criteria for the new system to be planted were low small-scale variability (as far as possible), comparability with the old rainforestation sites with respect to soil, exposition and slope, as well as a motivated owner experienced with establishment and maintenance of reforestation sites and willing to cede an area sufficient in size to a research experiment for long-term observations. Cienda San Vicente Farmers Association (CSVFA) seemed to fit the latter criteria and a Memorandum of Agreement on cooperation was signed. A plot with homogeneous planting lay-out was then planned and established from March-April 2004 in Cienda as a participatory research project. Plot position is N10°43'52.1'', E124°48'43.2'', about 300m distant from the plot installed in 1996, and size about one hectare. The area was completely cleared for the first time in 1950 and has since been planted to coconut and banana. In 1971 the present owners inherited the land and divided it along slope direction into two properties. On the northward side, land-use was continued as before; fertiliser or pesticides were not used. Banana is mainly grown in the centre part of the plot, a large fraction infested by Moko, a bacterial disease (Rolstonia solanacearum). Coconut was planted irregularly in time and space with an average distance of 8-10m. Usually, small amounts of coconut husks are burned to repel mosquitoes during harvest time, only once a wildfire struck the entire plot during a dry summer in the late 1980s. The lower part was sporadically browsed by a neighbour's water buffalo, it is characterised by water-logging in one part, indicated by Wedelia biflora, and Imperata sp. and ferns in the drier parts. In small strips along the lower boundary as well as on some patches in the centre cassava and sweet potato were planted in 2003. Extraction is minimal; three times per year 500-1000kg copra and 30kg bananas monthly are harvested, tubers were negligible. Apart from this, villagers take advantage of the remote situation and extract firewood and coconuts. The southward half has been managed even more extensively if at all. The area is covered by 7-12m high forest regeneration, mainly Ficus spp., in an early successional stage. Scattered bananas can be found in the plot centre and coconut palms are planted in similar distances as on the other half. With respect to land-use, this side is the more homogeneous. Main differences between both sides are canopy cover (denser in the southern half) and water-logging in the lower part of the northern half. Slope is similar on both sides (20-30%).

2.3.1 Species and planting material Aiming at a planting scheme comparable to the original rainforestation idea – i.e. with a focus on indigenous timber species – but with more emphasis on early yielding fruit trees, four timber and six fruit tree species were selected. Characteristics of these species as far as relevant for model parametrisation and performance are compiled below32. • Dipterocarpus validus Blume syn. Dipterocarpus warburgii Brandis, Dipterocarpaceae: Hagakhak or Apitong Hagakhak is a slender tree (with canopy diameter in solitary stand up to 10m; FERNANDEZ, pers. comm.) with straight bole of up to 50m height (SOERIANEGARA ET AL. 1993). Branches grow horizontally to downward, leaves are entire, oblong-elliptical, 7.5-12cm x 12-30cm and brownish when young. Wood specific gravity is 612-740kg/m3 (SOERIANEGARA ET AL. 1993). Flowering starts at about 10 years age (own observation) from Mar-May, fruits are mature in Sep-Oct (FERNANDEZ, pers. comm.). Hagakhak prefers half- shaded conditions during the first years and can survive even under full shade, but like

32 Not all synonyms are mentioned. Wood specific gravity refers to moisture contents of 15%.

• 38 • 2 Sites, material and methods most dipterocarps will grow at a maximum rate only under higher light intensity (MARGRAF & MILAN 1996). According to WHITFORD (1911) and LANGENBERGER (2003), hagakhak was the dominating species of the Lauan-Hagakhak type forest, which extended from well drained sea-level areas next to the beach forest habitat up to approx. 150m asl., especially along rivers and on gravel. The mostly evergreen species is adapted to a short or no dry season and tolerates short periods of flooding. Other species adapted to the same ecosystem are Toona calantas (s. below), Dracontomelon dao (Dao), Terminalia microcarpa (Kalumpit); erect palms, lianas and small trees are also typical members of the Lauan-Hagakhak forest. Hagakhak seedlings in PE bags of approximately one year age were provided in good shape by the LSU tree nursery. Average height at the time of the first inventory shortly after planting (June 2004) was 29cm. • Shorea palosapis (Blanco) Merr. syn. S. squamata (Turcz.) Benth. & Hook. f. ex A. DC., Dipterocarpaceae Palosapis or Mayapis is a fast-growing dipterocarp of up to 50m height with straight bole. Leaves measure 12-24 x 8-11cm (SOERIANEGARA ET AL. 1993), are entire, light green and show a characteristic vein pattern. Flowering season is Mar-May, fruits mature from Oct- Nov; first flowering and fruiting were observed at an age of 10a (FERNANDEZ, pers. comm.). 3 Wood density has been reported to be 310-578kg/m (SOERIANEGARA ET AL. 1993), or slightly above (CABI TREE COMPENDIUM 2002). Under the group name of light red Philippine mahogany, S. palosapis as a hardwood is used for construction, posts, furniture, containers and plywood, but has been over-exploited in its natural habitat. In the Philippines, palosapis grows naturally from sea level up to 300m asl. The species is adapted to uniform rainfall pattern, but tolerates a dry season of up to 3 months. Requirements to soil seem to be medium, but free drainage considered to be important (CABI TREE COMPENDIUM 2002). Planting material was obtained from LSU tree nursery; average height was 29cm and plants were in good condition. • Shorea contorta S. Vidal syn. Pentacme contorta (S. Vidal) Merr. & Rolfe, Dipterocarpaceae White Lauan is a large tree of up to 50m height with conical canopy of 15-25m diameter (FERNANDEZ; LANGENBERGER pers. comm.). Leaves are ovate-lanceolate (cordate) and 9-29 x 5.5-11cm in area. Flowers appear from Mar-May and develop into mature fruits until Sep- Oct. First flowering has been observed by FERNANDEZ after 11 years. Wood density is between 420 and 590kg/m3 (ICRAF online wood density data base33). White Lauan is representative of Yakal – Lauan and Lauan type forests, relicts of which were still found on a plateau area with deeply weathered soil such as under the Cienda demo farm (LANGENBERGER 2003). Lauan forest species are best adapted to climates without or with only short dry periods, but for white lauan up to 5 months of dry periods are reported to be tolerable (CABI TREE COMPENDIUM 2002). With respect to soils, white lauan is not exigent; they may be shallow, but should be free draining. Under the trade name of Philippine mahogany, white lauan is a highly demanded hardwood timber used for construction, furniture, boats, plywood and pulp. Planting material was bought from a farmers' cooperative in Patag; some plants were apparently weak and had been uprooted just before being potted in PE bags for sale. Coefficient of variance with respect to height and stem diameter was exceptionally high (43cm and 37%, respectively) indicating great heterogeneity of the planting material. The lowest individual measured only 18cm, the tallest 75cm (average 36cm). 33 www.worldagroforestry.org

• 39 • 2 Sites, material and methods

• Toona calantas Merr. & Rolfe syn. Cedrela calantas (Merr. & Rolfe) Burkill, Meliaceae Kalantas is a fast growing species common for Lauan-Hagakhak type forests; KOLB (2003) characterises kalantas as a pioneer tree. The species was recommended for reforestation purposes in the Philippines as early as 1916 (PULHIN ET AL. in CHOKKALINGAM ET AL. 2006) . The leaves are odd pinnate and up to 50cm long, each dark green leaflet measuring approx. 5-10cm. The tree forms a straight bole of about 25m, up to 20m branchless (SOERIANEGARA ET AL. 1993) and with a crown diameter of up to 10m. Flowering starts at an age of 6-10 years (FERNANDEZ; LANGENBERGER, pers. comm.) from Mar-May and fruits mature until Aug-Sep. Wood density is 365kg/m3. Seedlings obtained from LSU tree nursery were in good shape and measured 41cm in average. • Nephelium lappaceum L., Sapindaceae Rambutan is a small (REHM & ESPIG 1996) fruit tree of 12-25m height (FAO 1982), cloned material often only 4-7m (VERHEIJ & CORONEL 1991), with a dense, round canopy (SCHÜTT ET AL. 2004). Leaves are petioled, alternate or subopposite paripinnate, up to 6-jugate (VERHEIJ & CORONEL 1991) with obovate (to ovate) leaflets, which turn from an early th th brownish stage to dull dark green (SCHÜTT ET AL. 2004). Flowering starts from the 5 to 6 year (for budded plants earlier) and takes place during dry season (VERHEIJ & CORONEL 1991), the fruit ripening within 90-150d later (SCHÜTT ET AL. 2004). The fruit is bright red, ellipsoidal in shape, measures 3-4 x 6cm and bears soft hooks on the outside. The white arils are edible and commercialised, the seed contains about 30-45% fat and can be 3 eaten roasted (FAO 1982). Wood density is 1001kg/m (SCHÜTT ET AL. 2004). In the Malay Archipelago as its area of origin (FAO 1982) and also in Thailand, rambutan production is highly commercial and many different varieties have been bred. Most of the improved varieties are grafted. Yields range from 25-30kg/tree (own estimation after VERHEIJ & CORONEL 1991 and interviews with farmers) to 48-120kg/tree (SCHÜTT ET AL. 2004, data from the Philippines). Figures for maximum production under optimal conditions are th th given by VERHEIJ & CORONEL (1991) with up to 170kg/tree from the 8 to 10 year and even 250-300kg/tree have been reported (FAO 1982). Rambutan has been classified as medium-sized gap species, which tolerates light shade. However, in orchards and plantations it is usually planted under open conditions. As ecological optimum for rambutan, SCHÜTT ET AL. (2004) give the example of Mindoro Oriental with 1800mm evenly distributed annual rainfall, 27.3°C annual mean air temperature and 82% relative humidity. A dry season > 3 months as well as strong wind during flowering are noxious. According to FAO (1982), infertile soils are tolerated as long as rooting depth is sufficient. Grafted planting material was obtained from a commercial tree nursery specialised on import of fruit trees from Mindanao. Plants were strong and already overgrown in 1l PE bags. Initial height was 154cm in average, habitus was not balanced. Trees appeared as if kept in very dense stand in the nursery. After planting out many plants bent down and a new sapling from below took the lead. • Durio zibethinus Murray, Bombacaceae Durian is a tall tree (up to 35m) with straight bole, conical and dense crown formed by mainly horizontal branches (FAO 1982). Crown diameter is estimated approx. 15m. Leave position is alternate to subopposite, they are petioled, simple, entire, ellipsoid-lanceolate and acuminate. The upper surface colour varies among sites, the lower side is silver to brownish/copper. Flower is induced by dry season and two flowering periods per year are reported (FAO 1982). Yellow to pale brown round to ellipsoidal spiny fruits form 90-130 days after flowering (VERHEIJ & CORONEL 1991). They weigh 2-3kg and are approx. 20-35 x

• 40 • 2 Sites, material and methods

3 17cm in size. Wood density is 540 (FAO 1982) to 690kg/m (SCHÜTT ET AL. 2004), the timber being used for light construction. Durian is native to Malaysia where it grows naturally on forest margins up to 300m asl. It prefers half-shaded conditions, can survive under full shade, but grows only, under higher light intensity. The fruit is said to be promising with respect to commercialisation and many improved varieties have been developed. Yields start from the 5th to 7th year after planting (FAO 1982) and amount to 50 fruits/tree (FAO 1982; VERHEIJ & CORONEL 1991). Grafted plants from Mindanao were bought at a commercial nursery. Seedlings were uniform (average height 86cm) and in excellent shape. • Garcinia mangostana L., Clusiaceae (=Guttiferae) Mangosteen is a relatively small tree, 6-25m (PROSEA) tall, but in cultivation mostly smaller than 10m and 0.25-0.35m stem diameter (FAO 1982). Perfectly symmetrical growth and conical crown are characteristic for its architecture. Like branches, leaves are always opposite; they are petioled and elliptic to oblong in shape, measuring 12-23 x 4.5-10cm. Due to its poorly developed root system (FAO 1995), the plant is said not to be an efficient nutrient acquirer, which would be an explanation for the very slow development. Flowering is induced by a dry period. Only from the 10th to 15th year after planting, production of fruits starts, maximum yield can be expected after 20 years. The fruit is a purple berry of 5-7cm diameter (FAO 1982) containing white edible arils and a seed, which is edible when roasted. Wood is heavy, its density just below 1000kg/m3 (FAO). G. mangostana is high in demand in Thailand and Malaysia, where most progress in plant breeding has been achieved. Seeds or cuttings are used for propagation; grafting on other Garcinia species is common. As an average fruit yield for the Philippines, 1.8-2t/ha have been reported (FAO 1982). Mangosteen stems from the Sunda Islands and Malay Peninsula (FAO 1982). It grows well in clayey unless water-logging soils. As an understorey species, mangosteen still performs well under up to 50% shade; shelter from wind and sun is necessary. Plants were bought from a commercial tree nursery. Material was uniform and healthy, average was 30cm. Seedlings had been kept in a strongly shaded bed and already suffered during the short accommodation to a brighter place before transplanting. • Lansium domesticum Corr., Meliaceae: Lansones is a short-trunked tree of 10-15m height. Leaves are pinnate, 22.5-50 cm long, with 5-7 alternate, elliptic-oblong leaflets. Fruits are greyish-yellow, 2.5-5 cm in diameter and are borne in clusters. Fruiting starts after 12-20years, if trees are grown from seeds. (ICRAF online AgroforesTree database)Wood density ranges from 750-920kg/m3 (ICRAF online wood density data base). Lansones is native to Malaysia and common throughout Southeast Asia. The natural habitat is limited to areas lower than 700m asl and without pronounced dry periods. Lansones grow best under half-shade on humic and free draining soils and are highly sensitive to water-logging. Propagation is by seed, whereby viability lasts few days only. Grafting is common. Average yields in the Philippines are around 1000 fruits per tree (ICRAF online AgroforesTree database). Planting materials of about one year age were purchased at a commercial tree nursery. Plants were not too uniform (average height 46cm, span 6-62cm) and yellow leaves indicated a suboptimal health status. • Artocarpus heterophyllus Lam., syn. Artocarpus philippensis Lam., Moraceae: Jackfruit is a medium-sized evergreen tree of 15m height and trunk diameter of 0.3-0.5m. Leaves are alternate, stipulate, entire and of elliptical shape. They measure 10-20 x

• 41 • 2 Sites, material and methods

3-12cm. The tree is cauliflor, flowers being formed from the 2nd - 6th year after planting (BRACK EGG, 1999). Individual fruits can weigh 10-30kg. Maximum production is reached th th from 8 to 15 year onward, attaining 12-100 fruits (BRACK EGG 1999) or 250-750kg/tree/a. The wood is used for construction, furniture, handles and dye. Specific gravity is 420-700kg/m3 (ICRAF online wood density data base). Many varieties are known. Jackfruit is a lowland species native to India (CABI Tree Compendium) and characterised as medium-sized gap species. Seedlings tolerate light shade but grow best in full sunlight. In the rainforestation inventories of KOLB (2003), Jackfruit was the fastest growing fruit tree. Shallow soils are tolerated unless they are water-logging (FAO 1982). Healthy and very uniform grafted plants (average height 63cm) were purchased at a government-operated nursery in Balinsasayao, approx. 30km SE from LSU. • Artocarpus odoratissimus Blanco, Moraceae (synonymously A. odoratissima): Marang is an evergreen tree, up to 25m tall. Leaves are broadly elliptic or rhombic to obovate, 16-50 cm x 11-28 cm, with the upper half often 3-lobed; both surfaces are hairy. Fruits are subglobose, up to 16 cm x 13 cm, green-yellow, their flesh is white, juicy and fragrant. (ICRAF online AgroforesTree database). Fruiting season is from Aug-Dec (Mindanao) and flowering starts from the 4th to 6th year. Wood density is 580-780kg/m3 (ICRAF online wood density data base). Fruits are said to be superior in taste to jackfruit, but harvest is difficult and shelf life short. Seeds are edible when cooked. Marang stems from Borneo, but is widely cultivated in the Philippines. Its natural habitat are partly-shaded places in secondary forests from 0-800m asl in regions with evenly distributed rainfall. The tree grows best on rich loamy, well drained soils. Propagation by seed is easy with high germination rates, growth is fast and major pests and diseases have not been observed. Seedlings were uprooted wildlings from the rainforestation demo site in Cienda and brought by a CSVFA member. Age at planting time was estimated about 6 months, initial height was 15cm in average with high heterogeneity between individuals (coefficient of variance 35%). • Musa textilis Nee, Musaceae Abaca is botanically closely related to banana and similar in habitus. In contrast, abaca pseudostems are more slender and often reddish, the leaves are narrower, not rounded at the end and show a dark line on the upper leaf surface. Abaca is a typical shade-tolerant understorey plant. It is traditionally planted either under existing forest canopy or along forest margins, but can be cultivated in pure stands in open areas, if continuous weed management and soil cover are provided. Planting is best at the onset of the wetter season (for Leyte Jul-Aug) as evenly distributed rainfall is crucial for abaca. With respect to soil, the species is not demanding, but volcanic soil rich in humic matter are preferred (FIDA extension material, not dated). Propagation can theoretically be by seed or tissue culture, but cormus propagation (suckers) is most common. First fibre harvest is after 18-24 months and then every 3-4 months. The fruits are not edible. Most relevant diseases for abaca are viruses like bunchy top and abaca mosaic, which are transmitted by aphids like Pentalonia nigronervosa among others. Besides, Fusarium oxysporum and bacterial wilt (Ralstonia solanacearum) are relevant (BORINES & PALERMO 2006). Abaca or manila hemp leaf sheaths render a very resistant fibre used for cordage, pulp or craftswork made of fibre (BROWN 1919). The plant is expected to have some economic potential on the world market and efforts are made to improve properties, processing and cultivation (e.g. GUARTE & SINON 2005). The Visayas and Mindanao are the most important regions of abaca production in the Philippines, which are the only country worldwide to

• 42 • 2 Sites, material and methods produce considerable quantities of abaca fibres (minor production exists in Ecuador). Within the Visayas, Southern Leyte is an important centre of abaca production. For Leyte, abaca is the second most important agricultural product for export after copra. Different traditional and some improved varieties are available, the most popular among farmers in Baybay being Inosa. Other varieties are Laylay (promoted by the National Abaca Research Centre, NARC, at LSU), Musatex 80, 81 and 82 and Minenonga (recommended by FIDA) as well as local traditional varieties. Planting material of Laylay and Inosa varieties was purchased from tissue culture produced at NARC. The decision to use this kind of planting material was rather because of missing alternatives than to obtain virus-free plants: According to CSVFA members suckers of Inosa or other traditional varieties could not be obtained at the time of plot installation. At NARC supply was low, so that two different kinds and to some extent small plantlets (height about 10cm) had to be accepted.

2.3.2 Plot design The planting lay-out is based on a 10 x 10m grid of alternating timber trees, with interplanted fruit trees at 5 x 5m distance. In between, M. textilis was planted at 2.5m distance (see fig.23). Cover crops and annuals were not planted, first, because soil protection from erosion seemed to be ensured by present Pueraria phaseoloides and other creepers, and secondly, because farmers considered the plots too distant from the village for planting, maintaining and harvesting annuals.

Figure 23: Planting scheme new plot in Cienda All species were planted in a regular sequence along all slope positions and previous land-uses to facilitate evaluation of their performance under all given conditions. The two areas without abaca planted (corresponding to subplots 2 and 6, s. below) were intended

• 43 • 2 Sites, material and methods as controls to compare effects of abaca growth to adjacent subplots 1 and 7.

2.4 Meteorological data

2.4.1 Weather data Weather data were obtained from the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA). The meteorological station is situated on LSU campus at 7m asl. The data cover the period from 1999 to April 2006 on a daily basis, including minimum and maximum air temperature [°C], relative humidity [%], rainfall [mm], wind speed [ms-1] and direction [1], potential (pan) evaporation [mm] and bright sunshine [min]. Pan evaporation was evaluated using the Penman equation on the basis of PAGASA observations. As the experimental plot in Cienda is located at >100m elevation, approximately 5km from LSU on the western footslopes of the cordillera, a portable weather station was installed at Cienda village in April 2004 for comparison. This µMetos (Pessl) device logged hourly air temperature, relative humidity, rainfall and wind speed in the same units as PAGASA and additionally solar radiation [Wm-2]. These data were then used for comparisons between both sites. Tiny Talk II sensors (Gemini) and loggers were used for registration of soil temperature. Measurements were conducted on site under different vegetation types in 5cm soil depth and covered time spans from 24h up to 2 months. On the basis of these data, regressions were calculated to predict Cienda soil temperature from PAGASA-LSU air temperature and rainfall.

2.4.2 PAR measurements In order to assess the impact of radiation on plant growth and mortality, Photosynthetically Active Radiation (PAR) was measured with ecotec PAR 750 sensors and stored in DL424 data loggers (all by Colorlite). Measurements were carried out on the Cienda plot during cloud-free midday hours as singular measurements. Four sensors were used simultaneously: Two stationary reference sensors under open sky in five and one meter height, respectively, and two mobile sensors in one meter distance mounted on a slat to maintain a horizontal position. The latter were carried to each of the 400 tree positions and held over the tree denoting exact time of measurements. Later a mean of the two mobile sensors, accounting for small-scale variation, was related to the simultaneous reference values and the PAR value expressed as percentage of PAR above canopy. As sun flecks play an important role under closed canopy (KELLMAN 1970; see also fig.24), integrated measurements of high temporal resolution would have been desirable for each spot. As this was not possible, representative places on the plot were observed, usually during 48h in time-steps of 2s. These observations were extended to the nearby 10-year old rainforestation site in Cienda and additionally included air temperature and relative humidity (sensors were ecotec TLM 35 for temperature, FLM 15-ME (both Colorlite) for relative humidity).

• 44 • 2 Sites, material and methods

Figure 24: PAR in an open area and under canopy and relative humidity intraday during one day in the rainy season 2005

2.5 Soil analyses

As far as not otherwise stated, fine earth refers to soil sieved through a 2mm sieve and all results are expressed on an oven-dry basis.

2.5.1 Subplots and sampling scheme In Cienda, ten subplots were established after an exploratory auger transect assessment to account for small-scale variability. These subplots were used for further soil sampling and positioned in order to represent different canopy cover, slope position and land-use history due to tenure. Subplots 1 & 2 and 6 & 7, apart from being different treatments with respect to abaca later, served as quality and sensitivity checks for soil analyses during the first months. For the paired plot approach, adjacent areas were sampled in representative zones. Obviously disturbed zones like footpaths, eroded soil, burnt areas, piles of harvested Figure 25: Subplots and sampling scheme new plots Cienda coconut husks etc. as well as plot margins were excluded from sampling.

• 45 • 2 Sites, material and methods

2.5.2 Soil profiles For selection of representative profiling points, auger samples were taken along transects covering different slope positions and relief forms. For general classification and characterisation one profile was then dug in each rainforestation area per site. For the new plantation in Cienda, a more detailed scheme included three profiles along a transect in different slope positions. Profiles were dug to a depth of 1m according to FAO procedures, but with a reduced profile area of approx. 1x1.2m to minimise root damages in the plantations.

2.5.3 Soil sampling In Cienda, the ten established subplots were used for a first assessment of variability with respect to Corg. These plots were retained for further analyses. For the paired plots, different sampling approaches were tested with respect to deviations. These always aimed to representatively cover the entire plot and ranged from eight single samples per plot up to five composite samples containing 20 nested individual samples each. All analyses concerning the paired plot approach were subsequently carried out on the same samples.

2.5.4 pH pH of field fresh soil was measured in H2Odist., 0.01M CaCl2 and 1M KCl using a Corning 10 and, in 2006, a Metrohm 826 pH-meter. Control measurements were carried out in Hohenheim in 2005 (in air dry samples). In all cases, ratios were 10g soil : 25ml of solution/water. Measurements were conducted 30min after homogenising the solution with a plastic rod.

2.5.5 Bulk density and volumetric water contents

3 Aluminium cylinders of 100cm volume were used to determine bulk density and WCvol. as described in SCHLICHTING, BLUME & STAHR (1995). Gravimetric water contents could then be transformed into volumetric values. For each soil horizon, an average out of 3 replicates was calculated.

2.5.6 Gravimetric water contents and soil water potential Self-made FDR sensors (Frequency Domain Reflectometry, see fig.26) were used to estimate soil moisture. Principally, frequency of a swinging capacitor differs depending (exponentially) on water contents of soil located between the two conductor plates. This frequency can be converted into voltage and read out using a common household multimeter. In direct vicinity of profiles Cienda PN1 (middle slope closed canopy) and PN3 (footslope open area), sensors were installed in four depths per profile corresponding to the vertical centre of each horizon (none in the stony C horizon of PN1). Two to three sensors were installed per depth, horizontally in the three upper and – for technical reasons – slightly inclined in the lowest horizon. Calibration of all FDR devices had been carried out in the laboratory with air, oven-dried soil, rewetted soil of 5, 10, 15, 20% gravimetric water contents and water. For each sensor an individual calibration curve was plotted from measurements in sandy soil to avoid influence of compaction. As this proved not representative for clayey Cienda soils (field capacity was exceeded at < 20% gravimetric water contents) and calibration with clayey soil would have caused other types of artefacts, calibration was undertaken through gravimetric water content measurements

• 46 • 2 Sites, material and methods

(auger method) carried out simultaneously to FDR and tensiometer readings during representative weather conditions.

Two tensiometers per horizon, extended with rubber hose (joints fixed with clamps and sealed with epoxy resin), were installed next to the FDR pits. After augering holes to the depth aimed for, extracted soil suspended in water was poured into the hole before inserting the tensiometer; thus a tight contact between the ceramic head and soil matrix should be achieved (HARTGE & HORN 1989). Distilled boiled water was filled into each tensiometer, which was then closed with a rubber septum. Hoses ended shortly above ground level and readings were always Figure 26: FDR sensor carried out during the morning hours, before the high rising sun would have led to expansion of the water column (SCHLICHTING, BLUME & STAHR 1995). Tensiometers were mantled by a plastic collar to keep rain water from entering the holes and septa were protected with aluminium foil against heat and rodents. For readings a digital differential pressure manometer (Greisinger GDH 200-13) connected to an injection needle to penetrate the septum was used. Simplified, matrix potential, not taking into account osmotic and pressure potential, was calculated as

ΨM = -|ΨH| - |Ψz|

ΨM = matrix potential (negative value above groundwater level) and Ψz = gravitation potential. Gravitation potential Ψz was converted from depth as 1cm water column = 98.0665 Pa. The top ground surface was taken as reference level. Pressures were log10 transformed and expressed as pF. In addition, gravimetric water contents were measured from auger samples of the same spots and depths periodically, especially during extreme conditions. KELLMAN (1970) mentions, that soil water dynamics are more strongly influenced by rainfall events than by seasons. Measurements were thus conducted event-specific from Feb-May 2006, i.e. during the transition from rainy into dry season. Initially, three samples per horizon were bulked, but after an obvious weighing error (Feb 13th, horizon 5-14, PN3) three individual samples were gathered per horizon and plot.

2.5.7 Particle size distribution Particle Size Analysis was carried out at Dept. of Soil Sciences in Hohenheim following a standard procedure. Weight of air-dried samples was 20g. For samples with pH CaCl2 > 6.5, carbonate destruction (using HCl) was carried out first; where SOM exceeded 1%, destruction of humic matter in H2O2 at increasing temperatures (stepwise up to 80°C) was necessary as a pre-treatment. In a following step, H2O2 was removed with water. Once electric conductivity was below 40µS, analysis was carried out following the sieving- pipetting procedure given by SCHLICHTING, BLUME & STAHR (1995). As Cienda samples sedimented too fast for pipetting of clay, they were shaken in NH3 solution during one week (instead of just overnight) for dispersion.

• 47 • 2 Sites, material and methods

2.5.8 Total nitrogen Analysis for soils as well as fractionated organic matter (s. 2.5.14) was carried out in 1g finely ground air-dried samples with a C-N analyser.

2.5.9 Phosphorus Generally, Bray II method was preferred as it would allow extraction of occluded and aluminium- as well as calcium-bound PI from acidic and calcareous soils, respectively. Moreover, for the calcareous soil in Punta, P extracted with Olsen's procedure (NaCO3) was not detectable. Topsoils for the paired plot experiment were analysed in 2005 at LSU root crops laboratory according to Bray II method (as described by PAGEL ET AL. 1982), using ascorbic acid as a colourant. Where the filtrate interfered in its natural colour with the ascorbic acid complex, it was first cleared with activated charcoal. For samples showing very low P contents, sample weight was doubled for a second run. Readings were taken at 880nm using a Hitachi U-2000 spectrophotometer. PI of profile samples was analysed 2006 at LSU following the Soil Research Test Plant Analysis Laboratory (SRTPAL) standard procedure, which is based on the Bray II method. Bray solution is made of 1.11g solid NH4F dissolved in 0.1N HCl (gives 0.03N NH4F). For the colouring agent, the following stock solutions are prepared: 1 litre of solution A contains 5.35g ammonium molybdate and 0.12925 potassium antimony tartrate (dissolved separately) plus 66ml concentrated H2SO4. Solution B is prepared of 2.0922g ascorbic acid dissolved in 1l distilled water. A 4:1 mixture of B and A solution then gives the colourant. Standards are made of KH2PO4. 2.5g of ground soil were extracted with 25ml Bray II solution (shaking them during 5min), filtered (Whatman 42) and 10ml colourant were added to a 2ml filtrate aliquot. Transmittance was read at 880nm in a Bausch and Lomb Spectronic 20 fotometer. For Cienda profile samples Porg was determined after SCHLICHTING, BLUME & STAHR (1995) as difference of 0.1N H2SO4 extracted P of ashed and non-ashed samples. This method basically corresponds to Truog's method (extractant 0.002N H2SO4) as described by PAGEL ET AL. (1982). For SOM fractions, ICP-OES element analysis was chosen after HCl/HNO3 extraction.

2.5.10 CECeff, CECpot and base saturation

Potential Cation Exchange Capacity (CECpot) describes the amount of exchange sites for cations in a soil at a buffered pH of 7.0, while effective CEC refers to a soil at its natural pH. Both were analysed as described by SCHLICHTING, BLUME & STAHR (1995). For CECpot cations were displaced with Na- and NH4 -acetate. For CECeff, cations in 10g of fine earth 2+ were exchanged by 100ml 1M NH4Cl and percolated with BaCl2. Ba was then replaced + by H from HCl and quantified with a flame fotometer at 873nm. CECeff and CECpot were computed as  −  [ / ]= Abs.sample Abs.blank x100 x vol. flask [ml] xdilution CECeff/pot cmolc kg 1000 x weight sample [g] xmeq where for CECpot meq refers to the exchange solution and for CECeff to the respective cation. Base saturation was calculated as the ratio of basic cations (S-value) to potential CEC: S BS [ %]= CECpot Aluminium saturation was calculated as Al3+ / S-value.

• 48 • 2 Sites, material and methods

2.5.11 Exchangeable basic cations Plant-available Ca2+, Mg2+, K+ and Na+ were determined for top soil samples (0-20cm) of the paired plot experiment. Analysis was carried out April 2005 at root crops laboratory, LSU. 2.5g of air-dry fine earth were mixed with 25ml 1N ammonium-acetate adjusted to pH7, shaken for 5min and passed through Whatman 42 filter papers. For detection the AAS (Varian 220FS) was run with air/acetylene and set to 422.7nm for Ca2+, 285.2nm for Mg2+, 769.9nm for K+ and 589nm for Na+. Concentration of the respective cation in the extractant was then computed using a calibration curve produced from standard solutions.

2.5.12 Pedogenic oxides of Fe, Al and Mn Oxalate extracts the mobile or 'active' fraction of Fe, Mn and Al oxides such as ferrihydrite, allophane or organo-complexes. Dithionite-citrate extraction is used for crystalline pedogenic forms of Fe, Mn and Al. For determination of 'active' Fe, Al and Mn contents, 2g of air-dry fine earth were extracted with NH4 -oxalate solution and filtered as described by SCHLICHTING, BLUME & STAHR (1995). The more stable crystalline pedogenic oxides of Fe, Mn, Al were extracted, centrifugated and decanted with dithionite-citrate solution from 2g of air-dry fine earth according to the procedure given by SCHLICHTING, BLUME & STAHR (1995). Detection was by atomic absorption spectrometry for both extraction solutions (Fe 248.3nm air/acetylene, Mn 279.5nm air/acetylene, Al 309.3nm N2O/acetylene).

2.5.13 Soil organic carbon Different methods were employed depending on availability of equipment, research purpose and number of samples. Among the used approaches to analyse Corg contents, Loss on Ignition requires least technical expenditure and no chemicals. Walkley-Black method (WB) is standard in many laboratories in tropical countries, but hazardous and pollutant as it makes use of potassium dichromate. Elemental analysis (here: LECO C/N analyser) is most convenient and supposed to deliver the most exact values, which are not influenced by clay contents as is LoI.

2.5.13.1 Loss on Ignition

LoI analysis was based on method descriptions by SCHLICHTING, BLUME & STAHR (1995), PAGEL ET AL. (1982) and AGRICULTURAL EXPERIMENT STATIONS OF CONNECTICUT ET AL. (1995). Ignition temperature (ranging from 360°C to >800°C in literature) had to be selected sufficiently high to ignite all carbon including charcoal, but on the other hand minimise errors arising from evaporating crystal water. Crucibles were tared after ignition during 4h in a muffle furnace at 550°C (there was no significant difference between tare weights at 105°C and 550°C). Approximately 5g of air- dried earth were then oven-dried at 105°C until constant weight (12h) and, after cooling down in a desiccator, weighed to mg. The muffle furnace was pre-heated to 550°C and, once it had reached this temperature, samples were ignited during four hours. After turning off the furnace and cooling down to 250°C, crucibles were transferred into a desiccator and, later on, weighed again. Soil organic matter was computed as follows:

Gross Weight 105°C [g] - Gross Weight 550°C [g] SOM [%]= Gross Weight 105°C [g] - Tare Weight [g]

• 49 • 2 Sites, material and methods and organic carbon as

SOM [%] C [%]= org 1.724

2.5.13.2 Elemental analysis Carbon and nitrogen contents of finely ground profile samples were determined using a LECO C-N analyser. This analysis was carried out at Landesanstalt für Landwirtschaftliche Chemie (LACh), and Dept. of Soil Sciences, Hohenheim.

2.5.13.3 Wet combustion SRTPAL standard procedure is based on the Walkley-Black method, differing from other wet digestion methods in that Cr3+ is quantified by titration, not fotometrically. 0.5g of finely ground soil are weighed in a flask, then 10ml of 1N potassium dichromate (K2Cr2O7) and 10ml of concentrated H2SO4 are added. After swirling and leaving the flask for one hour, it is supposed, that all organic C has been oxidised, setting a corresponding 3+ amount of Cr ions free. The remaining K2Cr2O7 is backtitrated with 0.5N ferrous sulphate (FeSO4 x 7H2O) and 0.025M O-phenanthroline – ferrous complex as an indicator. The reduced Cr3+, which corresponds to the amount of organic C, can then be calculated as fraction of a blank subtracted from 1 as follows:

S [ml] 0.69 Organic Matter [%]=10 1−  B [ml] Sample Weight [g] where S and B are amounts of FeSO4 solution for the titrated sample and blank. The factor 0.69 is derived from

12 1.72 1N 100% = 0.69 4000 0.75 where 1N is the normality of K2Cr2O7; 12 / 4000 = meq of carbon, 1.72 conversion of organic C into organic matter and 0.75 an empirical recovery factor for organic C, taking into account, that only a certain proportion (75%, the easily oxidisable fraction according to AGRICULTURAL EXPERIMENT STATIONS OF CONNECTICUT ET AL. 1995) of total Corg is oxidised.

2.5.13.4 Assessment of methods Loss on Ignition is simple, efficient and robust, but often said to be of low accuracy due to crystal water bound to clay and oxides, which is accounted for as carbon when weighed. Walkley-Black is claimed by some to be more exact34 than LoI. Given large numbers of samples, LoI was first choice, after calibration with Walkley-Black and Elemental Analysis. A general source of error for WB is, that only easily oxidisable C is detected, which is then converted into Corg by multiplication with the recovery factor 0.75. On the other hand, recovery from 0.59-0.94 has been observed as a consequence of different temperatures

34 in contrast to SCHLICHTING, BLUME & STAHR (1995): [...] given the numerous presumptions [for Walkley-Black method], it must appear surprising, that results in most parallel experiments sufficiently match those obtained by elemental analysis [p.160].

• 50 • 2 Sites, material and methods

(SRTPAL procedures, unpublished). Chlorides, ferrous iron (Fe2+) as well as organic N or S can also be oxidised and hence lead to overestimation, whereas manganese or oxidised organic C will result in underestimation of Corg (SCHLICHTING, BLUME & STAHR 1995). Consequently, AGRICULTURAL EXPERIMENT STATIONS OF CONNECTICUT ET AL. (1995) recommend to collect a representative cross-section of soils from the area under research, calculate a regression Walkley-Black – LoI and continue analyses using only LoI. This regression would not take clay contents into consideration, in contrast to what has been suggested by SCHLICHTING, BLUME & STAHR (1995). Unlike WB procedure, LoI and other dry combustion methods include all organic carbon but, depending on combustion temperature, also other fractions like crystal water. Interrelation of LoI and WB was assessed for 20 SRTPAL standard soils from different sites in Leyte (fig.27). As recommended by SCHULTE (1995), a correction factor for clay was not subtracted.

Figure 27: Comparison of CLoI and CWB for 20 Leyte soils For profile samples, a CN-analyser was available in Hohenheim, so that the relationship between LoI and elemental analysis (EA) was of interest. Samples from the Cienda profiles with known clay contents were analysed, and regressions were calculated with and without the use of a correction factor for clay (subtraction of 0.1% Corg per % clay, as suggested by SCHLICHTING, BLUME & STAHR (1995)). This factor resulted in negative values for all horizons rich in clay or low in OM, so that an empiric linear regression including the independent variables CLoI and clay contents was formulated (fig.28).

• 51 • 2 Sites, material and methods

Figure 28: Observed vs predicted values for Corg at Cienda profiles Clay contents did not exercise substantial influence on the dependent model, so that coefficient of determination was r2 = 0.79 for the model as well as for LoI by itself. The formula was used for calculations of Corg in chapter 5. After converting SOMLoI into CLoI, Corg was calculated as follows:

Corg [%] = - 2.163 + 0.814 CLoI [%] + 0.001 clay [%]

The lower C contents of EA may be explained to a minor extent by losses in C contents during storage (approx. 3 months), but the main reason is due to the procedure, as discussed above.

2.5.14 Physical fractionation of soil organic matter One bulked sample containing 12 cores of almost 600cm3 each (Eijkelkamp root corer, sampling depth 0-15cm) was collected per subplot. The soil surface was not cleared at sampling time to include the litter layer, yet living aboveground biomass was sorted out as were live roots. The entire sample was air-dried and sieved to 2mm and the following fractions were separated following the procedure described by GAISER (1993), with some modifications: The Light Organic Matter > 2mm fraction was separated by flotation in water (decantation into a 2mm sieve), segregated manually into aboveground biomass and roots and then dried at 60°C and weighed. C, N and P contents of a subsample were determined. Stones and charcoal fraction > 2mm were also weighed. To obtain the Light Organic Matter < 2mm fraction, 1kg of fine earth was mixed in a pail with roughly 5l of water and stirred well. Floating organic matter was then carefully decanted onto a 0.25mm mesh sieve. This fraction was then washed out of the sieve with a small amount of water. It was dried at 60°C and weighed. A subsample was set aside for C, N and P analysis. To the remaining rest of Heavy Organic Matter + Mineral Soil < 2mm and water, 10ml of concentrated HCl were added. This mixture was left overnight for sedimentation and then decanted. A subsample of the sediment was dried and the moisture conversion factor calculated. Carbon, nitrogen and P contents of this fraction were analysed. Elemental

• 52 • 2 Sites, material and methods analyses were carried out using a C-N analyser and for P an ICP-OES at LACh Hohenheim. Carbon contents of the fractions' subsamples were then multiplied by the respective oven-dry weight of each fraction.

2.5.15 Substrate-Induced Respiration Substrate Induced Respiration (SIR) method is applied to estimate microbial biomass in soils. After addition of an easily accessible substrate, glucose35, to a soil sample, the increase in microbial metabolism, expressed as respiration, allows conclusions on the size of the actively metabolising microbial population. Field-fresh soil samples were stored at 8°C, then passed through a 2mm sieve and remaining roots were picked out with forceps. If the sieved samples could not be analysed immediately, they were stored in a deep-freezer. Prior to incubation, samples were adapted to room temperature during 5 days. In 2004, incubation temperature was approx. 32°C (dry season room temperature), in contrast to the original procedure (ANDERSON & DOMSCH 1978). Only in 2005, an air-conditioned room was available for the paired plot experiments, so that the analysis could be carried out at 22°C as described in most literature. As night topsoil temperatures in Leyte can be as low as 22°C, a shock, drastically reducing metabolism after conditioning at room temperature was not expected. Still, samples were acclimatised to 22°C the morning before the experiment actually started (preparations would usually take until noon). Gravimetric water contents were parallely determined in subsamples after sieving and samples were adjusted to 55% of water holding capacity on the 4th day of conditioning. Water holding capacity is computed as: fresh weight - dry weight [g] WHC [%]= x100 and thus can exceed 100%. dry weight [g]

SIR was carried out according to the procedure given by SCHINNER ET AL. (1993) with some modifications: Samples were incubated as fivefold replicates including one unamended control. For each replicate 20g of soil were filled in a fine-mesh nylon bag and placed in an airtight glass jar containing 20ml of 0.1M NaOH. After 4 hours, samples were removed from the jars. Following the Isermeyer approach, CO2 evolving through respiration is trapped by a known amount of sodium hydroxide and then precipitated with barium chloride (forming BaCO3). CO2 evolved from the soils was precipitated from the NaOH using 2ml of 0.5M BaCl2. NaOH not consumed by CO2 was backtitrated with 0.1M HCl adding 3-4 drops of phenolphthalein indicator. CO2 evolved from respiration can be computed as:

Blank−Sample[ml HCl] 2.2 100 100 = mgCO 100g-1 DMh-1 , 4 sample weight[g] %DM 2 which can then be converted into microbial biomass carbon (Cmic):

-1 -1 -1 (SCHINNER ET AL. 1993). 20.6 x 10 mgCO2 100g h = µgCmic g A pretest was first carried out to determine the minimum dosage of glucose required to render a maximum reaction of microbial respiration. Different dosages of 0, 75, 150, 300, 600 and 1000mg glucose per 100g dry soil were tested as well as different concentrations of NaOH and HCl to obtain adequate resolution of titrated values.

35 applied in solid form, not dissolved (see LIN & BROOKES (1999) for discussion).

• 53 • 2 Sites, material and methods

Soils need to be in equilibrium for SIR; recently fertilised or tilled sites with growing or decreasing microbial populations are not suited for SIR experiments (ANDERSON & DOMSCH 1978). This precondition was assessed for Cienda soils in 2004: In a pre-test, respiration had been observed after 1, 2, 3 and 4 hours of incubation to monitor the shape of the respiration curve for horizontality (SCHINNER ET AL. 1993). In 2005, interviews with landowners or tenants were used to assure that such changes had not occurred recently.

2.5.16 Basal respiration As for SIR, basal respiration (BR) of sieved soils, excluding roots and macrofauna, is measured as evolution of CO2. In contrast, no substrate is added, temperature is not adjusted and incubation periods exceed four hours. BR method is used to reflect microbial metabolism under controlled yet close-to-natural conditions. In 2004, several pre-tests were carried out: Different concentrations of NaOH and HCl were compared for trapping CO2 and titration; 0.1M NaOH and 0.1M HCl were identified as best concentrations to trap the entire amount of CO2 and at the same time allow for sufficient resolution during titration. Soil was broken up but not sieved; roots were sorted out with forceps. Water contents were adjusted to 50-60% of water holding capacity before the experiment started. During the incubation period of 30 days, water contents were controlled and losses in the airtight jars found to be negligible; thus moisture was not adjusted during the experiment. Transport and storage before the experiment were as for SIR including an incubation period of 5d at room temperature. 30g of field fresh soil were used per sample. The experiment was carried out following the procedure given in SCHINNER ET AL. (1993). Each sample was repeated in 7 technical replicates plus one blank (no soil) to evaluate the quality of the method. Samples were incubated at room temperature, transferred to new glass jars after 1, 3, 6, 12, 18, 24 and 30 days and titrated after each transfer. Samples from all subplots were collected, incubated and analysed simultaneously, so that there were no differences with respect to temperature. During the second year, water contents were adjusted to 55% of WHC to make measurements on different dates more comparable. This was necessary due to a larger number of different plots, which could not be processed parallely anymore. Also, technical replicates were reduced from 7 to 4, which had been found sufficient to prove statistically significant differences between plots. Experiment duration was reduced, too, (changes after 24, 96 and 168h) as longer experiments did not make findings more evident.

2.5.17 Soil respiration For Cienda subplots 1-13, soil respiration was also estimated in situ following the pail method described by ANDERSON & INGRAM 1993. In contrast to the aforementioned respiration methods, this analysis includes faunal and root respiration over the entire profile depth. As installation of pails disturbs relatively large areas of an experimental field, there were only two replicates installed per subplot to roughly assess if tendencies were comparable to BR analysis. After removing all aboveground biomass, a glass vial containing 1M NaOH was placed under a pail, which was inserted 3-4cm into the topsoil and the soil compressed against the pail; thus gas exchange of in- and outside atmospheres was minimised. After 24h of exposure, the set-up was opened, glass vials were sealed, removed and brought to the laboratory, where precipitation and titration (here with 1M HCl) were carried out as for -2 -1 basal respiration. Results are presented as g CO2 cm d , where area of the pail – area of vial = 484.15cm2 – 16.76cm2 = 467.39cm2.

• 54 • 2 Sites, material and methods

Shortly before the installation, samples for parallel BR analysis had been taken next to the pail positions. Simultaneously to the experiment, soil temperature in 5cm depth was monitored in two subplots (under closed canopy and in an open area) during the exposure period.

2.5.18 Phosphatase activity Measurements of phosphatase activity are used to assess activity of phosphate- transforming enzymes at soil pH, which includes phosphatases of plant roots as well as such of microorganisms. In the experiments, accordingly to SCHINNER ET AL. (1993), di-sodium phenylphosphate served as a substrate, which is split up by phosphatases into phosphate and phenol, the latter being detected photometrically in relation to phenol standard solutions. Samples were taken at the same dates as those for BR and SIR. After cooled transport to the laboratory and discarding roots, the field fresh samples were stored in a deep-freezer. Before analysis, soil samples were defreezed and conditioned at room temperature for five days. Each sample was processed as threefold technical replicate with one additional blank. To 5 g of soil, 10ml of 0.2M di-sodium phenylphosphate and 10ml H2Odist were added into Erlenmeyer flasks, for the blanks another 10ml distilled water were used instead of the substrate. Incubation of the samples was during 3 hours at 37°C. After incubation samples were immediately transferred to a cooled room, where activated carbon was added. After shaking, samples were filled quantitatively into 50ml volumetric flasks, volumed and filtered through Whatman #42 filter papers. 2ml of the filtrate were then transferred into 100ml volumetric flasks, which had been prepared with 5ml borate buffer solution and 1ml 2,6-Dibromoquinone-Chloroimide as colouring agent. Phenol was determined photometrically at 614nm against phenol standards. In modification to the procedure described by SCHINNER ET AL. (1993), 5ml of 0.2M instead of 0.1M substrate solution were used as a pre-test had shown that the indicated amount of substrate gave very low phenol readings. Further, measurements were conducted with distilled water at soil pH instead of buffer solutions used for acid or alkaline phosphatase. Nevertheless, a borate buffer (pH 10) was added after filtration to make the colourant work (and also for the standards). Incubation was at 37°C for 3 hours, the following steps were then carried out in an air-conditioned room at approximately 22°C to stop the enzymatic processes, which might have gone on at 32°C room temperature. After filtration, flasks were wrapped in aluminium foil to protect the solution from light. To avoid turbidity of extracts (probably due to iron oxides), which interferes with the wavelength to be detected, activated carbon (ground and acid-washed) was used after incubation to clear the solution. Erlenmeyer were used instead of volumetric flasks for incubation and shaking; then the solution was quantitatively transferred into volumetric flasks. Absorbance values obtained from the photometer readings were transformed into µg phenol concentrations per ml. In a second step, phosphatase activity is expressed as µg phenol per soil and incubation time:

Sample−Blank[µg/ml] 50 100 Phenol [µg] = , Filtrate [ml] Weight of Sample[ g] Dry Matter [ ] Dry Matter [g] 3h where sample (mean of replicates) and the respective blanks are given as µg phenol / ml, 50 = ml extractant volume, 100 / (% dry matter) = conversion for water contents. As suggested by DENICH & KANASHIRO (1998), pH and PI were parallely determined for every phosphatase sample.

• 55 • 2 Sites, material and methods

2.6 Biomass measurements

2.6.1 Mulched biomass Biomass cut during plot preparation in Cienda was estimated from subsamples. Lay-out of the five-by-five meter grid for trees to be planted was carried out the day after brushing. Within every second 25m2 quadrant (corresponding to all fields of the same colour on a chess board), a 1x1m square was randomly selected by drawing a number. All aboveground biomass within this 1m2, which had been cut during brushing36, was collected and necromass discarded. Woody and non-woody parts were separated . Samples were oven-dried at 70°C until constant weight; dry matter was recorded separately for both fractions.

2.6.2 Undergrowth biomass and growth rates A rough assessment of undergrowth biomass and growth rates as reference for modelling was carried out under canopies of different density. Five squares of 1x1m were laid out under dense, medium and no crown cover, respectively. AGB was cut in February 2006, separated into woody and non-woody parts, oven-dried (70°C for 24h) and weighed. The regrowing AGB was then cut again after 6 and 12 weeks, i.e. at the end of the rainy and during the dry period. Carbon contents of three composite samples were also determined.

2.6.3 Root length and weight density Soil samples were taken from 0-15 and 15-30cm37 using an Eijkelkamp soil corer of 15cm tube length and 7cm inner tube diameter. A maximum distance to trees was kept to ensure that RLD reflected stand density. Two replicates per subplot were sampled, but only one, A series, each was lastly analysed for root length density (RLD) due to labour- intensity of the method. The B series was used for a complementing RWD analysis. Roots were washed out of the soil cores in the lab and separated into live and dead: First by floating in water and then manually sorting out, using colour and elasticity as distinguishing criteria38. Roots were then dried for one hour, cut into pieces and scanned. Root length was determined in Hohenheim using WinRhizo 5.0 (BAUHUS & MESSIER 1999) and related to the volume of the soil core to calculate RLD. RWD was calculated for a replicate of every RLD sample, divided into fractions > and < 1.5mm diameter. Additionally RWD of roots >2mm from OM fractions analysis (2.5.14) were taken as reference. Dry matter per soil mass values were multiplied with bulk density obtained from the soil profiles. Bulk density entered as a mixed factor, weighted according to thickness of horizons.

2.6.4 Aboveground biomass growth of planted species In order to assess performance of the planted trees and abaca in different environments (regarding canopy closure, slope position and soil properties), various measurements were conducted. Six inventories were carried out during the first two years after planting, recording all plant heights and stem diameters. To relate these values to biomass increase, three individuals per species were measured non-destructively in detail for calibration and deduction of empiric allometric equations. Biomass could then be converted into amounts of sequestered carbon. Biomass values were also estimated

36 This refers to herbs, shrubs and small trees only, as medium-sized trees were not cut. 37 These were defined after root counts from the soil profiles. 38 As assignment was not unambiguous in many cases, this aspect was not considered for data evaluation.

• 56 • 2 Sites, material and methods using the WaNuLCAS model, which is based on functional branch analysis (FBA) of mature trees. While the 'manual' generation of allometric equations does not allow for extrapolation beyond the calibrated range, this was possible using WaNuLCAS, which requires only partial analysis of grown-up trees to estimate the entire tree biomass.

2.6.4.1 Inventories With regards to drought being the principal cause for mortality and growth retardation, a first tree inventory (June 27-28, 2004) was undertaken shortly after the dry season that followed planting to account for the established trees and abaca. Further dates were November 11, 2004 and April 14, 2005, both before, and June 29, 2005, shortly after dry season. Two more measurements were carried out in December 2005 and April/May 2006. Tree height of each plant was measured from ground level to the lower end of the terminal bud. For abaca the bifurcation between the two youngest leaves was chosen as upper point. Instead of conventional diameter at breast height (130cm), diameter was measured at 5cm above ground level. This point was marked with an indelible colour pen and diameter was determined using a calliper gauge. Mortality was also recorded. In addition, monitoring of leaf damage was conducted five times from the second growth inventory (Nov. 2004) until January 2006. Damages were estimated in percent of leaf area and categorised into pest damage, sunburn and others (mainly black, red and yellow spots). Mechanical breakage of shoots was also denoted.

2.6.4.2 Calibration for allometric equations As destructive methods were not appropriate, three individual trees per species were gauged in detail for biomass determination. These individuals included the tallest, one of the smallest and one medium exemplar to cover the whole range of planted trees and avoid extrapolations. Length and diameters at both ends of every stem, branch or twig were determined separately. Where segments' cross-sections were oval, two perpendicular diameters were taken and volumes were calculated for cylinders/frustra with elliptical basal areas. For compatibility, the same diameter classes for wood, branch and twig as for the FBA method (see 2.7.2.1) described by MULIA (2001) and MULIA, PURNOMOSIDDHI & LUSIANA (2001) were used. Wood is thereby defined as >5cm in diameter, branch 5-2cm and twig <2cm. Wood, branch and twig density (at 0% moisture) of each species were determined by weighing oven-dry subsamples of known volumes. Displacement of water (with detergent) and geometric calculations were both used to compare precision of volume measurements. Where literature values (ICRAF wood density database at www.worldagroforestrycentre.org, SOERIANEGARA ET AL. 1993 and FAO 1982) for air-dry wood density complemented own measurements, these were adjusted to 0% moisture neglecting shrinkage. Multiplication of woody volume and density gave then woody aboveground biomass of the respective tree. Previous leaf measurements had served to establish length – width ratios. For average leaf area, scanned leaves were measured using WinRhizo 5.0 software and then regressions were formulated to estimate area from either leaf length, width or length multiplied by width. As criterion r2 of regressions was employed. Specific Leaf Area (SLA) had been determined as average of 10-20 oven-dried leaf circles of known diameter and gave the inverse specific dry weight per unit area. The method has been described by MULIA, PURNOMOSIDDHI & LUSIANA (2001). For the calibration procedure, leaf length of a number of single leaves was measured and

• 57 • 2 Sites, material and methods an average leaf length was calculated. Given the abovementioned ratios and the number of leaves per tree, leaf dry matter for every species could be estimated. This was added to woody biomass to account for total aboveground biomass (AGB).

2.6.4.3 Allometric equations Several types of alternative regressions were fitted to predict biomass values from diameter and/or height. In a first step, relationships between stem diameter and height of all trees were tested to decide, whether both parameters or only diameter should be considered for the allometric equations. Generally, including height into the regression equation would add accuracy, but in cases of distorted growth, mechanical damage etc. would lead to greater deviation. Allometric equations were then fitted on the basis of the calibration measurements described beforehand. Regressions were formed following the exponential shape

B = a Db or B = a (HD2)b where B is biomass, H height, D diameter at 5cm height and a and b are empirical species-specific values (see KETTERINGS ET AL. 2001).

2.6.5 C, N and P contents of plant tissues Composite samples of each tissue type – leaves, twigs, branches, wood and roots of the ten planted tree species plus Gmelina arborea and leaves, pseudostem and roots for abaca - were analysed at SRTPAL Leyte following standard procedures:

2.6.5.1 Plant C analysis Sample preparation for C and P analysis: 0.5g oven-dry plant tissue is ashed at 550°C in a muffle furnace during 6-8 hours, until the sample turns white (black particles are removed with 3ml of concentrated HNO3). The ash is then dissolved in 3ml concentrated HCl overnight, quantitatively transferred into a 100ml flask and volumed with 0.1N HCl. After filtering through a Whatman 42 or equivalent filter paper, C and P are analysed. C analysis was conducted according to the modified Walkley-Black procedure as described under 2.5.13.3 for soils.

2.6.5.2 Plant P analysis Plant phosphorus is analysed in 1ml ashed tissue sample extract, which is amended with a reagent 'C' composed of two solutions (Stock Solutions A and B as described under 2.5.9). 10ml of reagent 'C' are mixed with the tissue extract and left for 15min before being measured fotometrically (here: Bausch and Lomb Spectronic 20) at 880nm. Standards are made from 0.4394g oven-dried KH2PO4 plus 5 drops toluene in 1l H2Odd. From this 100ppm stock solution, working standards are prepared.

2.6.5.3 Plant N analysis Plant N was analysed following the SRTPAL standard procedure based on the Kjeldahl method. In the presence of a catalysing mixture (K2SO4, cupric sulphate pentahydrate and selenium), organic N is converted to (NH4)2SO4 through reaction with H2SO4 and then to NH4OH by addition of NaOH. During the subsequent distillation, NH3 is set free and bound - by H3BO3 as (NH4)H2BO3. N is then quantified indirectly by titration of H2BO3 with HCl. To 0.5g of ground air-dry plant sample, 0.2g of selenium catalyst mixture and 3ml of conc.

• 58 • 2 Sites, material and methods

H2SO4 are added. The solution is heated until clear. After digestion, the liquid is allowed to cool and 30ml of distilled water are slowly added after quantitative transfer into a Buchi flask. Distilled NH3 is collected in an Erlenmeyer flask containing 25ml of 2% H3BO3, 3 drops of mixed bromcresol / methyl red indicator and 20ml of 40% NaOH. Titration starts directly after distillation with 0.05N H2SO4 as titrant. Parallely to the samples a blank is analysed and the results are computed according to the formula:

H SO − H SO [ml] N 0.014 N [%] = 2 4 sample 2 4 blank titrant 100 sample weight [g] where Ntitrant stands for normality of the H2SO4, and 0.014 is the meq of nitrogen.

2.6.6 Litter production To assess litter production under different canopy cover in Cienda, litter traps were randomly installed inside the subplots as proposed by ANDERSON & INGRAM (1993). Six litter traps with a surface area of 0.25m2 each (50x50cm) were set up per subplot. The collected plant material was segregated into leaves, branches & bark and flowers & fruits, dried at 60°C and weighed. For larger branches and coco leaves, only the part lying on the trap was cut off for weighing. Traps had first been installed by DIEHL (2005) and litter recollected irregularly. From early 2005 up to the destruction of most traps in February 2006, litter harvesting was carried out monthly. The same experiment was set up in Marcos plots from Feb – May 2006 to compare litter production in reforestation plots with indigenous vs. exotic (Gmelina arborea) trees.

2.6.7 Litter decomposition PVC rods containing 12 minicontainer capsules each (as described by EISENBEIS 2004 and DUNGER & FIEDLER 1997; fig.29) were filled with 0.1500g dry matter of different standard plant materials39 and inserted horizontally into the soil at a depth of 5cm. Exposition periods varied according to plant materials (2 weeks and 5 weeks). Two different types of nylon nets where used as cover to allow access for specific decomposer groups: 4mm (access for most decomposers) and 0.1mm mesh (to exclude the macro- and mesofauna). Plant material was oven-dried at 70°C for 48h before and after exposure. Figure 29: Minicontainers in PVC rod, mesh The same set-up was used for an experiment 4mm and 0.1mm, randomly distributed in Marcos comparing Ficus-dominated substrate as well as plot-specific leaf litter of rainforestation vs. Gmelina plantation during wet and dry season.

39 easily decomposable Leucosyke capitellata leaves and Cocos nucifera fine roots in a second run

• 59 • 2 Sites, material and methods

2.7 Plant measurements required for modelling

Numerous models have been formulated to predict soil organic matter in different pools on a plot (CHERTOV & KOMAROV 1996; COLEMAN & JENKINSON 1996; COLEMAN ET AL. 1997; FRANKO 1996, KELLY ET AL. 1997, MOLINA ET AL. 1996; for an overview POWLSON ET AL. 1996) and landscape level (PAUSTIAN ET AL. 1997a; FALLOON ET AL. 1998; V. NOORDWIJK 2002). Among the existing plant growth models, WaNuLCAS was chosen because it is designed explicitly for agroforestry and can simulate interactions and simultaneous growth of up to four plant species in a mixed planting system. The model was written for tropical agroecosystems and parametrisations for some of the trees used here was already available for comparison with own measurements. Relevant modules for this research, especially with regards to the carbon cycle, allow extensive options to feed own data into the model. The model has a user-friendly interface, but the underlying algorithms can be modified on the Stella 'subsurface'. Updated technical support, also with respect to gathering of input data, is facilitated by ICRAF through manuals, publications and procedures. WaNuLCAS is freeware, but requires a commercial environment, Stella. The version used for this study was WaNuLCAS 3.1.

2.7.1 Crop parametrisation Apart from soil and weather data (on a daily time-step), working with WaNuLCAS requires specific inputs for plant parametrisation. These include a tree survey focussing on habitus, phenology and physiology of the selected species. Different support programs have been written to derive inputs for crop parametrisation in WaNuLCAS, which are usually not included in standard measurements. These are Light Use Efficiency (LUE), Specific Leaf Area (SLA), Leaf Weight Ratio (LWR), Harvest Allocation and Root Allocation. WOFOST (BOOGARD ET AL. 1998), a software recommended for WaNuLCAS (RAHAYU ET AL., available online) simulates optimum crop growth depending on growth stage; outputs are then entered into a WanHelp spread sheet, which delivers input parameters for the crop library in WaNuLCAS. While SLA and LWR data were collected on site, LUE as well as harvest and root allocation were derived from WanHelp, based on dry matter of different plant fractions at various growth stages.

2.7.2 Tree parametrisation This part of the survey can be addressed through interviews with experienced farmers, extensionists or other resource persons. Other parameters need to be quantified experimentally; these are specific leaf area (SLA), wood density, C, N and P (see previous sections) as well as Functional Branch Analysis, leaf weight ratio, polyphenolics and lignin contents of leaves and fine roots (see below).

2.7.2.1 Functional Branch Analysis and Leaf Weight Ratio FBA is a method employed to describe branching patterns of trees based on non- destructive measurements. Roots and stems are seen as transport channels for water and dissolved substances and their shape, especially cross sectional area, is determined by magnitudes of these fluxes (V. NOORDWIJK, SPEK & DE WILLIGEN 1994). Assuming a self- repeating pattern of branching from fine roots to proximal roots and then from the stem towards branches and twigs, species-specific geometric ratios can be observed. To come to these ratios, length and diameters of aboveground plant segments or links per plant are

• 60 • 2 Sites, material and methods measured. The first aboveground link would always be the stem from the ground to the origin of its lowest first-order branch; the second segment would then extend from the beginning of the first branch to the first second-order branch turn-off and so on. Length, two diameters (measured perpendicularly to account for elliptical cross sectional area in the middle of the segment), linkage to parent and number of leaves are noted for at least 100 aboveground segments (table 6). Table 6: Schematic FBA protocol

Link # Length Diameter 1 Diameter 2 Parent # Number of leaves 1 : : : : : : : : : : : 100 +x : : : : :

These data are copied into WanFBA, a pre-processing software for WaNuLCAS, where inputs are checked for plausibility. Factors p and q and their respective means and ranges are the most important variables dominating the branching pattern. The first describes the squared diameter (representing cross sectional area) of a parent link in relation to that of all of its daughters

D2 p= parent ∑ 2 D daughters and q the ratio of the thickest daughters' squared diameter in relation to the sum of all its sisters' squared diameter:

D2 q= max ∑ 2 D daughters

As the branching pattern is fractal, a user-defined minimum diameter Dmin needs to be set to stop the downscaling process; Dmin may be chosen < Dlow, the smallest measured diameter on a twig. Bare Tip Length (from the junction of the youngest leaf to the tip) and Leaf Weight Ratio (LWR), describing dry weight of leaves relative to leaves + woody parts, are further inputs for the parametrisation. These data are usually collected in the process of FBA for the same branches. WaNuLCAS will then fit allometric regressions for each individual based on the processed exports from WanFBA. Averages of three mature individuals of S. contorta, G. arborea, A. heterophyllus, A. odoratissimus, D. zibethinus, N. lappaceum, G. mangostana and two replicates for D. validus were used for allometric equations derived from FBA. While the procedure demands a minimum of 50-100, at least 100 links were measured per plant. If exceeding 100 links, the entire branch, at times >200 links, was included. For roots, generally the same procedure is applied, starting from the proximal root and denoting number of fine roots instead of leaves. Additionally, average length of fine roots and length per dry weight are determined. Especially in heavy soils roots are not easily accessible without breakage. In consequence the method is labour- and time-intensive, so that for this study less segments were measured per plant and not all species could be sampled. Alternatively, WaNuLCAS offers two more options to estimate roots biomass, one being maximum root length density per zone, which were obtained for some cases

• 61 • 2 Sites, material and methods following the method described under 2.6.3. All other procedures mentioned in this section were described by MULIA (2001) and MULIA, 40 PURNOMOSIDDHI & LUSIANA (2001) and are available online .

2.7.2.2 Total extractable polyphenol (Folin – Ciocalteau method) and lignin contents 0.2g of ground plant sample are placed in a 50ml beaker and 20ml extractant (50% methanol dissolved in distilled water) is added. The beaker is covered with a watch glass and heated in a water bath at 75oC for one hour. The extract is filtered (filter paper #1) and volumed to 50ml with extractant. 1ml filtrate is transferred into a 100ml volumetric flask and 60ml distilled water, 5ml Folin-Ciocalteau reagent and 15ml of 20% sodium carbonate solution are added before voluming to 100ml with distilled water. Standards are prepared from a 1mg ml-1 tannic acid stock solution. 0, 1, 2, 3, 4, 5, and 6ml of stock solution are pipetted into 50ml volumetric flasks; 2.5 ml of Folin – Ciocalteau reagent and 10 ml of 17% sodium carbonate are added before voluming with distilled water. Absorbance of standards and samples is read at 760 nm using a spectrometer and expressed as tannic acid equivalent (TAE). Total Extractable Polyphenol is computed as follows:

TAE −TAE [mg] 50 [ ]= sample blank TEP % [ ] 10 sample weight g

2.8 Statistics

Statistical evaluation was carried out with SAS 8.2, Minitab 13 and SPSS 12.0. Averages, standard deviations, coefficients of variance, coefficient of determination and correlations were calculated in OpenOffice Calc 1.1.5 spread sheets. For statistical evaluation of model goodness of fit, the following parameters as suggested by LOAGUE & GREEN (1991) were employed:

Coefficient of determination: Root mean squared error:

n n ∑  − 2 =[∑  − 2 / ]0.5 100 oi o RMSE pi o n = i=1 i=1 o CD n ∑  − 2 pi o i=1

Modelling efficiency:

n n [∑  − 2−∑  − 2 ] ,where o are observed and p predicted values; oi o pi oi i i = i=1 i=1 overlined o stands for the mean of observed values EF n and n for the number of observations. ∑  − 2 oi o i=1

40 At www.worldagroforestrycentre.org

• 62 • 3 Characterisation of soils

3 Characterisation of soils Variability of relief and soils in Leyte is high and both exercise an important influence on plant growth. LANGENBERGER (2003) described relief-dependency of natural vegetation in the research area. This chapter will give an insight in main soil forming processes as caused by climate and relief. Assets and constraints of soils at different sites will be discussed to better understand growth conditions for plants in agroforestry systems.

3.1 Profile descriptions

Samples were collected from soil profiles in 2004 at Cienda site and in 2005 on the paired plots (s. chapter 4, also for land use history). One soil profile per paired plot site was excavated, while on the new Cienda plot three profiles were installed. The nearest meteorological station for long-term data is PAGASA-ViSCA. Soil climate for all profiles is isohyperthermic (mean annual soil temperature ≥ 22°C with seasonal amplitude in 50cm depth ≤ 6°C) and udic (not dry in any part for longer than 90 cumulative days in normal years). All sites are situated in formerly forested areas, which have been used for agriculture during the last decades. None of the sites is still covered by the original vegetation, which was evergreen rainforest. All analysis were carried out in duplicate in the following laboratories: pH, bulk density and PBray were analysed at Dept. of Agriculture and Soil Sciences, Leyte State University; Texture at Dept. of Soil Sciences, Hohenheim; C, N, Fe, Mn and Al at Landesanstalt für Landwirtschaftliche Chemie, Hohenheim; CEC and part of Corg/NT at Dept. of Plant Nutrition and Soil Sciences, University of Halle. Ecological evaluation followed the procedures and calculations described by JAHN, BLUME & ASIO (2003). These include aggregate stability as estimated on the basis of structure and pH; permeability (Ksat) derived from texture and bulk density and erodibility read from a nomograph of texture, organic matter, structure and Ksat. Erodibility covers only the soil- related component of erosion. For estimates of erosion (e.g. after Wischmeier), rainfall intensity, soil cover and slope need to be considered. Air capacity was evaluated after AG BODENKUNDE (1982).

• 63 • 3 Characterisation of soils

3.1.1 Haplic Cambisol, Cienda PN1 PN1, 2 and 3 are on the same site, 1 located on the upper middle slope position, 2 at the lower middle slope and PN3 at the footslope on a natural terrace. Land-use and present vegetation: Coconut, secondary forest, bush fallow. Scarce undergrowth, low creeping grass and ferns, some Pueraria. Weather: Sampling follows a rainy night after 4 weeks dry period.

Profile # 1 Date of description 040427 Location Cienda, Gabas Elevation [m asl] 130 Coordinate N 10° 43' 54.6'' Coordinate E 124° 48' 45.2'' Major landform High-gradient hill Profile position Middle slope Slope form Slightly concave Slope gradient [%] 50 Orientation W Parent material Andesitic Gen. observations Young colluvial soil

Horizon Depth Tex- Matrix Struc Voids Roots Boun- Observations41 [cm] ture colour -ture (n, ∅) >2mm/dm2 dary Ah 0-6 CL 10YR 3/2 GR C-M; M-C >50 CS Worm casts, Cocos roots AB 6-12 CL 10YR 3/3 GR C-M; M-C >50 GW Cocos roots Bw1 12-32 CL 10YR 4/3 SB M; FF 11-20 DI Mottles (rust), charcoal Bw2 32-62 C 10YR 4/3 SA C-(M); F-M 6-10 CI Charcoal lCw 62-(100) C 10YR 5/4 SG F; V 6-10 Saprolite

41Ah and AB differ mainly in colour, AB and Bv1 in structure, Bv2 and lCv in material and texture.

• 64 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-6 1 26.9 37.8 35.3 0.72 AB 6-12 2 24.5 36.3 39.2 1.05 Bw1 12-32 3-5 24.7 36.0 39.3 1.11 Bw2 32-62 5 22.2 33.1 44.7 1.17 lCw 62-(100) 5 25.7 28.1 46.2 1.14

Horizon pH pH Corg NT C/N PBray II PTruog CEC Ca Mg K Na S BS CaCl2 KCl

0.01M 1M % mg/kg cmolc/kg fine earth % Ah 5.69 5.14 4.65 0.41 11.4 0.98 15.7 48.66 13.83 6.41 2.55 0.22 23.01 47 AB 5.10 4.50 1.74 0.18 9.8 0.26 - 45.20 10.60 5.49 2.02 0.35 18.46 41 Bw1 4.94 4.24 1.97 0.19 10.2 0.45 6.0 43.68 10.74 5.29 1.40 0.37 17.81 41 Bw2 5.16 4.37 1.50 0.15 9.9 0.18 7.0 45.36 12.40 5.73 0.81 0.49 19.43 43 lCw 5.18 4.09 0.45 0.05 8.4 25.33 37.6 39.69 13.35 4.59 0.16 0.92 19.02 48

Horizon Feo Fed Alo Ald Mno Mnd [g/kg] Ah 2.85 20.09 1.85 2.01 1.17 1.45 AB 3.05 22.38 2.30 2.55 1.05 1.47 Bw1 3.38 22.66 2.07 2.40 1.16 1.43 Bw2 3.45 21.63 2.06 2.14 1.08 1.31 lCw 1.69 11.76 2.03 1.30 0.67 0.78

Ecological evaluation Soil depth and rootability: Mechanical depth: >100cm. Physiological depth: Deep due to the soft bedrock. Rootability is very good in Ah and AB (low bulk density, few stones, granular structure), good-moderate in B (subangular structure), moderate in Bv (subangular-angular structure) and poor in the cemented C horizon. Effective rooting space (ERS) 100cm. Air and water budget [%] Total pore volume Air capacity Av. field Field capacity capacity Ah: CL; 8%OM; BD 0.72g/cm3 69 17 (high) 27 52 AB: CL; 3.0%OM; BD 1.05g/cm3 62 16 (high) 24 47 Bw1: CL; 3.4%OM; BD 1.11g/cm3 56 11 (medium) 22 46 Bw2: C; 2.6%OM; BD 1.17g/cm3 61 6 (low) 16 55 lCw: C; 0.78%OM; BD 1.14g/cm3 53 6 (low) 11 47 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is moderately high for all horizons. 2 S value: High: 182.,76 molc/m ERS 2 N supply: High: NT 1.42 kg/m ERS P supply: Low: 10.62g/m2 ERS

Erodibility (8% OM; CL; Aggregate stability high; Ksat very high): Very low.

• 65 • 3 Characterisation of soils

3.1.2 Haplic Cambisol, Cienda PN2 PN1, 2 and 3 are on the same site, 1 located on the upper middle slope position, 2 at the lower middle slope and PN3 at the bottom on a natural terrace. Land-use and present vegetation: Coconut, banana. Relatively open canopy. Pioneers: Low creeping grass, ferns, kudzu. Weather: 90% overcast.

Profile # 2 Date of 040427 description Location Gabas, Sitio Cienda Elevation [m asl] 120 Coordinate N 10° 43' 55.3'' Coordinate E 124° 48' 43.7'' Major landform High-gradient hill Profile position Middle slope Slope form Slightly concave Slope gradient [%] 40 Orientation W Parent material Volcanic General Colluvial observations

Horizon Depth Tex- Matrix Struc- Voids Roots Boun- Observations [cm] ture colour ture (n, ∅) >2mm/dm2 dary Of 1-0 Thin fragmentary litter layer; topsoil with less litter, roots, fauna than PN1. Ah 0-5 CL 10YR 3/3 GR M; F-M 11-20 GS Charcoal AB 5-14 CL 10YR 4/3 SB F-C; F-M 6-10 DS Few mottles (rust), little charcoal Bw 14-39 C 10YR 3/4-4/4 AB F; V 1-5 DW Little charcoal BC 39- C 10YR 4/4 AB F;V 1-5 Mottles (rust) (100)

• 66 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-5 5 28.5 33.8 37.7 1.16 AB 5-14 2 27.4 33.7 38.9 1.26 Bw 14-39 10 23.9 29.7 46.4 1.33 BC 39-(100) 50 20.9 28.9 50.2 1.35

Horizon pH PH Corg NT C/N PBray II PTruog CEC Ca Mg K Na S BS CaCl2 KCl

0.01M 1M % mg/kg cmolc/kg fine earth % Ah 5.30 4.89 3.13 0.34 9.2 0.70 4.7 39.27 10.49 5.39 0.5 0.17 16.54 42 AB 5.06 4.54 1.77 0.20 8.8 0.25 4.2 37.90 9.63 4.35 0.32 0.16 14.46 38

BT 5.07 4.49 0.97 0.12 8.3 0.18 5.3 38.13 10.18 3.93 0.25 0.29 14.65 38 BC 5.14 4.45 0.74 0.09 8.6 0.11 7.7 38.82 10.80 3.90 0.11 0.46 15.27 39

Horizon Feo Fed Alo Ald Mno Mnd [g/kg] Ah 4.06 29.61 1.55 2.97 1.72 1.54 AB 3.66 33.20 1.66 3.40 1.60 1.79 Bw 2.74 34.20 1.82 3.75 1.36 1.68 BC 2.01 35.78 1.63 3.88 0.94 1.35

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth: 100cm. Rootability: Ah good, AB moderate (due to structure), Bv moderate-poor and BC poor (due to angular-blocky structure and 50% rocks in BC). Effective rooting space 100cm. Air and water budget [%] Total pore volume Air capacity Av. field capacity Field capacity Ah: CL; 5.4%OM; BD 12 (medium) 1.16g/cm3 63 25 51 AB: CL; 3.1%OM; BD 11 (medium) 1.26g/cm3 56 22 46 Bw: C; 1.7%OM; BD 1.33g/cm3 51 4 (low) 10 48 BC: C; 1.3%OM; BD 1.35g/cm3 51 4 (low) 10 48 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is moderately high for all horizons except BC (moderate). 2 S value: High: 100.15 molc/m ERS 2 N supply: High: NT 1.11kg/m ERS P supply: Very low: 0.167g/m2 ERS

Erodibility (5.4% OM; CL; Aggregate stability moderate; Ksat high): Low.

• 67 • 3 Characterisation of soils

3.1.3 Stagnic Luvisol, Cienda PN3 PN1, 2 and 3 are on the same site, 1 located on the upper middle slope position, 2 at the lower middle slope and PN3 at the bottom on a natural terrace. Land-use and present vegetation: Open area, approx. 20% canopy cover. Coconut, previously annuals. Imperata and ferns in drier, yellow-flowering Asteraceae in wet parts. Crop residues. Weather: 50% overcast, changing.

Profile # 3 Date of description 040427 Location Gabas, Sitio Cienda Elevation [m asl] 102 Coordinate N 10° 43' 54.3'' Coordinate E 124° 48' 41.7'' Major landform High-gradient hill Profile position Bottom Slope form Terraced Slope gradient [%] 2 Orientation W Parent material Volcanic

Horizon Depth Tex- Matrix Struc- Voids Roots Boun- Observations [cm] ture colour ture (n, ∅) >2mm/dm dary 2

Ah 0-5 C 10YR GR C-M; F; P 20-50 GS 3/3-4/3 AB 5-14 C 10YR 3/ 4 SB C; FF; P 11-20 GS

Stagnic BT1 14-49 HC 10YR 4/4 AB C-F; FF; P 1-5 DI Stagnic, Mn concretions

Stagnic BT2 49-(100) HC 10YR 4/6 AB C; FF; P 1-5 Stagnic, 20% Mn, larger than in BT1

• 68 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-5 1 21.4 33.0 45.6 0.89 AB 5-14 5 19.7 31.6 48.7 1.04

BT1 14-49 15 12.2 21.0 66.8 1.19

BT2 49-(100) 15 10.2 16.7 73.1 1.17

Horizon pH PH Corg NT C/N PBray II PTruo CEC Ca Mg K Na S BS CaCl2 KCl g

0.01M 1M % mg/kg cmolc/kg fine earth % Ah 4.52 4.23 3.97 0.34 11.7 0.73 12.0 38.30 5.30 4.76 0.31 0.11 10.47 27 AB 4.28 4.05 2.30 0.22 10.4 0.33 6.8 32.72 4.29 3.45 0.14 0.14 8.03 25

BT1 4.45 4.13 1.00 0.11 9.4 0.07 3.9 35.58 5.77 2.77 0.04 0.24 8.82 25

BT2 4.80 4.52 0.51 0.05 9.6 0.03 8.2 37.14 6.74 3.46 0.04 0.26 10.52 28

Horizon Feo Fed Alo Ald Mno Mnd [g/kg] Ah 3.38 49.46 2.20 6.54 2.44 2.97 AB 3.69 50.30 2.07 6.50 2.47 2.98

BT1 2.18 58.25 1.85 7.67 2.08 3.65

BT2 1.35 68.93 1.53 8.29 1.30 2.40

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth: Deep (100cm). Rootability: Ah good (low bulk density, granular structure), AB good-moderate (subangular structure), BT1, BT2 moderate-poor (high clay contents, bulk density, stagnic). Presence of very few trees in this part of the plot may indicate problems for deep-rooting plants. Effective rooting space (ERS): 100cm. Air and water budget [%] Total pore Air capacity Av. field capacity Field capacity volume Ah: C; 6.8%OM; BD 0.89g/cm3 73 10 (medium) 25 63 AB: C; 4.0%OM; BD 1.04g/cm3 73 10 (medium) 22 63

3 BT1: HC; 1.7%OM; BD 1.19g/cm 57 6 (low) 13 51

3 BT2: HC; 0.9%OM; BD 1.17g/cm 53 6 (low) 11 47 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is moderately high for all horizons. 2 S value: High: 83.89 molc/m ERS 2 N supply: High: NT 0.99kg/m ERS P supply: Very low: 0.105g/m2 ERS

Erodibility (6.8% OM; C; Aggregate stability moderate; Ksat high): Very low.

• 69 • 3 Characterisation of soils

3.1.4 Dystric42 Nitisol, Cienda (Rainforestation demo plot) Land-use and present vegetation: Rainforestation, densely (2 x 1m) planted in 1994, average height of trees 2004 approximately 8m. Scarce undergrowth: Poaceae in openings; pineapple. Plot size 1ha. Reference plot: Grass and Pueraria 40cm, bush 1m, coconut 8x8m. Weather: Overcast, no rain.

Profile # 4 Date of description 040512 Location Gabas, Sitio Cienda Elevation [m asl] 80 Coordinate N 10° 43' 41.7'' Coordinate E 124° 48' 38.6'' Major landform High-gradient hill Profile position Lower part of Plateau Slope form - Slope gradient [%] 3-5 Orientation - Parent material Volcanic

Horizon Depth Tex- Matrix Struc- Voids Roots Boun- Observations [cm] ture colour ture (n, ∅) >2mm/dm2 dary Ah 0-8 HC 7.5YR3/3 GR-SB M; FF 11-20 D, S AB 8-27 HC 7.5YR3/4 SB M; V 6-10 D, S Bw 27-73 HC 7.5YR4/4 SA M; V 6-10 D, S Mn concretions ∅ BT 73-(100) HC 7.5YR4/4 SA M; V 1-5 Mn concretions ( > in Bw)

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-8 - 12.0 23.8 64.2 0.98 AB 8-27 - 9.4 21.0 69.5 1.15 Bw 27-73 - 7.2 24.6 68.2 1.11

BT 73-(100) - 5.6 14.2 80.2 1.05

42 Formally, this soil would match the category hyperdystric, as BS < 50% between 20-100cm and <20% in some parts of the profile.

• 70 • 3 Characterisation of soils

Horizon pH pH Corg NT C/N PBray PTruo CEC Ca Mg K Na Al S Al- BS 43 CaCl2 KCl II g S

0.01M 1M % mg/kg cmolc/kg fine earth % Ah 4.33 4.20 2.77 0.27 10.2 1.44 9.8 32.25 3.40 3.42 0.17 0.10 0.52 7.09 7 22 AB 4.03 3.90 1.53 0.16 9.8 0.17 11.5 28.12 1.43 1.32 0.03 0.09 1.56 2.87 39 10 Bw 4.08 3.85 0.85 0.09 9.0 0.18 - 35.45 1.81 1.26 0.02 0.13 1.89 3.22 37 9

BT 4.04 3.84 0.67 0.07 9.1 0.18 4.4 31.22 1.56 1.20 0.04 0.14 2.11 2.94 42 9

Horizon Feo Fed Alo Ald Mno Mnd [g/kg] Ah 3.60 49.49 2.12 6.98 2.64 2.56 AB 2.86 50.95 1.88 6.87 2.41 2.83 Bw 1.54 51.15 1.76 6.93 1.27 1.94

BT 1.06 50.04 1.78 6.68 0.88 1.41

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth: Deep (100cm). Rootability: Due to low pH and high Al-saturation, high clay contents and blocky structure restricted below Ah. Effective rooting space (ERS): 100cm. Air and water budget [%] Total pore volume Air capacity Av. field capacity Field capacity Ah: HC; 4.8%OM; BD 0.98g/cm3 73 10 (medium) 25 63 AB: HC; 2.6%OM; BD 1.15g/cm3 61 6 (low) 16 55 Bw: HC; 1.5%OM; BD 1.11g/cm3 57 6(low) 13 51

3 BT: HC; 1.2%OM; BD 1.05g/cm 63 9 (medium) 16 54 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is moderately high for all horizons. 2 S value: Moderate: 33.88 molc/m ERS 2 N supply: High: NT 1.25kg/m ERS P supply: Very low: 0.292g/m2 ERS

Erodibility (4.8% OM; HC; Aggregate stability moderate; Ksat high): Very low.

43 Al3+ saturation as percentage of (S-value + Al)

• 71 • 3 Characterisation of soils

3.1.5 Chromic Cambisol, LSU Land-use and present vegetation: Rainforestation, 15-20m height. Plot size approx. 2ha. Reference plots: a.) Adjacent, but less steep plot recently planted to abaca, yams and sweet potato. Well maintained, bare soil, where not covered by sweet potato. b.) Pasture (grass 40cm, ferns, kudzú) in approx. 70m distance, same exposition and slope. Weather: Overcast, 30h after long heavy rain, 12h after less intense rainfall.

Profile # 5 Date of description 050318 Location LSU Elevation [m asl] Approx. 120 Coordinate N 10° 44' 50.2'' Coordinate E 124° 48' 14.8'' Major landform High-gradient hill Profile position Middle slope, 80m below ridge Slope form Slightly convex Slope gradient [%] 70 Orientation SW Parent material Volcanic General Strong erosion in observations parts of the site (not sampled), observable from stones on topsoil and horizonation of auger cores.

Horizon Depth Texture Matrix colour Structure Roots Boundary Observations [cm] >2mm/dm2 Of 3-0 Litter layer varying from 2-10cm thickness Ah 0-3 CL 7.5YR 3/3 GR 2 G, S AB 3-15 CL 7.5YR 4/4 GR- SB 15-20 G, W 5% rust

Chromic BT 15-60 CL 7.5YR 4/6 SG- SB 1-2 G, S Pseudo sand C 60- SCL 10YR 5/3 (70%), SB 1 (100) 7.5YR 4/6 (30%)

• 72 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-3 < 5 31.9 36.9 31.2 0.95 AB 3-15 5 25.6 35.7 38.7 1.00

BT 15-60 7 30.5 30.6 38.9 0.94 C 60-(100) 80 51.0 25.7 23.3 0.94

Horizon pH Corg NT C/N PBray II CEC Ca Mg K Na S BS CaCl2

0.01M % mg/kg cmolc/kg fine earth % Ah 5.95 2.42 0.23 10.5 1.98 41.83 18.64 8.53 0.69 0.45 28.31 68 AB 4.65 1.24 0.13 9.6 0.41 39.89 9.74 8.25 0.11 0.38 18.49 46

BT 4.85 0.69 0.07 9.6 0.18 42.05 7.74 9.97 0.07 0.34 18.12 43 C 4.70 0.31 0.03 9.0 0.89 51.61 11.90 16.83 0.07 0.35 29.14 56

Ecological evaluation Soil depth and rootability: Mechanical depth: 100cm, physiological depth 60cm due to (even though soft) rock contents in the C horizon. Effective rooting space (ERS): 60cm. Air and water budget [%] Total pore volume Air capacity Av. field capacity Field capacity Ah: CL; 4.2%OM; BD 0.95g/cm3 69 17 (high) 27 52 AB: CL; 2.1%OM; BD 1.00g/cm3 62 16 (high) 24 47

3 BT: CL; 1.2%OM; BD 0.94g/cm 59 15 (high) 23 45 C: SCL; 0.5%OM; BD 0.94g/cm3 56 11 (medium) 24 45 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is high in Ah, moderately high in AB, BT and moderate in the C horizon. 2 2 S value: molc/m for ERS is 105.06 cmol/m (high) assuming an effective rooting space of 60cm. 2 N supply: Moderately high: NT 0.52kg/m ERS P supply: Very low: 0.17g/m2 ERS

Erodibility (4.2% OM; CL; Aggregate stability moderate; Ksat very high): Low.

• 73 • 3 Characterisation of soils

3.1.5.1 Ferri-stagnic Luvisol, Marcos Land-use and present vegetation: Rainforestation, height 15m, few coconut. Undergrowth pineapple, some creeping grasses, 20cm. Plot size 0.3ha. Reference plots: a.) Short grass 10cm, plot used for annuals and/or as pasture, probably burnt in previous years, stones outcropping. b.) Gmelina plot of same size and age as rainforestation, height 20m, undergrowth 70-100cm, dominated by fern. Topsoil notably darker and wetter than under rainforestation.

Profile # 6 Date of description 050225 Location Marcos Elevation [m asl] 30 Coordinate N 10° 45' 55.3'' Coordinate E 124° 47' 26.0'' Major landform High-gradient hill Profile position Middle slope Slope form Straight Slope gradient [%] 60 Orientation W Parent material Andesitic General Aquiferous horizons, observations spring/well nearby.

Horizon Depth Texture Matrix Structure Roots Boundary Observations [cm] colour >2mm/dm2 Of 2-0 Scarce, no cover. Very dry soil between pineapples. Ah 0-4 C 10YR 4/3 SB 80 GS Sw44 4-15 C 10YR 4/3 GR 50 GS Cracks, Mn concretions <1mm, lateral water-flow Ferric 15-40 HC 10YR 5/6 SB 5-10 DW 20% rust, 10% Mn BT -Sw concretions, illuvial, hydromorphic Stagnic 40-65 CL 10YR 5/4 SG+MA 1 DW Sandy andesitic material, Bw 10% rust, 5% Mn concretions

44 Sw is used in the German classification for stagnic horizons, when stagnic properties are caused by infiltration or lateral flow, but not groundwater (gleyic properties).

• 74 • 3 Characterisation of soils

Stagnic 65-(100) L 10YR 5/3 SG+MA 1 Strongly weathered, Mn C <1mm

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-4 - 14.5 31.1 54.4 0.93 Sw 4-15 - 14.6 28.8 56.7 0.98

BT - Sw 15-40 - 6.2 21.7 72.1 0.88 Bw 40-65 10 26.6 34.0 39.5 0.90 C 65-(100) - 49.9 35.8 14.4 0.97

Horizon pH Corg NT C/N PBray II CEC Ca Mg K Na S BS CaCl2

0.01M % mg/kg cmolc/kg fine earth % Ah 5.55 3.50 0.34 10.2 3.10 56.61 33.08 5.72 0.24 0.53 39.56 70 Sw 5.20 1.68 0.16 10.6 0.47 50.44 28.16 5.16 0.14 0.42 33.88 67

BT - Sw 5.40 1.07 0.12 9.2 0.61 54.78 38.70 6.28 0.26 0.63 45.87 84 Bw 5.75 0.40 0.05 8.1 0.95 54.90 35.74 4.76 0.08 0.78 41.37 75 C 6.30 0.05 0.01 6.1 0.7345 33.65 22.05 2.45 0.05 0.54 25.09 75

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth: 65cm, soft (andesitic) rock or cemented material in C horizon. Restricted rootability below BT horizon because of lateral water flow and heavy texture (air budget). Effective rooting space (ERS): 65cm. Air and water budget [%] Total pore Air capacity Av. field capacity Field capacity volume Ah: C; 6.0%OM; BD 0.93g/cm3 73 10 (medium) 25 63 Sw: C; 2.9%OM; BD 0.98g/cm3 67 9 (medium) 19 58

3 BT -Sw: HC; 1.8%OM; BD 0.88g/cm 63 9 (medium) 16 54 Bw: CL; 0.7%OM; BD 0.90g/cm3 56 14 (high) 22 42 C: L; 0.1%OM; BD 0.97g/cm3 55 19 (v. high) 24 35 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is high throughout all horizons except C (moderate). S value is 260.85 cmol/m2 and thus very high, even when assuming an effective rooting space of only 65cm. 2 N supply: Moderately high: NT 0.68kg/m ERS P supply: Very low: 0.491g/m2 ERS, but moderate: 113.75g/m2 ERS, if C is included to ERS.

Erodibility (6.0% OM; C; Aggregate stability moderate; Ksat intermediate): Very low.

45Additional samples (all in duplicate) nearby from the same depth gave 334.34mg/kg, 358.84mg/kg, 0.80mg/kg and 20.83mg/kg, respectively.

• 75 • 3 Characterisation of soils

3.1.6 Ferri-chromic Luvisol, Pangasugan Land-use and present vegetation: Rainforestation, height approx. 15m, mostly native species. Plot size 0.6ha. Reference plot: Short-grass (10cm) pasture with single shrubs. Weather: 25% clouds, no rain for > 48h.

Profile # 7 Date of description 050311 Location Pangasugan Elevation [m asl] 20 Coordinate N 10°45' 15.4'' Coordinate E 124° 47' 35.5'' Major landform Medium-gradient hill Profile position Middle slope Slope form Straight Slope gradient [%] 40 Orientation WSW Parent material Volcanic ashes and (below 160cm) eroded material General Some horizons with observations 20% cutanic or redoximorphic features (MCP) on peds. Ah, AB formed by erosion (stone contents > than horizons below). Aquiferous horizons, spring nearby.

Horizon Depth Texture Matrix Structure Roots Boundary Observations [cm] colour >2mm/dm2 Of 2-0 Leaf cover composed of native species. M146 0-3 C 7.5YR 4/3 GR >10 GS Termites, ants. Boulders. Colluvial origin. M2 3-30 HC 7.5YR 3/3 SB 7-10 CW Ferric, chromic 30-50 HC 5YR 4/6 AB 2 D Few Mn concretions; Bts1 strong red mottles. 20% of aggregates with Ferric, chromic 50- HC 5YR 5/8 AB 1 greyish cover. Bts2 (100)

46 From the German classification: Colluvial mineral soil with organic matter contents similar to those of an Ah.

• 76 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 M1 0-3 10 16.6 28.7 54.7 0.73 M2 3-30 20 13.1 25.4 61.5 1.04 Bts1 30-50 - 7.1 19.1 73.8 0.96 Bts2 50-(100) - 5.3 18.4 76.2 0.95

Horizon pH Corg NT C/N PBray II CEC Ca Mg K Na S BS CaCl2

0.01M % mg/kg cmolc/kg fine earth % M1 5.35 3.26 0.28 11.7 1.40 35.76 7.86 7.79 0.45 0.62 16.72 47 M2 4.70 1.28 0.12 10.4 0.17 29.35 5.11 6.61 0.06 0.46 12.24 42 Bts1 4.30 0.87 0.09 9.4 0.20 28.56 3.82 4.65 0.05 0.29 8.81 31 Bts2 4.25 0.64 0.07 9.1 0.07 27.92 2.87 4.56 0.04 0.20 7.67 27

Ecological evaluation Soil depth and rootability: Mechanical depth: 100cm, physiological depth only to a limited extent below 30cm because of pH gradient and abrupt textural / structural change. Still, effective rooting space was assumed 100cm. Air and water budget [%] Total pore Air capacity Av. field capacity Field capacity volume M1: C; 5.6%OM; BD 0.73g/cm3 73 10 (medium) 25 63 M2: HC; 2.2%OM; BD 1.04g/cm3 67 9 (medium) 19 58 Bts1: HC; 1.5%OM; BD 0.96g/cm3 63 9 (medium) 16 54 Bts2: HC; 1.1%OM; BD 0.95g/cm3 63 9 (medium) 16 54 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is moderate in all horizons. ² ² S value is high (86.92 cmolc /m ) for the entire to 100cm depth and moderate for the upper 3 horizons (48.57 cmolc /m ) as ERS. 2 N supply: Moderately high: NT 0.84kg/m ERS P supply: Very low: 0.139g/m2 ERS

Erodibility (5.6% OM; C; Aggregate stability moderate; Ksat high): Very low.

• 77 • 3 Characterisation of soils

3.1.7 Hypereutric Cambisol, stagnic properties, Maitum Land-use and present vegetation: Rainforestation area naturally terraced by Gmelina. Plot size 0.5ha. Reference plot: Pasture under common use; eroded, short grass 10cm, stone outcrops. Weather: Rain during previous night.

Profile # 8 Date of description 050322 Location Maitum Elevation [m asl] 60 Coordinate N 10° 37' 23.5'' Coordinate E 124° 46' 23.8'' Major landform Medium- to high- gradient hill Profile position Lower middle slope Slope form Concave Slope gradient [%] 15-70 (25 at profile) Orientation NW Parent material Volcanic

Horizon Depth Texture Matrix colour Structure Roots Boundary Observations [cm] >2mm/dm2 Of 4-0 Thick Gmelina litter layer (up to 10cm). Ah 0-1 C 7.5YR 3/2 GR n.d. GS AB 1-15 C 7.5YR 3/2 SA 20 AB 10% orange Stagnic 15-60 C s. AB-MA 5 AB 50% grey, 35% orange, Sw147 observations 15% stone Sw2 60-(80) HC (est.) s. AB-MA n.d. 55% orange, 20% grey, observations 20% stone, 5% black Explanations on colours: 'orange' = 10YR 6/8, sandy, weathered parent material with iron oxides 'grey' = 2.5Y 4/2, HC 'black' = illuvial humic substances, HC 'stone' = 10YR 8/2, weathered parent material, friable, no reaction with 10% HCl.

47 Sw is used in the German classification for stagnic horizons, when stagnic properties are caused by infiltration or lateral flow, but not groundwater (gleyic properties).

• 78 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ah 0-1 2 19.0 35.8 45.2 n.d. AB 1-15 < 5 13.3 36.4 50.4 1.00 Sw1 15-60 15 24.3 31.8 43.9 0.98 Sw2 60-(80) 20 n.d. n.d. n.d. n.d.

Horizon pH Corg NT C/N PBray II CEC Ca Mg K Na S BS CaCl2

0.01M % mg/kg cmolc/kg fine earth % Ah 6.15 3.94 0.30 13.3 65.40 59.48 37.68 9.43 0.84 0.40 48.34 81 AB 5.60 1.78 0.13 13.8 7.98 53.44 33.63 9.12 0.20 0.44 43.39 81 Sw1 6.75 0.20 0.02 12.8 4.04 66.84 48.27 10.60 0.13 1.31 60.30 90

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth 60cm. Rootability impeded by clayey and cemented material below 15cm. Effective rooting space (ERS): 60cm. Air and water budget [%] Total pore volume Air capacity Av. field capacity Field capacity Ah: C; 6.8%OM; BD approx. 1g/cm3 73 10 (medium) 25 63 AB: C; 3.1%OM; BD 1.00g/cm3 67 9 (medium) 19 58 Sw1: C; 0.3%OM; BD 0.98g/cm3 59 9 (medium) 14 50 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is high in all horizons. 2 S value: Very high (293.7 cmolc / m for 60cm ERS) 2 N supply: Moderate: NT 0.26kg/m ERS P supply: Very low: 3.207g/m2 ERS

Erodibility (6.8% OM; C; Aggregate stability high; Ksat intermediate): Low.

• 79 • 3 Characterisation of soils

3.1.8 Stagnic Cambisol, Patag Land-use and present vegetation: Rainforestation 10-15m height, sparse undergrowth. Plot size 1ha. Reference plots: Neighbouring Gmelina plot, 20m height, with some native species planted later and dense fern undergrowth. Low grass, bush fallow, bare soil. Cleared 1-2 years ago for construction of transmission line.

Profile # 9 Date of description 050418 Location Patag Elevation [m asl] 40 Coordinate N 10°44' 10.5'' Coordinate E 124° 48' 15.5'' Major landform High-gradient hill Profile position Middle slope Slope form Straight Slope gradient [%] 75 Orientation WNW Parent material Volcanic General Erosion visible on observations the surface; Ah eroded, profile truncated and new material deposited.

Horizon Depth Tex- Matrix Struc- Voids Roots > Boundary Observations [cm] ture colour ture (n, ∅) 2mm/dm2 Of 1-0 Thin fragmentary litter layer composed mainly of dipterocarp leaves. AB1 0-17 SiC 7.5YR 3/4 GR- SB M; V 5 DS Structure smeary under pressure. Grey eroded stones. AB2 17-35 SiC 10YR 4/4 SB M; F 2 DS Grey eroded material. Bw 35-70 C 10YR 4/6 SA C; F 2 CW Weathered material, outsides rusty. Mn concretions. Bw 70-(100) C 10YR 5/6 SB F; nd 1 5% rust, Mn concretions <1mm.

• 80 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 AB1 0-17 <5 15.4 42.0 42.7 1.07 AB2 17-35 5 17.0 42.4 40.5 1.14 Bw 35-70 15 19.0 35.1 46.0 1.09 Bw 70-(100) <5 12.2 38.5 49.3 0.98

Horizon pH Corg NT C/N PBray II CEC Ca Mg K Na S BS CaCl2

0.01M % mg/kg cmolc/kg fine earth % AB1 4.43 1.53 0.14 10.9 0.36 37.07 10.01 7.33 0.07 0.30 17.72 48 AB2 5.20 0.93 0.10 9.7 0.30 41.44 13.53 8.45 0.05 0.31 22.34 54 Bw 5.61 0.52 0.06 8.4 0.15 45.53 18.94 9.33 0.05 0.46 28.78 63 Bw 4.94 0.44 0.05 8.4 0.29 47.98 21.38 10.68 0.05 0.65 32.76 68

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth: 100cm. Rootability poor in stagnic horizons. Effective rooting space (ERS): 100cm. Air and water budget [%] Total pore volume Air capacity Av. field capacity Field capacity AB1: SIC; 2.6%OM; BD 8 (medium) 1.07g/cm3 66 25 57 AB2: SiC; 1.6%OM; BD 1.14g/cm3 54 5 (low) 19 50 Bw: C; 0.9%OM; BD 1.09g/cm3 59 9 (medium) 14 50 Bw: C; 0.8%OM; BD 0.98g/cm3 59 9 (medium) 14 50 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is moderate from Ab to BT and high in Bv. 2 S value amounts to 247.09 cmolc / m for 100cm ERS (very high). 2 N supply: Moderate: NT 0.78kg/m ERS P supply: Very low: 0.249g/m2 ERS

Erodibility (2.6% OM; SiC; aggregate stability moderate; Ksat high): Low.

• 81 • 3 Characterisation of soils

3.1.8.1 Calcari-Mollic Leptosol, Punta Land-use and present vegetation: Rainforestation mixed with Swietenia sp. and Gmelina sp., 10-15 (20)m high. Plot size 5.4ha, sampled portion mainly consisting of native species. No undergrowth. Reference plot: Reference 20m upslope, above a fertile plateau; not under agricultural use. Cocos, Imperata and ferns 30-40cm high. Has been burnt in the past. Weather: Overcast and rainy after 3 dry weeks (soil extremely dry).

Profile # 10 Date of description 050524 Location Punta Elevation [m asl] Approx. 80 Coordinate N 10° 37' 47.5'' Coordinate E 124° 46' 58.9'' Major landform High-gradient hill Profile position Footslope-middle slope Slope form Concave Slope gradient [%] 50 (upper part - 150) Orientation WNW Parent material Coralline limestone General Entire slope formed observations by landslide

Horizon Depth Tex- Matrix Struc- Voids Roots Boun- Observations [cm] ture colour ture (#, size) >2mm/dm2 dary Of 2-0 Slowly decomposable litter of native species (in contrast to exotics farther downhill). Ap 0-6 HC 10YR4/2 GR, M; F 11-20 GS Ants; channels by soil fauna crusted at surface. Stones < in Ap. Ap 6-20 HC 10YR4/2 GR, M; F 6-10 CS Termites. Less and smaller crusted stones than in Bv. Bw 20-38 HC s. below1 M; F 3-5 GW Some stones covered with rust. BC 38-47 HC s. below2 C; V 1-2 AI C 47-(60) 10YR8/3 MA - 1 Colour and structure: Matrix (45%) 10YR5/3, GR-SB; 35% 10YR7/6, GR cemented calcareous boulders; 20% 10YR3/2 AB, illuvial clay. 2 Colour and structure: Matrix 10YR7/6 (70%); 20%10YR5/3; 10% 10YR3/2.

• 82 • 3 Characterisation of soils

Horizon Depth Rocks ∑ Sand ∑ Silt Clay Bulk density cm % % of fine earth g/cm3 Ap 0-6 5 5.0 11.1 83.9 0.94 Ap 6-20 20 4.3 9.9 85.7 1.03 Bw 20-38 35 3.8 11.8 84.5 0.94 BC 38-47 40 2.6 16.9 80.5 n.d. C 47-(60) 100 n.d. n.d. n.d. n.d.

Hor. pH NT CT Corg Ccarb XCO3 PBray II CEC Ca Mg K Na S BS CaCl2

0.01M % mg/kg cmolc/kg fine earth % Ap 7.35 0.41 7.73 4.65 3.08 25.61 7.74 75.80 69.36 3.93 0.48 0.20 73.97 98 Ap 7.48 0.19 6.08 1.91 4.17 34.73 0.98 65.42 67.28 1.32 0.11 0.12 68.84 100 Bw 7.44 0.16 5.35 1.10 4.20 35.01 0.60 71.18 68.23 0.64 0.09 0.09 69.05 97 BC 7.49 0.09 7.67 1.14 6.53 54.36 0.81 36.69 45.16 0.36 0.04 0.05 45.62 100

Ecological evaluation Soil depth and rootability: Mechanical and physiological depth are 47cm. Rootability below 38cm strongly restricted by cemented material. Thus, effective rooting space was assumed to be only 38cm. Air and water budget [%] Total pore volume Air capacity Av. field capacity Field capacity Ap: HC; 8.0%OM; BD 0.94g/cm3 76 10 (medium) 28 66 Ap: HC; 3.3%OM; BD 1.03g/cm3 67 9 (medium) 19 58 Bw: HC; 1.9%OM; BD 0.94g/cm3 65 9 (medium) 18 56 BC: HC; 2.0%OM; BD ≈1g/cm3 65 9 (medium) 18 56 Nutrient status (considering rocks, bulk density, thickness of horizons and effective rooting space): CEC (cmolc/kg clay) is high from Ah to Bv and moderate in BC. ² S value over an ERS of 38cm is 200.05 cmolc / m², for 47cm ERS would be 224.69 cmolc / m (both very high). 2 N supply: Moderately high: NT 0.61kg/m for 38cm ERS P supply: Very low: 0.593g/m2 for 38cm ERS

Erodibility (8.0% OM; HC; aggregate stability high; Ksat very high): Very low.

• 83 • 3 Characterisation of soils

3.2 Synopsis and Discussion Soil Profiles

Profiles presented in the previous section will be discussed in the light of their parent material, soil forming processes and topography. Magnitudes of values for soil parameters will be compared to ranges given by other authors and methods will be discussed.

3.2.1 Parent Material Leyte forest soils can be divided into two major groups with respect to parent material. These are volcanic rocks and calcareous sediments. Among the presented, only the Punta Leptosol developed on calcareous material, namely coralline / marly limestone. In the Maitum Cambisol, calcareous and igneous rocks coexist, possibly due to the influence of volcanic ashes as has been hypothesised for some Punta soils by ASIO ET AL. (2006). Soils at all other sites were formed from basaltic volcanic rock, mainly Andesites and Saprolite. In the landscape, calcareous terrain often appears more gentle than the rugged volcanic mountains. On the plot, outcrops can provide information on the parent material. Greyish, white and dark brown colours dominate the profile wall, while the volcanic soils tend towards yellow and reddish colours. Punta profile is shallow and its B horizon cemented showing massive structure. Texture is heavy clay (>80%) throughout the profile, in contrast to most volcanic soils, which have at least loamy topsoils. Organic matter contents do not differ significantly between Punta and any of the volcanic soils, nor do -3 pore volumes. Bulk density is 0.94 to 1.03g cm , which is less dense than what ASIO ET AL. (2006) found for Punta limestone soils (1.25-1.51g cm-3). The most conspicuous chemical criterion to differentiate calcareous from volcanic soils is pH (fig. 30), which leads to a strongly acidic (red lines), a moderately acidic (yellow lines) group, Maitum and Punta categories. For volcanic soils in Leyte, ZIKELI (1998) measured pH-values between 4 and 5 which generally increased in C horizons. This was confirmed for all profiles with a C horizon within 100cm depth. Maitum is a transition zone where coralline limestone, andesite and tuff have been superimposed as well as mixed. The lowest accessible layer in the Cambisol is volcanic material, but pH in the solum is clearly higher than for any volcanic soil. So are basic cations and CEC (see fig.33). This may be due to limestone or dissolved secondary lime. Strong differences in the profile with respect to texture Figure 30: pH of profiles in Leyte occur mainly in the sand fraction. Clay contents are similar to a colluvial volcanic profile.

• 84 • 3 Characterisation of soils

3.2.2 Formation of volcanic soils Loss of bases, acidification and formation of Fe-oxides have been described as principal drivers for soil genesis on volcanic bedrock in Leyte (ASIO 1996, JAHN & ASIO 1998). Consequently, older and more weathered soils will be more acidic than young undeveloped stages. On the other hand, landslides as a natural phenomenon have always disturbed in situ soil development through truncation of upslope profiles or burying at the footslopes. The studied volcanic soils can be categorised into two groups: Yellowish soils of 10YR hue (Munsell soil colour charts), which are of colluvial nature and still in an early phase of development. Goethite is the predominant form of iron responsible for the yellow colour. Haplic Cambisol PN1, Haplic Cambisol PN2, Ferri- Stagnic Luvisol Marcos and the Stagnic Cambisol in Patag form part of this group. Reddish (7.5 and 5YR) colours in soils are characteristic for hematite as dominant form of iron. The Stagnic Luvisol and Dystric Nitisol in Cienda, the Chromic Cambisol at LSU and Ferri-Chromic Luvisol Pangasugan are examples for this type of soils. The differentiation of yellowish and reddish soils coincides with the grouping after pH (fig. 30). Reddish soils are the more acidic and the older ones. According to ASIO (1994) red colours also indicate degradation, where the topsoil has been eroded. Similar distinction has been made for Ferralsols, where the red soils are found on plateaus and yellow ones along the slopes (SCHACHTSCHABEL ET AL. 1992). Mixed forms occur, where hematite has formed under warm and dry conditions during prehistoric times. As climatic conditions turned more humid, hematite was dissolved. Ferrihydrite and hematite could not be formed due to climate and organic matter, so that goethite dominates the upper part (ASIO 1996). In the case of the Ferri-chromic Luvisol in Pangasugan, erosion has caused the subdivision into 5YR subsoil and superimposed 7.5YR material. Apart from hue, this leap can be observed in an abrupt decrease of sand and silt and increasing clay, and at the same time a drop in pH and base saturation. Another factor that changes during ageing of volcanic soils, is bulk density. Supposing a chronosequence Andosol → Cambisol → Luvisol → Acrisol / Alisol, as suggested by ASIO (1996), Andosols as starting point have extremely low bulk densities, which is one -3 classification criterion. ZIKELI (1998) found values as low as 0.34 - 0.6g cm in Andosols at Mt. Pangasugan. Although significant differences in bulk density could not be observed among the studied soils with respect to age, the two eroded profiles, PN2 and Patag, were the only soils with BD >1 in the top horizon48.

3.2.3 Topography On volcanic grounds, mass fluxes along the slopes play an important role for arrested soil development. This will be illustrated for the example of Cienda profiles. Along the toposequence PN1-3 in Cienda, it is PN1 Cambisol, located on the upper middle slope, which shows the least developed solum with parent material present at 62cm depth. For PN2 Cambisol at the middle slope, the parent material rests below 100cm depth, but from 14 cm downwards rocks have been mixed with the solum through landslides. At PN3 (footslope), rocks still occur in the B horizon, but to less extent and generally smaller in size. Cienda demo plot on a plateau has not been affected by these movements and is deeply weathered with stone contents close to zero. With respect to the topsoil, apart from landslides, man-made erosion has contributed its share to mass fluxes. Comparing topsoil colour as an indicator of humification, land use over the last decades is reflected. PN1 under secondary forest and PN3 at the footslope, 48 For Maitum, BD sampling of the top horizon was not possible due to only 1cm thickness.

• 85 • 3 Characterisation of soils where transported material accumulates, have 12 and 14cm of dark humic topsoil (Munsell value 3), while at PN2, in the middle of the slope and under banana for decades, thickness of dark humic horizons has not surpassed 5cm (fig.31). For PN2, lower contents in organic matter go hand in hand with lower aggregate stability and higher bulk density, reinforcing erodibility through reduced water infiltration. Soils on the slope are in arrested development through the repeated or continuous downhill transport of material and with the topsoil the most fertile portion of the soil is carried downhill and accumulated at the footslope.

Figure 31: Topsoil contents of different parameters along a slope in Cienda. Munsell value 3 refers to thickness of humic horizons in [cm]

Contents of topsoil clay, FeD, AlO/D, MnO/D steadily increase downslope (fig.34), while 49 parameters influenced by vegetation (Corg , PBray) first decrease from forest-like PN1 to banana PN2 and then increase again at PN3 (fig. 31). Contrarily to what was expected, topsoil silt as the most erodible textural fraction decreases downhill (fig.32).

49 Section 4.2.1 will show, that Corg contents on a small scale are a function of vegetation rather than slope.

• 86 • 3 Characterisation of soils

Haplic Cambisol PN1 Haplic Cambisol PN2

Stagnic Luvisol PN3 Dystric Nitisol DS

Figure 32: Soil texture of Cienda profiles

Subsurface water flow carries clay, basic cations and pedogenous oxides (SCHACHTSCHABEL ET AL. 1992) and dissolved organic carbon (DOC). ZÖFEL (2004) found DOC contents of 183mgC kg-1 in a rainforestation soil at LSU equivalent to >300kg DOC ha-1. ZECH ET AL. (1997) state that DOC fluxes can amount to 5-60kg ha-1 a-1. Even P, which is not leached in its inorganic form, can be easily displaced as Porg, especially in neutral hydrophobic dissolved organic matter fractions (DONALD ET AL. 1993). Water flows on top of the clayey Bw layers (s. PN1 and 2) and gives PN3 its stagnic properties. This could be observed at PN3 during the rainy season 2006 even days after rainfall events and is also indicated by the presence of Wedelia biflora, which usually grows along creeks. Illuvial processes concern vertical flows in profiles. In situ clay formation cannot be easily distinguished from clay illuviation, unless clear indicators like clay cutans or bleached horizons are observed. As mentioned, PN1 and PN2 are colluvial and disturbed, while PN3 has developed for a longer time. Clay formation and accumulation, loss of bases and decline of pH in Cienda are more advanced towards the deeper horizons in each profile

• 87 • 3 Characterisation of soils and between profiles from PN1 to DS (fig.33).

pH-related parameters PN1 pH-related parameters PN2

pH-related parameters PN3 pH-related parameters DS

Figure 33: Depth functions of pH, CEC, S value, base saturation, organic carbon and clay contents in profiles along the toposequence PN1-3-DS at Cienda

Like clay minerals, sesquioxides are formed during weathering and also accumulate due to downhill flow. Concentrations of mobile and crystalline forms of pedogenous oxides can

• 88 • 3 Characterisation of soils provide information on the different processes. Oxalate extracts the mobile or 'active' fraction of Fe, Mn and Al oxides such as ferrihydrite, allophane or organo-complexes. FeO and AlO represent the younger fractions, which have not yet formed crystalline bonds as is the case during clay formation. Dithionite-citrate extraction is employed to quantify crystalline pedogenic forms of Fe, Mn and Al. For FeD, this includes more stable oxides such as goethite, hematite and lepidokrokite, but not pyrogenic oxides like magnetite. Concentrations of Fe, Al and Mn extracted with oxalate and dithionite are shown in fig.34 in context with clay contents.

Fe Mn Al and Clay contents PN1 Fe Mn Al and Clay contents PN2

Fe Mn Al and Clay contents PN3 Fe Mn Al and Clay contents DS

Figure 34: Clay contents, oxalate- and dithionite-extractable fractions of Fe, Mn for Cienda profiles

FeD and AlD follow almost identical trends throughout the Cienda profiles: Contents

• 89 • 3 Characterisation of soils increase substantially from PN1 and PN2 to PN3 and DS, and in each profile concentrations shift downwards parallely to the formation of a BT horizon. Elevated FeD and MnD contents often coincide with maxima of clay in BT horizons (SCHLICHTING ET AL. 1995). Parallel trends of clay contents and FeD as in PN2 and PN3 are interpreted as 50 comigration and of dominating illuvial clay over in situ formation (ASIO 1996) . MnD accumulates in the BT horizon of PN3. In DS, contents of FeD and AlD are stable throughout the profile. Magnitudes in PN3 and DS are very similar to those found in a Baybay Alisol by ASIO (1996) for FeD (5-6%) and AlD (0.7-0.6%). For Mn, the range is from about 0.08 to 0.36%, beyond the 0.01-0.02% found by ASIO (1996). Still, Mn might have been underestimated as soils were sieved to 2mm and concretions were not crushed. In contrast to the dithionite-soluble fraction, translocation of FeO can extend even below a BT. This could be observed in all Cienda profiles. Contents of FeO and MnO in the topsoil increase from PN1 to PN3, probably due to erosion, and decrease in the subsoils, due to illuviation and possibly due to crystallisation into more stable minerals like goethite for iron (ASIO 1996). AlO remains relatively stable over profiles and horizons with a slight decrease in subsoils. Magnitudes for FeO and AlO are around 0.1-0.4% and thus far below the 8%, that ZIKELI (1998) found in Andosols at Mt. Pangasugan. None of the Cienda soils would meet the criteria of Andosols according to the WRB classification (AlO + 0.5 FeO > 2%; FAO-ISSS- ISRIC 1999). All Cienda soils have already passed the Andosol stage in a topo- / chronosequence Andosol → Cambisol (PN1, 2) → Luvisol (PN3) → Acrisol / Alisol (ASIO 1996). MnO and MnD values are generally similar in magnitude (SCHLICHTING ET AL. 1995), a typical MnO/MnD ratio in Cienda being around 80%. For Al and Fe, ratios differ considerably, decreasing along the sequence and downwards inside most profiles (tab.7). Generally, a low ratio of oxalate- to dithionite- extractable Fe and Al indicates higher shares of crystalline pedogenous oxides characteristic for more developed soils. FeO/FeD-ratios are around 0.1, higher than those found by ASIO (1996), but different to ratios of 1 assumed by SCHLICHTING ET AL. (1995) and found by ZÖFEL (2004) for LSU soils. Table 7: FeO/FeD- and AlO/AlD-ratios in Cienda profiles.

PN1 FeO/FeD AlO/AlD PN2 FeO/FeD AlO/AlD PN3 FeO/FeD AlO/AlD DS FeO/FeD AlO/AlD Ah 0.14 0.92 Ah 0.14 0.52 Ah 0.07 0.34 Ah 0.07 0.30 AB 0.14 0.90 AB 0.11 0.49 AB 0.07 0.32 AB 0.06 0.27

Bw1 0.15 0.86 Bw 0.08 0.48 BT 0.04 0.24 Bw 0.03 0.25

Bw2 0.16 0.97 BC 0.06 0.42 BT -Go 0.02 0.19 BT 0.02 0.27 lCw 0.14 n.d.

In summary, Fe, Mn and Al concentrations confirm, that PN3 and DS are advanced in development and that erosion and subsurface fluxes play an important role along slopes. Soil formation is most advanced in DS, the dystric Nitisol51, which is deeply weathered, acidic, poor in bases and clayey in the subsoil.

50 The Cw horizon at PN1 mainly consists of parent material and is not taken into account. 51 JAHN & ASIO (1998) classified a similar Baybay soil as Alisol. Nevertheless, Al-saturation for Cienda DS was <60% and a decrease of clay below 100cm was not assessed.

• 90 • 3 Characterisation of soils

3.2.4 Single parameters compared across all study sites S-value and base saturation increase with pH across all soils. CEC and pH are also positively correlated for the studied soils (fig.35), but the correlation is less distinct. This can be explained from the influence of acidic humic substances, which are also connected with high CEC.

Figure 35: Correlations of pH with CEC, S-value and base saturation in Leyte soils

Clay contents per se allow a distinction of 10YR and 7.5YR groups again, with higher clay contents throughout the profiles and stronger increase in the BT of the red group. However, LSU Cambisol and Marcos Luvisol (fig.36) are exceptions, which do not fit into the scheme. In the case of Marcos Luvisol high clay contents in the BTSw may be a result of lateral water flow. There is no clear relationship between clay contents and CEC (r2 = 0.09), even if Ah and AB horizons, which are often low in clay but high in CEC and would distort any trend, are omitted (overall correlation r2 = -0.73, but for individual profiles r2 varies strongly). CEC of fine earth shows (fig. 37a), that the more weathered reddish profiles are lower in CEC than the colluvial than the calcaric ones. As for clay, LSU and Marcos do not match the categories. Looking at CEC per kg clay (fig.37b), these two are very similar and form an Figure 36: Clay contents and distribution in Leyte profiles extra group apart, which seems to originate from similar parent material.

• 91 • 3 Characterisation of soils

CEC per kg fine earth CEC per kg clay

Figure 37: CEC per fine earth and per kg clay in Leyte profiles On the other hand great differences exist with respect to present cation loads and base saturation: Ca/Mg-ratios in the LSU Cambisol decrease from about two to 0.7 towards the subsoil, while they increase from six to nine in Marcos Luvisol. Among the other volcanic profiles, Ca/Mg-ratio is around one to two, in Maitum about four and in Punta 18 (topsoil) to 125 in limestone (probably overestimated because of free CaCO3 in the solum, s. ASIO ET AL. 2006). Base saturation (fig.38) underlines this trend with Marcos Luvisol tending towards higher values and LSU among the colluvial soils. The striking difference in soil colour in Pangasugan ferri-chromic Cambisol is reflected (but not caused) by lower BS in the subsoil. As for the data presented by ZIKELI (1998) for Andosols, all CEC values are above 25cmolc kg-1 fine earth, but towards the upper end, the more developed volcanic soils analysed -1 during this study reach about 55cmolc kg . With respect to base saturation for the volcanic 52 soils, DS Nitisol is similar to the Baybay Alisol studied by ASIO (1996) and Andosols in ZIKELI (1998) are in the range of Marcos Luvisol. For most soils, BS increases towards the parent basaltic/calcareous material.

52 BS = 28 to 6% from top to bottom

• 92 • 3 Characterisation of soils

Figure 38: Base saturation of Leyte profiles

Not only for classification purposes (Alisol vs. Nitisol), it is of importance, that Al-saturation even in the most acidic DS Nitisol is below levels found by ASIO (1996), where Al has by far the highest share among all cations. Due to pH clearly below 5, Al-toxicity may be problematic for plants. PAGEL ET AL. (1982) state that S-values alone do not provide too much information on a soil as long as the saturation of each basic cation is not calculated. High saturation of a cation implies high risk of leaching and also displacement of other cations from the exchange sites. Substitution of Mg and K through Ca certainly occurs in Marcos subsoils, Maitum and Punta. Constraints of soils due to low contents of cations will be discussed for each soil under 3.2.6. -1 Potassium contents are below 0.9cmolc kg and K-saturation is below 2% (very low to medium for some topsoils) for all profiles except PN1 with high contents (2.55 to 0.8cmolc and 5.2 to 1.8% saturation from Ah to Bw2). Levels in Punta are clearly lower than those found by ASIO ET AL. (2006) for the same site. Sodium levels are all low (0.1-2% of CEC) and, especially in the topsoils, appear to be related to seawater spray, as they increase with vicinity to the sea and with W exposition. Land use and present vegetation play a key role for soil organic matter (see KELLMAN 1970) and shall be discussed with more detail in chapters 4 and 5. In general, unfavourable conditions for decomposers, e.g. alkalinity and excessive drainage on limestone or free 3+ Al ions (VELDKAMP 1994) in strongly acidic soils, can lead to humus accumulation. For soils in Sumatra, V. NOORDWIJK ET AL. (1997) found that pH <5 and >6, dry or anaerobic conditions and high clay contents enhanced Corg levels. However, in this study correlation of pH and Corg was weak (r = 0.18) as pH is only one among several factors influencing Corg contents. Mean topsoil Corg was 3.4%, ranging from 4.65% in Punta, where alkaline conditions hamper SOM decomposition (ASIO ET AL. 2006), and the Haplic Cambisol under secondary forest in Cienda to 1.53% for the truncated profile in Patag. Clay contents exercise an indirect influence on decomposition of OM through the formation of organo-mineral complexes (GAUNT ET AL. 2000, BALESDENT ET AL. 1996). CHENU ET AL. (in REES ET AL. 2001)

• 93 • 3 Characterisation of soils found, that organo-mineral aggregates in a French pine forest drastically decreased during cultivation of 35 years, leading to increased wettability of clay and thus to higher rates of erosion. In reverse, organic matter protected by minerals will become exposed to microbial attack once the aggregates are destroyed. The effect of erosion and accumulation at footslopes on Corg contents can be seen in fig.31. An impact of strongly acidic or alumic environments, that would protect SOM (ZIKELI 1998) could not be deducted from the two sufficiently acidic examples (Pangasugan Cambisol, Cienda Nitisol). From Ah to the underlying horizons, Corg contents dropped to about 50% in all profiles except Patag and so did NT. Subsoil contents of Corg were between 0.3 and 0.7% in all soils except Marcos (0.05%, C horizon) and Punta (Corg 1.14%, CCarb 6.53%; C:N is higher than in the other horizons, which might imply an overestimation of Corg). This is close to results of ASIO (1996). Parallel trends can be seen for NT contents: High values (0.41%) for PN1 and Punta, 0.23-0.34% for all other Ah and a leap from Ah to AB horizons. Subsoil NT contents ranged from 0.01 to 0.07% for all profiles except PN1 and Punta (0.09%). This is in the range of N contents in Pangasugan Andosols studied by ZIKELI (1998) for subsoils and about double of her topsoils. According to PAGEL ET AL. (1982), 0.41% NT is a maximum expected for soils of the Humid Tropics, while a mean would be around 0.16%. PAGEL ET AL. (1982) state that Andosols are rich in nitrogen, and for the studied soils, atmospheric inputs and land use are supposed to have contributed additional N (JAHN 1998). Available phosphorus is a limiting factor for many tropical soils. It is determined mainly by parent material and degree of weathering; at similar stages of development, acidic soils tend to be comparatively lower in available P than others (PAGEL ET AL. 1982). In soils with low pH (<5), P is adsorbed by Fe, Al and clay minerals or fixed, either by Fe or Al ions. In calcareous soils P is fixed by Ca ions (forming apatites). Generally, P-sorption is more reversible and thus less problematic than fixation, which is typical for Andosols and Alisols in Leyte (ASIO 1996). An important share of available P is bound in organic molecules. While PI is generally not leached in significant amounts, Porg has been found to be more susceptible than C and N (SCHOENAU & BETTANY 1987 for a boreal soil). GOLLER ET AL. (2006) found that 2/3 of all P leached in an Ecuadorian montane forest was organic. Losses of P through erosion, leaching and slash and burn have been found to make up for 70-80% of total P losses including harvest in Indonesian timber plantations (MACKENSEN ET AL. 2003). For AB and B horizons of the studied profiles, higher contents of available P with increasing pH could be observed in the pH - range 4 to 7 (fig.39). The maximum P contents of AB and B horizons (7.98 and 4.04mg kg-1, encircled) were found in the Hypereutric Cambisol, Maitum, at pH 5.60 and 6.75). -1 All other PBray values >1mg kg in fig.39 were measured in Ah and C horizons, with litter or rocks as sources. Generally, PBray decreases from the Ah horizon towards the AB and B and increases again where the parent material is not weathered strongly. An exception is Maitum, where all horizons show relatively high contents.

• 94 • 3 Characterisation of soils

Extraordinarily high PBray contents were observed in samples from C horizons in the upslope profile PN1 in Cienda and in Marcos; these were 25 and >300mg kg-1, respectively. Repeated sampling in Marcos revealed a high small-scale variability in 65-100cm depth, with PBray of 359, 334, 21 and 0.8mgP kg-1. It seems that tree roots are able to penetrate the stagnic clayey (72%) BTSw, tap the P resources and return them to the topsoil, which shows PBray values three times above those in the other volcanic soils. For PN1 Cambisol, DS Nitisol and LSU Cambisol, accumulation of P in the topsoil can Figure 39: Available P (Bray II) of Leyte profiles, all be observed to some extent, too. horizons, plotted against pH This shows the importance of a close P-cycle to avoid that P is adsorbed and then fixed by minerals. Depending on vegetation, a large proportion of plant P can be supplied by organic P (GAISER 1993). The proportion of Porg increases with weathering and SOM can contain 40-90% of total P in ferralitic soils of humid regions (PAGEL ET AL. 1982). For Cienda profiles, Porg, calculated as difference of PTruog in digested and non-digested samples was between 85 and 90% with exception to the C horizons of the PN1 Cambisol and the Cienda Nitisol. ZECH ET AL. (1997) estimate, that 20-75% of P reserves in tropical soils are stored in topsoil SOM, 60-80% of them as Porg. Generally all measured phosphorus contents save the mentioned exceptions are extremely low, about one tenth of what is considered an average level in soils of the -1 -1 Humid Tropics: PAGEL ET AL. (1982) give typical values of 750mg kg total P and 27mg kg available PBray, which is 300% of the Maitum soil. For soils of Andosol chronosequences, -1 -1 contents <40mgPBray kg and for calcaric soils <20mg kg are considered low. For P -1 extracted after Truog's method (0.002N H2SO4) <30mg kg for soils rich in Fe and Al and -1 <80mg kg for calcareous soils have been defined as low (PAGEL ET AL. 1982). -1 ZIKELI (1998) found levels of available PBray of 0.4-1mg kg and negative correlations to Fe and Al in Andosols between 350 and 550m asl on Mt. Pangasugan slopes; there were no clear tendencies of accumulation or depletion within the profiles. ASIO (1996) determined -1 PBray of two Alisols under forest in Baybay and measured 0.1 to 0.6 mgP kg with maxima in Ah, BT and BC horizons. ASIO ET AL. (2006) classified calcaric soils in Leyte as Calcaric Phaeozems (upper positions) and Calcaric Cambisols (middle and lower slope). They found, that solum thickness differed on a small scale and common limitations in nutrients were often aggravated through shallow rooting space. In contrast, high levels of basic cations and moderate nitrogen supply were measured in Punta, the only serious constraint in nutrients being phosphorus, as in all former Leyte forest soils. Advantages for plant production due to favourable chemical properties are outweighed by the shallow effective rooting space and excessive drainage, which can lead to drought stress during periods of scarce rainfall. This fact, temporarily, and the high pH can be the causes for strong humus accumulation. DAUB (2002) found biological activity of

• 95 • 3 Characterisation of soils the mesofauna in Punta significantly higher than at LSU. Animals are able to evade into deeper soil layers during dry periods and are better adapted to higher pH than most bacteria. Microorganisms, which are responsible for the major part of decomposition, would be more affected by adverse environmental conditions.

3.2.5 Water Balance Favourable water capacity as well as drainage have been observed for volcanic soils in Baybay (JAHN & ASIO 1998). Even during dry season (March 2005), water contents of soil samples from grassland53 were still relatively high. On the other hand, water stress shown by rolled leaves was observed during dry season even for wild plants under closed canopy. In clayey soils, water contents may be relatively high but bound in fine pores and thus not available for plants. In 2006, water contents (WC) and hydraulic or soil water potential (ΨH) measurements were conducted using Frequency Domain Reflectometry (FDR) sensors for WC and tensiometers for ΨH. Two permanent plots were installed in Cienda at middle and footslope position (next to profiles PN1 and 3). The first plot was under closed canopy and the second in open grassland (average height of vegetation 30cm). Sensors and tensiometers were installed on Feb 13th after four days of continuous rain, assuming that this represented a soil status of saturation. A few days with light rains later, the soil was assumed to be at field capacity and a second measurement was undertaken. The following readings were also event-specific during the transition from rainy into dry season. A first test of FDR sensors in sandy soils of known gravimetric water contents showed, that deviation between sensors was considerable, especially towards the upper end of the scale, so that individual calibration was necessary. For each sensor, repeated readings in the same sample were reproducible and gave coefficients of variation between two and four percent. On the other hand, even laboratory calibrations with unstructured sandy soils seemed to produce artefacts, so it was decided to install the sensors directly in the respective soil layers and use parallel auger samples as control. Correlations between gravimetric water contents (auger samples) and voltage read on FDR had to be determined for every individual FDR sensor. Data were not sufficient to develop meaningful regressions, so that only auger results were used for WC.

53 Determined for moisture correction factors of chemical soil analyses

• 96 • 3 Characterisation of soils

60 60 Gravimetric Water Contents Gravimetric Water Contents PN3 55 Cienda PN1, Feb-Apr 2006 55 Cienda PN3, Feb-Apr 2006 50 50

45 45

40 40

35 35

30 30

25 25 0-5 0-12 [% ] contents water Gravimetric Gravimetric water contents Gravimetricwater [% ] 20 12-32 20 5-14 14-49 32-62 49-100 15 62-100 15

10 10

Feb 13th Feb 16th Feb 21st Mar 1st Mar 29th Apr 25th Feb 13th Feb 16th Feb 21st Mar 1st Mar 29th Apr 25th Figure 40: Gravimetric water contents at Cienda Figure 41: Gravimetric water contents at Cienda PN1 profile determined by auger method PN3 profile determined by auger method

Gravimetric water contents sampled by auger showed an overall decrease of soil water contents (fig.40 and 41). As expected, superficial horizons and PN3 profile were most affected by desiccation, whereas WC in the lower horizons of PN1 remained stable. At the peak and end of the dry season (April 25th), the upper two horizons at PN1 decreased more sharply and WC of Ah dropped below those of Bw2. At PN3, the WC of Ah and AB fell below those of BT1 and 2 during the month of March. Opposed to the decreasing water contents, matrix potential of the soils rose towards the end of the dry period (fig. 42 and 43).

• 97 • 3 Characterisation of soils

Figure 42: Isobares of pF in different soil depths of Cienda PN1 from Feb – May 2006 and rainfall at LSU during the same time; minor time lag between both sites possible

Figure 43: Isobares of pF in different soil depths of Cienda PN3 from Feb – May 2006

• 98 • 3 Characterisation of soils

Tendencies are parallel in both plots, but although PN3 is fully exposed to sunlight and evaporation exceeds that at PN1, water reserves decrease earlier in the upslope position. The reason for this is better drainage of the andesitic subsoil compared to the stagnic BT at PN3. With respect to plant water supply, the upper rooting zone will be replenished from lateral flow and capillary rise, which cannot be expected to the same degree at PN1. The range of measured soil water potentials did not exceed pF -2.8 at any time, which would not constitute a limiting factor for plants yet54. During the peak of the dry season lower potentials were supposedly reached, but the range shown by tensiometers is limited to pF 2.8. Comparing both methods under field conditions, soil water potential, even though osmotic and pressure potentials were not accounted for, gives a better insight of root water stress than water contents do. Curves relating pF to water contents should have been based on water removal under controlled pressures in the laboratory. For auger measurements, sufficient points over a broad range to plot a pF-WC curve were not available. Hydraulic conductivity of the saturated soil (Ksat), which is relevant for erodibility, was roughly estimated from literature (JAHN, BLUME & ASIO 2002; SCHLICHTING ET AL. 1995). Ksat ranges between kf 4-5 (high to very high) for all horizons, which corresponds to 40-300cm/d. These estimates are based on texture and bulk density, which govern Ksat.

3.2.6 Ecological evaluation - summary In a broader context, the most relevant constraints of Southeast Asian soils have been subsumed by JAHN (1998). Low base saturation ranks before low CEC, P retention and limited rooting space. For lowland soils stagnic properties also play an important role. Generally, soils in Leyte are geologically young, so that CEC has not been found to be a limitation for the studied sites. Low base saturation is problematic on the more acidic soils, but the most common problem is low plant-available phosphorus content. All studied Leyte soils are characterised by moderate to high N supply. CEC is moderate to high and S value high to very high (except Cienda Nitisol, moderate). Available P contents are low, except in some C horizons. P retention is an important feature of Andosols, and during their development towards Cambisols, Luvi- and Acrisols, mainly sesquioxides and low activity two-lattice minerals like kaolinite are formed, which also retain P. Contents of exchangeable K are low for all soils except for PN1. As a heritage of their Andosol past, bulk density of the volcanic soils is generally low and pore volume is about 55-70%. Air capacity is low (<7%) for some clayey horizons, if estimated from texture and BD; in the field, often a sandy structure of aggregates was felt, which turned plastic under pressure. This indicates, that pore volume and infiltration are not a problem as long as the soils are not compacted. Pore volume of all horizons is very high. Available water capacity is 10-25%. All profiles have been estimated to be of very low to low erodibility, which is owed to high infiltration rates (hydraulic conductivity of the saturated soil, Ksat, estimated from texture and bulk density) and SOM contents. In the field it becomes obvious, that steep slopes without sufficient soil cover and rain events of extreme erosivity cannot be compensated for by low soil erodibility (example PN2 Cambisol). Another source of underestimation may be the fact, that the nomograph for erodibility does not consider depth of the soil profile; once the soil is saturated, water will simply overflow, which can be observed on site after few days of constant rain. In practice, all studied soils which are not nearly level are in danger of being eroded, if not covered and managed properly.

54 Permanent Wilting Point (PWP) = pF 4.2. Plants can survive at pF 3.8, but not grow (SCHLICHTING ET AL. 1995). These values were established for more drought-resistant plants (Helianthus sp. and Pinus sp.).

• 99 • 3 Characterisation of soils

As erodibility concerns only the topsoil, it does not give any information about the risk of landslides, which are caused through sliding of a water-saturated soil on a plastic boundary layer of clay (SCHACHTSCHABEL ET AL. 1992). Evaluating each soil by constraints after PAGEL ET AL. (1982) and SCHLICHTING, JAHN & ASIO (2003), plant production on PN1 Cambisol is only slightly limited by air capacity in the lower horizons and Na-saturation. PN1 is the only of all studied soils not deficient in K+ and one out of two with reserves of available P in the lower horizons. As rootability is not restricted to a depth of 100cm, these P sources should be accessible at least for trees. Due to erosion processes, PN2 Cambisol profile is more shallow, bulk density is higher than for the other profiles and air supply can be a problem for roots in lower horizons in addition to 50% rocks. P availability is low throughout the solum and so are K+ and Na+. Stagnic properties in PN3 Luvisol are caused by an abrupt change of clay contents from 49 to 67% in 49cm depth and increasing bulk density. As a consequence and due to small pore size, the lower horizons have limited air capacity. Chemical limitations are low pH and deficiencies of available P, exchangeable Ca, K and Na. Even more than the previous profile, DS is strongly acidic and aluminium toxicity may pose a problem in the future. Due to the low pH, contents of available P and all basic cations are also low. Heavy clay may cause problematically low air capacity, but in the field this did not seem to be a threat: Bulk density was not too high and this soil seemed to be an example of isovolumetric weathering (ASIO 1996). Horizons of LSU Cambisol are deficient in K and Na cations. S-value (at an average CEC) per m2 is high even when the shallow rooting space (60cm) is considered. Available P contents are very low over the profile depth, but relatively high in the Ah horizons, as in DS Nitisol and Marcos, where reforestation is well established. Characteristics of Marcos Luvisol depend very much on rootability, which may be species- specific. If roots are able to penetrate the BTSw horizon, then air supply and even available P are minor problems. In the C horizon, 334mg kg-1 available P and, taking more -1 samples nearby from the same depth, >300 to <1mgPBray kg were found. The number of roots counted on the profile wall gave the impression, that the BTSw is hardly rootable. On the other hand, PBray of the Ah was by far the highest of all volcanic soils, which leads to the conclusion that P is pumped up from the subsoil by trees. The profile shows extremely good Ca-supply and sufficient Mg (which may be displaced by the abundant Ca), so that base saturation is high, although exchangeable K and Na are deficient. Effective rooting space in Pangasugan Luvisol may be mechanically restricted by an abrupt textural change, but as bulk density remains low, is not seen as a major constraint. The sudden drop of pH from M2 to BTS1 may also reduce effective rooting space to some extent. Ca, K and Na contents are low and Ca concentrations are even less than Mg, as Ca is more easily leached under acidic conditions. Patag Cambisol shows stagnic properties such as manganese concretions and rust, which point to a changing water regime. Air capacity may be problematic, but only in the AB2 horizon. Corg and NT contents are the lowest of all profiles, but for NT still moderate. P, K and Na are low as in most profiles. Maitum Cambisol and Punta Leptosol both have structural limitations as their B and Sw horizons quickly turn from smeary-plastic (poor air supply) during rainy weather to extremely hard when dry. Moreover, rooting space of both profiles is shallow, which makes the overall nutrient supply problematic: Even though NT in the Ah horizon is the highest of all profiles, Punta soil disposes of less than half of the nitrogen compared to Cienda dystric Nitisol on a square meter basis. The dominating Ca-saturation of exchange sites may displace the other basic cations. Overall P supply is very low for both profiles due to the shallow profiles and, in the case of Punta, pH. KAISER ET AL. (in REES ET AL. 2001)

• 100 • 3 Characterisation of soils underlined the relevance of nutrient losses through dissolved organic matter (DOM) in shallow calcareous soils in temperate regions. This is also true for tropical soils and especially affects P, which is leached as DOP.

• 101 • 4 Effects of land use on soil rehabilitation – a paired plot approach

4 Effects of land use on soil rehabilitation – a paired plot approach In this section, effects of rainforestation on soils are assessed in a false time series approach on adjacent plots. Each paired plot represents a chronosequence as all sites had been under grassland, fallow or annuals before. In the 1990s one part per site was reforested, while the reference plot was kept under the previous management55. After ten years of increased biomass production and litter circulation, effects of the trees on soil were looked at to find out, whether rainforestation could contribute to site rehabilitation as claimed by the project. For the soil samples, the approach was to use t-tests at an α-level of 0.05 for each pair and parameter. Apart from soil sampling, litter production and decomposition were quantified on one paired plot under rainforestation vs. Gmelina arborea.

4.1 Land use history

Information on rainforestation and adjacent reference land uses as shown in tab.8 was obtained from semi-structured interviews with the respective land owners or tenants in 2005, through observations on site and from KOLB (2003).

Table 8: Land use history of the research sites in Leyte

Site Denomi- Land use nation Cienda Ci-1 to -9 see chapter 5 • Ci-RF Planted March – June 1996 (KOLB 2003) distance 2x2, then 2x1m, one species per line Ci-Grass • Pueraria and grasses (60cm high) between Cocos (approx. 10x10m). LSU LSU-RF • Rainforestation random-planted 1993-5 successively, after experimental plots and fallow dominated by grasses, Lantana sp. and bamboo. No external inputs. Intensive weeding during the first year. Approx. 50% mortality of fruit trees after 5 years. LSU-Ann • Annuals: Abaca, sweet potato and tubers planted 2004 after fallow and, previously, annuals and banana. LSU-Grass • Grassland: Pueraria, grasses and ferns, approx. 60cm high, Cocos approx. 10x10m. Marcos Mar-RF • Rainforestation random-planted in Feb 1995 after years of Imperata, then fruit trees and Ipil-ipil, a tree legume. Mulch (rice hulls) is applied around the trees; no synthetic inputs. Mar-Grass • Short grass (10cm) after annuals, burned Mar-Gme • Gmelina: Age approx. as adjacent rainforestation plot. Pure Gmelina, undergrowth dominated by ferns 60-80cm high. Pangasugan Pang-RF • Rainforestation: Planted Nov 96-Jan 97 as 2x2m lines in slope direction after Cocos / Imperata; high mortality. Pang-Gras • Short grass (<10cm) used as pasture Maitum Mai-RF • Planted 'early 90s'. Contains Gmelina sp., Swietenia sp., Acacia sp. and other exotics. Previously Imperata. 2x2m distance, urea application at planting. Mai-Grass • Short grass (<10cm) with rock outcrops. Degraded, under communal use.

55 This was due to property, not intentionally as an experimental lay-out.

• 102 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Site Denomi- Land use nation Patag Pat-RF • Planted 1994 as 2x2m – 2x3m lines in slope direction, each line corresponding to one species. Urea was applied at planting. Previously Cocos and pasture. Pat-Grass • Cleared 2003 for a transmission line, this plot consists of grass (10cm) and small bushes. Previously Cocos and pasture. Pat-Gme • Gmelina was planted 1993 Punta Pun-RF • Planted in the 'early 90s', maintained by LSU. Intensive weeding during the first year. Contains Gmelina sp., Swietenia sp., Acacia sp. and other exotics. The plot was ploughed and urea applied before planting. Previous land use Cocos and banana in the sampled middle slope area. Pun-Grass • Grassland after cocoa and annuals, all between Cocos (approx. 10x10m).

During the first years after planting of the rainforestation plots, dead trees were replaced by the project. Regular weeding was carried out by LSU personnel for some plots during the first year after planting. Apart from sporadic coconut harvest, only one site, Marcos, yielded noteworthy amount of fruits in 2005.

4.2 Soil samples

All soil analyses presented in section 4 refer to topsoil from 0-20cm depth. This layer was expected to most clearly reflect biological and litter-related processes influenced by the recent land use changes. Results obtained at Cienda in 2004 (see chapter 5) had provided an orientation on the spatial variability of parameters. In consequence, composite samples were collected to increase representativeness of the plots for the experiments in 2005. From all but two plots, eight composite samples consisting of three individual samples each were taken. Exceptions were Cienda (eight individual samples) and Punta (five composites containing 20 individual samples each). All samples were analysed for pH, C, N, BR, SIR and phosphatase. For available cations and PI less samples were analysed due to limited resources.

4.2.1 Soil carbon, nitrogen and pH Soil organic carbon and pH were analysed for a project status report by Asio and co- workers (CENIZA ET AL. 2004). Increases over time were found for both on the rainforestation plot at Cienda, but with considerable fluctuations of pH. At first sight, obvious differences in organic matter were observed between sites depending on rock, drainage and erosion as long-term factors. For the different land uses at each site, soil organic carbon, total nitrogen and C:N-ratio at the respective pH are shown in fig.44. pH was a relatively sensitive parameter to indicate differences between land uses. Values at rainforestation plots were lower than those of the paired land uses in 7 of 10 cases. This goes conform with observations of KELLMAN (1970) along a successional chronosequence in Mindanao, finding that pH decreased during (secondary) succession, especially after initial burning. Elevated pH could be observed at the recently burned Marcos grassland plot. The same tendency was found for C and N contents.

• 103 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Figure 44: Organic C, total N, C:N-ratio and pH for rainforestation and reference land use plots in Leyte

With respect to Corg and NT, tendencies between land uses were not obvious across sites. On the calcareous sites, both C and N were higher under rainforestation then under grassland. Where Gmelina was involved, C and N were elevated. Plots with the highest C and N contents were also the ones with the widest C:N-ratios. These are located at Cienda and Punta sites, where 'extreme' pH values prevail, and at the Gmelina plot at Marcos. Litter and humus accumulation were clearly visible at Punta, most probably due to the excessive drainage and high pH of the terrain. C:N-ratio turned out to be one of two most sensitive parameters for detecting significant (α = 0.05) differences between land uses (7 of 13 cases). In 3 of 4 significant cases between rainforestation and grassland, C:N-ratios were tighter under the first. At the two sites with adjacent rainforestation and Gmelina, one difference was significant, also with rainforestation showing the tighter C:N. For C (LANGI in GEROLD ET AL. 2004) and N (KELLMAN 1970), a gradual increase has been reported following an initial drop after disturbance. V.NOORDWIJK ET AL. (1997) state, that belowground carbon stocks of well-managed pasture do not necessarily fall below those of tree-based systems. LEITE ET AL. (2004) simulated and measured decrease and recovery of Corg fractions after clear-cutting of Amazonian rainforest and organic inputs, while mineral fertiliser at least maintained the low levels.

4.2.2 Available Ca2+, Mg2+, K+ and Na+ Loss of basic cations after logging of forests in Sabah, Borneo, has been described by NYKVIST ET AL. (1994), who found particularly high leaching of potassium (>100kg/ha). A potential increase of basic cations in the topsoil as a consequence of tree-growing could be ascribed to transfer from the subsoil via tree roots and leaf litter or through leaching from leaves. Fig.45 shows concentrations of plant-available calcium, magnesium, potassium and sodium in the topsoils of the paired plots. Calcium concentrations were dominated by parent material and highest at the calcareous Punta and Maitum sites. Lowest Ca2+ values were found at the acidic Cienda plots. Contents were rather site- than plot-specific. Significantly lower values (t-test, α = 0.05) attributable to land use were found for LSU-grassland compared to rainforestation and Marcos rainforestation compared to grassland. Considering also non-significant

• 104 • 4 Effects of land use on soil rehabilitation – a paired plot approach differences, Ca was lower in rainforestation soils for 7 of 10 pairs.

Figure 45: Available basic cations at rainforestation and reference plots in Leyte. Note separate scale for Calcium

Mg2+, apart from C:N, proved to be the most sensitive of all parameters. Seven of ten paired land uses differed significantly with respect to available Mg. However, in 5 of 6 significant cases involving rainforestation, this system showed the lower Mg2+ values. The exceptionally low Mg2+ values at Punta site are supposedly due to replacement by Ca2+. 2+ NYKVIST (1997) calculated for forested Acrisols in Sabah, that up to 50% of ecosystem Ca can be bound in the vegetation. Although this ratio was only 3% for Mg2+, amounts in tree- based systems (e.g. 0.6kg Mg2+ ton-1 tree biomass, same author) are considerably higher than for grassland, which may explain the lower contents in soil under rainforestation. Levels of potassium and sodium were statistically not distinguishable between plots and even non-significant differences showed no trend in favour of any particular land use.

4.2.3 Basal Respiration

Basal respiration describes the evolution of CO2 during equilibrium microbial metabolism. Basal respiration (BR) as well as microbial biomass is strongly influenced by microclimate, with soil moisture ranging before temperature (WARDLE & PARKINSON 1990). Microclimate in turn is affected by canopy cover (through shading and litterfall), so that differences between open areas and closed-canopy systems could be expected. On a smaller scale, substrate quantity (SOM) and decomposability, reflected by C:N-ratios, are relevant for microbial parameters. Superior litter quality and decomposition of Gmelina leaves compared to those of indigenous trees (s.4.7.2) were likely to affect microbial activity reflected by BR. Basal respiration rates are shown in fig. 46 for the incubation period of 24 to 96h after conditioning as average per hour. Results display the same tendencies as during the first 24h (not shown), to the exception of Punta site, where BR was extraordinarily high at the beginning56. A long-term experiment (5.3.1.2) confirmed that BR rates usually take some time, even after conditioning to the lab environment, to reach an equilibrium.

56 Probably due to carbonate set free from the calcareous soil material (SCHINNER ET AL. 1993)

• 105 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Figure 46: Basal respiration and qCO2 after 4 days incubation for paired plots under different land uses A clear site effect was the elevated BR at Punta, where rates surpassed those of all other sites by far. This may be attributed to the combination of high contents in organic matter (4.2) and of high biological activity (DAUB 2002), but also to stress caused by high pH and 57 excessive drainage . An influence of stress is also suggested by elevated qCO2, or basal respiration per unit microbial biomass, which is often interpreted as opposed to metabolic efficiency. Elevated BR rates were also found under easily decomposable Gmelina litter. BR under Gmelina exceeded that of adjacent land uses in all 4 cases, for 3 of them significantly at α = 0.05. For the Cienda subplots 1-9, the eroded subplot 8 under banana could be distinguished at α = 0.05 from Ci-1, 4 and 9. For Cienda-RF and -grassland, high BR was associated to high microbial biomass, resulting in low qCO2. Basically, qCO2 followed BR to the exception of Cienda subplot 9, Ci-RF and Ci-grass; some peculiarities will be discussed under 4.2.4 in context with the microbial biomass Cmic. Overall, most plots exceeded levels measured by MAO ET AL. (1992) under reforestation in -1 -1 tropical China (0.11-0.94µg CO2 h g ), but remained below data reported by MENYAILO ET -1 -1 -1 -1 AL. (2003) around 0.1 – 0.15gC kg d (equivalent to 15-25µgCO2 g h ) for Amazonian soil samples incubated at 28°C and 60% WHC.

57 Samples were collected during dry season. This can be influential in spite of adjusted water contents of the samples.

• 106 • 4 Effects of land use on soil rehabilitation – a paired plot approach

4.2.4 Microbial carbon, Q10 and qCO2

In contrast to other methods for determining microbial carbon (Cmic) in soils, substrate- induced respiration (SIR) is based on a flush of microbial respiration after addition of glucose and thus determines only Cmic of the actively metabolising biomass. This is of relevance, since most bacteria are in a dormant state at any given time, not participating in decomposition processes. While SIR focuses on young bacterial biomass, the standard fumigation- extraction method emphasises the fungal biomass (BRAKE ET AL. 1999). Metabolic quotient or specific respiration qCO2 is the proportion of CO2-C evolved from basal respiration per unit microbial biomass C. The quotient characterises the energy necessary to maintain a certain population biomass (ANDERSON & DOMSCH 1986). High qCO2 values are interpreted as low metabolic efficiency. This may be due to stress, caused e.g. by drought or substrate quality and land use (WARDLE ET AL. 1999) or soil depth (MAO ET AL. 1992). qCO2 has been interpreted as an indicator of the successional maturity of a microbial community (INSAM & HASELWANDTER 1989; FLIESSBACH ET AL. in REES ET AL. 2001), following the theory, that community respiration decreases relatively to community growth from pioneer to maturity stage of an ecosystem (ODUM 1969). Another potential cause for comparatively higher qCO2 mentioned by DILLY & MUNCH (1998) is a dominance of organisms specialised in easily decomposable substrates, e.g. of r-strategists over K- strategists or of bacteria over fungi. This argument seems to lead into the same direction as Odum's theory, if r-strategists are understood as pioneer species and K- or autochthonous species as typical for advanced stages of succession.

Theoretical digression: Incubation temperature and Q10 : Incubation temperature is critical for microbial metabolism in general and for SIR in particular. During the first year, an air- conditioned room was not available and experiments had to be carried out at 32°C. In 2005, the set-up was 22°C, in line with standard procedures. In order to assess the influence of temperature on SIR, a simultaneous SIR experiment including both temperatures was undertaken as a pre-test in 2005 (fig.47) with samples across a broad range of soils and land uses. Following Arrhenius dynamics, an increase of factor two per 10K was expected for elevated temperatures. This Q10 factor has been assessed for qCO2 of two European soils (Luvisol and Phaeozem), found to be 1.4 to 2.0 per 10K between 0°C and 45°C (ANDERSON & DOMSCH 1986). FANG ET AL. (2005) Figure 47: Assessment of Q10 for SIR experiments report values around

• 107 • 4 Effects of land use on soil rehabilitation – a paired plot approach

2.1. In this study, not basal, but glucose-induced respiration as used for SIR was measured, and Q10 values between 22 and 32°C were found to be 1.2 to 1.9, depending on soil.

As stated, experiments in 2005 were carried out at 22°C. This is a prerequisite for the application of the conversion factor suggested by ANDERSON & DOMSCH (1978) to derive microbial biomass Cmic from respiration rates. Assuming a respiration coefficient of 1.0, a -1 -1 rate of 1mgCO2 100g soil h corresponds to 20.6mgCmic per 100g soil for temperate regions. Since a more specific conversion for tropical regions could not be found in literature, the original factor was used to express Cmic in fig.48 and later for Cmic/Corg ratios. Alternatively, qCO2 was expressed as crude respiration rates, calculated as gCO2-C evolved per kg soil and day. This is sufficient as long as samples are compared under equal conditions. However, rates were exactly in the range given by MENYAILO ET AL. (2003) 58 for humid-tropical soils in Brazil . Cmic-SIR quantified by MOREIRA (2004) in the Brazilian Amazon was at the lower end of the range determined in this study (150-200µg g-1 soil) with no significant differences between different land uses. In contrast to BR measurements, Cmic values at Punta site were in the normal range and even slightly below. For this reason an overestimation of Cmic-SIR in the calcareous Punta soil due to changed O2 partial pressure (BECK ET AL. 1997) is not likely. Concerning land uses, Cmic-SIR did not prove an appropriate distinguishing parameter at α = 0.05 level, however, as a tendency, rainforestation plots were relatively lower in Cmic for 7 of 10 pairs. The largest microbial population size was found at Cienda-RF and -grassland. Cienda -9 and LSU-Annuals had been disturbed in 2003 and 2004, so that the soil was not in equilibrium, and Cmic can be expected to fall back to normal levels during coming years. The peak is due to an exponential growth phase of microbial biomass after addition of substrate (here: mulch or crop residues) leading to higher production of CO2 per biomass (ANDERSON & JOERGENSEN 1997). The same phenomenon was observed for the two Gmelina plots with easily decomposable litter: High qCO2 was associated with intensive turnover of substrate.

58 SIR carried out at 28°C

• 108 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Figure 48: Substrate-induced respiration and qCO2 under rainforestation and reference land uses in Leyte

Apart from intensive turnover, high qCO2 can be the result of stress as in the case of the high pH value at Punta or the burned Marcos grassland. For the annuals plot at LSU as for the disturbed subplot Ci-9, the surprisingly low qCO2 may be due to a build-up of microbial population about one year after disturbance through planting. qCO2-rates were very similar to those presented by ANDERSON & DOMSCH (1986) in a laboratory experiment assessing effects of different temperatures on qCO2. As assumed before, BR in the temporarily dry and alkaline Punta soil was an expression of stress, which is corroborated by the relatively low microbial biomass and consequently high qCO2. In contrast, low BR presented for Ci-8 and Ci-RF was associated with high microbial biomass and thus small qCO2, pointing to unfavourable but stable conditions for microorganisms (erosion and low OM for Ci-8 and acidity for Ci-RF). Thus, for Cienda RF, the large glucose-responsive microbial population is well adapted and efficient, meaning that maintenance requirements are low. Following this logic of short-term adaptation, temporal drought would have been the main stressor at Punta rather than pH. Considering the more temperate climate (21°N, 23°C, 1600mm annual rainfall), results obtained by MAO ET AL. (1992) in China were in comparable magnitudes to those from -1 -1 -1 Leyte: Cmic-SIR between 20 and 350µg Cmic g and qCO2 0.8 to 1.8µgCO2-C g Cmic h . The microbial portion of organic carbon, expressed as quotient Cmic/Corg, describes relative availability of substrate for soil microorganisms (ANDERSON & DOMSCH 1986) and can give an impression of amounts of organic inputs into a system and thus of management (ANDERSON & JOERGENSEN 1997). In this study, Cmic/Corg turned out to be rather site- than land use-specific with exception to Maitum and LSU grassland as well as Patag rainforestation, which were higher than their alternative land uses (in Maitum significantly at α = 0.05). For acidic Amazonian soils, MOREIRA & MALAVOLTA (2004) found, that Cmic/Corg did not even differ substantially between primary forest and fruit tree plantations (around 2%). In acidic Amazonian soils analysed by FEIGL ET AL. (1995), microbial carbon amounted to 3-4% of Corg and in humid subtropical China ratios from 0.48 to 2.31% were found in different (agro-) forestry systems (YAN ET AL. 2003). In this study, Cmic ranged over a wider scale, from 0.67 to 4.1% of Corg.

• 109 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Figure 49: Cmic , Corg and the quotient Cmic / Corg for the paired plots in Leyte

4.2.5 Available PI and phosphatase activity Enzyme activities in soils have been interpreted as indicators for land-use, organic and microbial carbon and even 'sustainability' of a given land-use system. Enzymes react even more quickly to management changes than microbial biomass (DODOR & TABATABAI 2003). Phosphatases are secreted by plant roots and soil microorganisms. They transform P bound to organic macromolecules into plant-available inorganic forms. Acid and alkaline phosphatases are often analysed separately because they may originate from different sources (SCHINNER ET AL. 1993) and depend on soil conditions (GEORGE ET AL. 2002). For this study, phosphatases were determined at the respective given soil pH to account for different site conditions. Phosphatase activity for the different plots in Leyte is shown in fig.50 in context with pH and PBray II as influential parameters. Soil moisture and temperature were controlled and calibration curves were very similar for all batches, thus photometer readings of different dates could be compared. In addition, composite samples were used in a separate experiment to confirm relative magnitudes of the different batches.

• 110 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Figure 50: Phosphatase and PBray at paired plots

Results show especially for the Cienda subplots that phosphatase activity was inversely related to pH. This seemed to be the most influential factor as found by correlation analysis. Increasing phosphatase activity with duration of the fallow period – as found by DENICH & KANASHIRO (1998) for acid phosphatase – can be observed for the Cienda subplots 1 - 9, if decreasing relative PAR is associated with fallowing (closing canopy). Vegetation as a source of phosphatases may have been of minor importance since roots had been sorted out before analyses. However, vegetation as a source of litter and thus Porg, the substrate for phosphatases, can be related to Cienda subplots 1-9, even if P-recycling through leaf litter may be limited to some extent by retranslocation before leaf fall (KHANNA 1998). Phosphatase production is usually negatively correlated to PI contents (YADAV & TARAFDAR 2001, DENICH & KANASHIRO 1998), which were expected to limit plant growth on all volcanic sites in Leyte (s. chapter 3; ASIO, personal communication; ZIKELI 1998). For Marcos rainforestation and Gmelina, however, levels of phosphatase exceeded those under grassland. This could indicate, that trees were not able to significantly increase PI levels through nutrient pumping from the subsoil. If root phosphatases would have been involved, these would probably have had a higher share in the topsoil under grassland. This may have been the case for Cienda and LSU, where fallows were of relatively high biomass. Decreasing contents of PI with progress of succession have been reported by LEHMANN ET AL. (2001), but for the sites in Leyte, Bray II - PI contents were too low in most cases to give distinguishable tendencies. Even values at Maitum would still classify as low according to the evaluation scheme by PAGEL ET AL. (1981). The grassland plot at Marcos was the only place where signs of burning during recent years could be observed in the field. Insofar increased pH and slightly higher PBray contents are typical (KELLMAN 1970).

• 111 • 4 Effects of land use on soil rehabilitation – a paired plot approach

4.3 Leaf litter production and decomposition under different tree systems

A paired plot of rainforestation on one side and Gmelina on the other was selected to assess litter cycling. To account for seasonal effects, litter traps and minicontainers were set up during the transition period from rainy into dry and again into rainy season 2006 (March 1st – June 15th).

4.3.1 Leaf litter production With respect to litterfall, orders of magnitude on a dry matter basis were similar for both systems. Clear seasonal leaf-shedding was also observed in both. On the rainforestation plot, the peak of leaf shedding preceded that of Gmelina. Simultaneously to leaf fall, a period of flowering (and then abscission of flowers) could be observed in Gmelina (fig. 51). Due to the higher species diversity, fruits and flowers (F&F fraction) fell more uniformly on the rainforestation plot. Bark and branches (B&B fraction) were similar for both land uses with an increase towards the dry season. Figure 51: Time series of different litter fractions at Marcos Generally, standard rainforestation and Gmelina sp. plots. Daily amounts of rainfall deviations of F&F as well as of B&B were high. At ten replicates per plot, coefficients of variation were between 150 and >200% for both fractions and systems. For leaves, CV were 40% for Gmelina and 51% for rainforestation, thus temporal patterns can be described best only on a leaf basis. An influence of the dry season can be deduced for Gmelina, but not as obviously for the rainforestation plot with its more diverse species composition.

4.3.2 Litter decomposition Rates of litter decomposition influence nutrient cycling in tree-based systems, with rapid turnover providing limiting minerals like P to plants but also increasing the risk of leaching, especially for K and N (ISAAC & NAIR 2005). On the other hand, slow decomposition maintains physical characteristics of organic matter like pore volume and structure, which improve the soil water balance. Mini-container experiments for an assessment of litter decomposition were installed on both plots during three different short periods with decreasing amounts of rainfall (fig.52). For the first set-up (Feb 18th – Mar 8th), a leaf mixture based on Ficus spp. from Cienda was used as a neutral reference. The second and third experiment were carried out with leaf litter from the respective plots. Minicontainer capsules covered with 4mm mesh tissue allowed access for the mesofauna, while 0.1mm did not, so that both decomposer groups could be differentiated.

• 112 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Dispersion and deviations were clearly higher for 4mm than for 0.1mm mesh width. A t-test between 0.1 and 4mm meshs showed, that for the period from Mar 28th to April 11th, decomposition in Gmelina and Rainforestation did not differ significantly (α=0.05), but from April 6th to 20th, Gmelina was decomposed to a significantly higher degree than rainforestation litter (α=0.01). Decomposition for rainforestation was clearly lower during the latter period Figure 52: Decomposition of plot-specific leaf litter under (for both mesh widths, t-test P different land uses in Marcos, 2006. R = rainforestation, G = = 0.0000), while Gmelina Gmelina; 0.1 and 4mm are mesh widths of the minicontainer capsules maintained the same level or even increased slightly (P = 0.031). This coincided with scarce rainfall during both periods (60 and 58mm) and the observation that soil moisture decreased more in the rainforestation plot, which suggests inhibited decomposition caused by drought. ISAAC & NAIR (2005) explained increased leaf decomposition during the rainy season in Kerala, India, with higher microbial activity and substrate quality. BECK & GASPAROTTO (2000) underlined the importance of microclimate on decomposition in context with different land uses in Amazonia. Leaf analyses of trees for WaNuLCAS tree parametrisation (chapter 6) show, that litter quality probably played an important role for the faster decomposition of Gmelina (table 9). Tighter C:N and C:P ratios and lower percentages of polyphenolics can also be observed for the pioneer S. palosapis relative to the other indigenous trees. Polyphenolics such as tannins can hamper and have stronger effects on decomposition than C:N-ratios (ISAAC & NAIR 2005). Table 9: C:N- and C:P-ratios and total extractable polyphenolics for leaves of Gmelina arborea and native trees

Species Leaf C:N - Leaf C:P - Polyphenolics ratio ratio [% TEP59] Gmelina arborea 12.3 188.8 1.56 Dipterocarpus validus 20.5 330.4 6.65 Shorea contorta 24.5 508.0 9.75 Shorea palosapis 20.0 266.0 6.33 Toona calantas 25.9 377.4 9.24

Faster decomposition of Gmelina arborea relative to Dipterocarpus validus was also found by ARAGON (2004) and BATISTEL (2004), who compared leaf decomposition of Swietenia sp., Tectona sp. and Gmelina arborea to that of several indigenous trees. 59 Total Extractable Polyphenolics

• 113 • 4 Effects of land use on soil rehabilitation – a paired plot approach

4.4 Synopsis

Rainforestation has been massively promoted in Leyte during the last decade. Discussions have often intended to prove the superiority of the rainforestation system, especially over the DENR approach that includes pure stands of exotic trees. Apart from general ecological questions60, an important issue has often been overlooked or ignored: Rainforestation is a flexible concept that requires adaptation to each farmer's reality. Consequently there are as many systems as rainforestation farms. In practice, indigenous and exotic plants were often freely mixed, species diversity was not always high, understorey fruit trees often died and were not always indigenous species61 and there was no uniform planting scheme or distance at all. This made it difficult to compare 'rainforestation' to other land uses, which were not always as pure as the Gmelina stand in Marcos. Especially the available grassland areas were not uniform, including Imperata- and Pueraria-dominated areas as well as pasture. Given these restrictions, some interesting tendencies appeared across most sites with regards to differences between land uses and correlations of parameters: Summarising the t-tests, pH under rainforestation was mostly below that of any reference use (7 of 10 cases, four of them significant). The same was true for contents of Mg (six of ten cases, 5 of them significant). Ca was lower in rainforestation in 7 of 10 cases, but only once significantly. For P, contents were determined from one composite sample, so that no t- test was performed; contents under rainforestation were lower in 7 of 10 comparisons. C, N, BR and Cmic were also each below the reference values in 7 of 10 cases each, but with only one significant case for Cmic, 2 for BR and NT and 3 for Corg. This is opposed to the decline in total microbial biomass (fumigation-extraction method) after conversion of forest into agricultural areas observed by WALDROP ET AL. (2000) in Tahiti. C and N were lower than the reference plots in all rainforestation plots of low pH (volcanic rock) and higher at Punta and Maitum. The quotient Cmic / Corg was significantly different in only one case (Maitum), where rainforestation was below the reference land use. With respect to differences, cases of the remaining pairs were balanced. Two sites showed clear overall trends across parameters: In Marcos, growth conditions under rainforestation were less favourable for all measured parameters except qCO2 (which was higher under Gmelina) when compared to Gmelina, and except Cmic and qCO2 when compared to grassland. The same was true for a comparison of LSU rainforestation with the LSU annuals field with exception to C:N and phosphatase. This shows that recently disturbed plots can contain higher levels of available cations and P as well as higher pH, especially after burning (Marcos). Elevated qCO2 would then point to a young and growing microbial population, that has not yet adapted to the new conditions and is still to become more efficient. The same applies to Gmelina, where litter quality and / or quantity surpassed that of the reference land uses: Turnover and thus qCO2 were elevated. MAO ET AL. (1992) ascribe elevated qCO2 under Eucalyptus sp. compared to mixed monsoon forest to lower microbial diversity and thus reduced efficiency under pure stands. Comparing Gmelina to rainforestation at Patag and Marcos, Gmelina was more favourable in both cases with respect to C, N, phosphatase (lower), K, BR and Cmic. Grassland showed higher BR and Cmic, P and Ca contents than rainforestation in 6 of 8 pairs. On the other hand, qCO2 was also higher in these cases. Looking at the potential of parameters to differentiate land uses, these belong to three

60 e.g. how far can indigenous trees be replaced by exotic species, and to which extent can structural or functional diversity substitute species diversity (RUSSELL 2002) 61 Most are cultivars from Mindanao and at least Durian is exotic (LANGENBERGER pers. comm.)

• 114 • 4 Effects of land use on soil rehabilitation – a paired plot approach categories of different time-scale. Firstly, enzymatic and microbial parameters indicating short-term effects of land use changes (DODOR & TABATABAI 2003; ANDERSON & DOMSCH 1989) as shown on the Cienda subplots for phosphatase or qCO2 for the recently disturbed LSU- annuals plot. Secondly, pH as a medium term-indicator, and thirdly, available nutrients on a longer scale. Under this aspect it was surprising, that clear differences between land uses could already be observed with respect to available cations, especially Mg, within 10 years after reforestation. With regard to relevance of mechanisms for soil development, it was of interest to look at relationships between the evaluated parameters across all sites and plots. Correlations between the most relevant parameters are shown in tab.10, values for BR and qCO2 - 24h being omitted.

• 115 Table 10: Correlations between analysed soil parameters across different sites and land uses in Leyte62. Significance at α = 0.05 is indicated by *, at α = 0.01 by **.

2+ 2+ + + pH C N C : N P-ase PBray Ca av Mg av K av Na av BR 72h Cmic QCO2 72 pH 1 0,527** 0,395** 0,408** -0,449** -0.377 0,808** 0,609** 0.145 -0.014 0,597** 0.061 0,434** C 0,527** 1 0,909** 0,476** 0.062 0,722* -0,388* -0,325* -0.107 0,487** 0,796** 0,449** 0,348** N 0,395** 0,909** 1 0.077 0.081 0,758* -0,341* -0.208 0.125 0,361* 0,699** 0,309** 0,367** C : N 0,408** 0,476** 0.077 1 0.001 0.255 -0.140 -0.289 -0,416** 0.288 0,390** 0,404** 0.043 P-ase -0,449** 0.062 0.081 0.001 1 -0.244 -0,520** -0.220 -0.168 0.325 -0.067 0,410** -0,308**

PBray -0.377 0,722* 0,758* 0.255 -0.244 1 0,951** -0.445 -0.010 -0.118 0.377 0.378 -0.039 Ca2+ av 0,808** -0,388* -0,341* -0.140 -0,520** 0,951** 1 0.247 0.144 -0.069 -0.188 -0,394* 0.079 Mg2+ av 0,609** -0,325* -0.208 -0.289 -0.220 -0.445 0.247 1 0.224 0.134 0.101 -0,414* 0.337 K+ av 0.145 -0.107 0.125 -0,416** -0.168 -0.010 0.144 0.224 1 0.109 -0.037 -0.271 0.265 Na+ av -0.014 0,487** 0,361* 0.288 0.325 -0.118 -0.069 0.134 0.109 1 0,467** 0.332 -0.144 BR 72h 0,597** 0,796** 0,699** 0,390** -0.067 0.377 -0.188 0.101 -0.037 0,467** 1 0,276** 0,637**

Cmic 0.061 0,449** 0,309** 0,404** 0,410** 0.378 -0,394* -0,414* -0.271 0.332 0,276** 1 -0,444**

QCO2 72 0,434** 0,348** 0,367** 0.043 -0,308** -0.039 0.079 0.337 0.265 -0.144 0,637** -0,444** 1

Cmic/Corg -0.090 -0.104 -0.210* 0.183 0.325** -0.298 -0.104 -0.197 -0.287 0.039 -0.201* -0.789** -0.739**

62 BR and qCO2 after 24h incubation are not shown, because they tended to be less balanced than those after 72 hours. In the case of qCO2 - 24h, correlations with pH (0.657**), C (0.591**), N (0.473**) and C:N (0.391**) were more pronounced than those for qCO2 72h. • 4 Effects of land use on soil rehabilitation – a paired plot approach

Strong correlations can point to dependence of one parameter on a limiting (for positive correlations) or restricting (for negative correlations) factor. pH and Corg were the most influential among all measured parameters. However, both were not too strongly interconnected across all plots as humus accumulation was typical for both ends of the pH scale. Under strongly acidic conditions free Al3+ can attenuate OM decomposition. At the calcareous sites, humus accumulation can be attributed to metabolic optima of most soil bacteria at slightly acidic conditions. For both extremes, characteristic fixation of P to minerals can be another reason for reduced microbial activity leading to humus accumulation. Overall, pH and PBray were inversely correlated. The strong positive correlation between Ca and P may point to apatite as a source of PBray dissolved by the HCl component of the extractant (PAGEL ET AL. 1982). This was supported by the fact, that 63 for POlsen only traces were found in Punta soils . Comparing correlations to those of KAISER ET AL. (1992), Cmic was significantly positively correlated to Corg and BR and so was C:N, for which KAISER ET AL. reported a negative correlation. For Cmic and N, correlations were positive in both studies, but weaker here, probably because N was not a limiting factor. A correlation of Cmic to pH as reported by ANDERSON & JOERGENSEN (1997) for temperate soils and in this study for BR to pH was not found; this is in line with studies by KAISER ET AL. (1992). In addition, the correlations were not very strong, probably due to the different soil types, including underlying factors of influence. Correlation between BR and pH was stronger, as shown before by JOERGENSEN & SCHEU (1999) and V.NOORDWIJK ET AL. (1997) and so were correlations between BR and C, N 64 and C:N. Microbial activity (BR) and qCO2 were both positively correlated to pH and consequently high at Punta and low at the acidic Cienda Nitisol. In this sense, positive correlations between C and N on one hand and qCO2 are not interpreted as a relationship of stress, but as due to substrate supply. Under unfavourable conditions, metabolic efficiency and microbial population size would decrease – a highly significant and strong negative correlation qCO2 – Cmic / Corg was found. This was also observed by JOERGENSEN & CASTILLO (2001) for young volcanic soils in Nicaragua. Microbial activity seemed to be more dependent on substrate quality than on P and cations. Both Cmic and BR were positively correlated to P, but not significantly. Cmic was negatively related to available Ca and Mg and positively correlated to phosphatase suggesting a common correlation of these parameters to an underlying effect. The strong statistical weight of the Punta and Maitum sites can be noted from the clear correlation PI – Caav. Sensitivity of the correlations to these two sites was strong for pH, C, phosphatase, Ca, Mg and all BR-related parameters including qCO2. Correlations of N, K, Na and Cmic, however, did not change much, when the two sites were excluded from the calculation of the correlation table65. An important objective of the rainforestation project was rehabilitation of degraded lands. This would combine improvement of soil fertility, i.e. yield-related parameters, and soil quality as related to ecological functions of soils (FILIP 2002). ASIO (1996) identified soil colour, reduced OM contents and soil respiration, increased erodibility and available basic cations as indicators of degradation and tested the hypothesis, that land use change would not necessarily lead to soil degradation. A principal conclusion was, that degradation was owed to the effect of clear-cutting, but

63 values obtained by Bray II roughly double those by Olsen and quadruple such by Truog extraction – s. PAGEL ET AL. 1982 64 Contrary to a negative correlation in soils of similar pH range found by YAN ET AL. 2003 65 Both tables were compared counting cases per parameter, where a correlation coefficient differed >0.4 between the full matrix and the one that excluded Maitum and Punta data. For the pH-sensitive group of parameters r changed in ≥5 of 16 cases, for the less sensitive group in ≤3 cases.

• 117 • 4 Effects of land use on soil rehabilitation – a paired plot approach could not be attributed to any particular form of 'secondary' land use (here: Cocos + Pueraria, Cocos + bush fallow, pasture, slash & burn). Concerning soil functions for sloping lands in Leyte, soil physical parameters counteracting erodibility appear most relevant to reducing erosion and the risk of landslides. These include organic matter, pore volume and bulk density as determinants of water holding capacity. Any build-up of biomass, preferably a closed layer producing large quantities of mulch, would improve these soil properties. Under the soil water and structure aspect, a thick mat of creepers such as Pueraria (s. OM contents in ASIO 1996 or at LSU, Cienda, section 4.2) can be of similar efficacy as a multi-storey system, but interception and transpiration would be lower and deep-rooting less intensive. Thickness of litter layers under rainforestation has been described in chapter 3 for each plot. For grassland at Marcos, Pangasugan, Patag, and Maitum, litter was virtually absent. Under this aspect rainforestation would be preferable to grassland because of higher litter inputs, and to Gmelina due to slower litter decomposition. The role of plant species diversity on microbial efficiency has been discussed with respect to qCO2 and the influence of different species on soil was illustrated by canopy and leaf litter production. An open question regarding ecological functions is, to what extent small-scale variability of soils influences species diversity, e.g. during germination. Effects of slope position on the composition of plant communities have been described by LANGENBERGER (2003) for Mt. Pangasugan in the hinterlands of LSU, and for Cienda, but at a smaller scale, studies have not been conducted. It has been stated, that rehabilitation of soils through natural succession is rather independent of soil nutrients, even though with some confinement with respect to Mg (HARCOMBE 1980). FILIP (2002) used BR and SIR measurements, among others, to draw conclusions on soil functions at different levels of contamination. However, as soon as yields and rentability are concerned, nutrients would be a limiting factor. With respect to soil fertility, chemical parameters are of great importance and this was a focus of the present study since fertility is crucial for any cash crop growing under trees. Soil quality and fertility can be opposed under certain perspectives. While for the ecological functions, a slow decomposition of organic matter increases structural stability, growing organisms may benefit from rapid turnover of OM and cycling of nutrients. Having started on grassland areas, the rainforestation plots represent an advanced successional stage on the way to a potential forest vegetation, in contrast to the arrested development on grassland plots. It is well-known that nutrient cycles in tropical forests are tight and a substantial share of ecosystem nutrients is tied up in vegetation as succession progresses 66 towards forest (NYKVIST 1997), leading to depletion of soils (HARCOMBE 1980) . In addition, KELLMAN (1970) reports a slight decrease of pH and PI under advancing secondary succession. Mg2+ is subject to leaching and uptake into plants is depressed by other 2+ + cations such as Ca , but also H (MARSCHNER 1993). As a consequence, soil acidification accompanying succession would reduce nutrient availability (see the strong positive correlations of Ca2+ and Mg2+ with pH in tab. 10). Interveinal chlorosis denoted on L. domesticum under canopy at Cienda may be a prove for this. The results presented in this chapter confirm the relevance of these mechanisms for Leyte. ASIO, JAHN & STAHR (1999) found higher concentrations of basic cations on plots after a conversion of forest soils in

-1 66 On a basis of roughly 300Mg biomass ha and 0.3% Mg in average plant tissues (after MARSCHNER 1993) -1 and about 5cmolcMg kg soil at PN1 this would lead to a ratio of 12 in soil/plant within the rootable depth. Availability and leaching of Mg are not considered in this estimate, nor is leaching from leaves into the soil. -1 For P, 5kg PBray ha in the soil would support roughly 1000kg P in plants. Plant available P, however, makes up for only a small part contained in the soil resource (total P in soils is 0.01-3.4%; PAGEL ET AL. 1982), while magnitudes for Mg2+ are similar in plant and soil.

• 118 • 4 Effects of land use on soil rehabilitation – a paired plot approach

Leyte, but ascribed this to enhanced weathering, presence of ash and a rejuvenation effect of soil erosion. While nutrients are accumulated in the system as a whole, supply in the soil may decrease. While shifting cultivation makes destructive use of the minerals stored in the fallow biomass, simultaneous agroforestry systems must necessarily aim at increasing soil fertility of the developing system, including tree and crop components. For any shade- tolerant plant grown under canopy it is of vital importance that additional minerals provided by the nutrient pump compensate for the competitive effect of the tree component. The bias between conservation and economic feasibility may be one reason for the lacking adoption of the system among target groups67. Rainforestation has not yet transgressed from a conservation concept to an economically viable alternative for farmers. Similar observations have been made by CRASWELL ET AL. (1997), who state, that labour is a major constraint to the adoption of complex agroforestry systems in Asia, as long as external incentives are not provided. The important role of early-yielding cash crops for the profitability of agroforestry systems has been discussed earlier. From this perspective, growth of the early-yielding species Musa textilis under varying vegetation and environmental conditions will be evaluated in chapter 5.

67 More than ten years after starting the project, 28 plots exist, integrating an area of approximately 10ha. The free distribution of seedlings and assumption of weeding through the project have not led to broader acceptance.

• 119 • 5 Plant growth in an agroforestry system under different small-scale environments

5 Plant growth in an agroforestry system under different small- scale environments The experiments presented here were carried out at Cienda site in 2004, the first year of the study. Objectives in section 5.1 were to assess small-scale variability of parameters and obtain a detailed characterisation of the situation 'before planting'. Section 5.2 examines plant performance on the same plot in response to spatial heterogeneity.

5.1 Site parameters

Data refer to the ten Cienda subplots described in section 2.5.1 with some reference measurements on the Cienda paired plots. Vegetation and land uses on the respective subplots are summarised in table 11.

Table 11: Vegetation and land uses for Cienda subplots

Subplot Slope Present vegetation and land use before 2004 position 1 and 2 Lower Young secondary forest, mainly Ficus spp.; coconut 3 Middle 4 Upper middle 5 Middle 6 and 7 Lower Open fallow dominated by Imperata sp.; coconut 8 Middle Banana and coconut 9 Middle Fallow dominated by Pueraria sp. and grasses; coconut; cassava until 2003 10 Upper middle Young secondary forest, mainly Ficus spp.; coconut 11 and 13 Plateau Rainforestation; coconut (extensive use for palm wine) 12 Plateau Fallow dominated by grasses and bush; coconut; previously annual crops

Subplots 1 to 5 and 6 to 10 belong to two different owners and land use was more intensive for the last 10-15 years on subplots 6-10. Subplots 11-13 are located on a plateau approximately 500m away as described in chapter 3 as dystric Nitisol and in chapter 4 as Cienda RF / grassland. Cocos palms are planted randomly across all subplots in distances of 10x10 to 15x15m.

5.1.1 Soil organic carbon (Corg) Responding relatively quickly to management or land use change, soil organic matter was expected to be a sensitive indicator for small-scale variability. 20 auger samples from 0-5 and 7-12cm depths each (Ah and AB horizons) were taken in a systematic pattern68 in each of the 10 subplots and soil organic matter (SOM) was determined by Loss on Ignition method. Values of Corg were calculated on the basis of LoI and clay contents according to the regression given in 2.5.13.4. Fig.53 gives an overview of magnitudes and dispersion of the data. Variability within subplots was lower than expected: Coefficients of variation ranged between 8 and 16% at 0-5cm and 5 – 15% at 7-12cm depth.

68 i.e. in an equidistant grid with every sampling point referenced

• 120 • 5 Plant growth in an agroforestry system under different small-scale environments

Corg 0-5cm Cienda subplots Corg 7-12cm Cienda subplots

Figure 53: Contents and small-scale heterogeneity of organic carbon contents 0-5cm and 7-12cm at Cienda subplots. Note different scales. Letters a-e indicate homogeneous groups based on Waller- Duncan statistics – same letters belong to same groups

Identical letters group subplots, that did not differ significantly (α <0.05) from each other in a one-way ANOVA applying the Waller-Duncan post-hoc test. As not all subplot values were normally distributed at 7-12cm69 depth, the non-parametric Mann-Whitney test was additionally applied on this set. This test is more conservative than the t-test-based Waller-Duncan statistics, but resulted in a very similar grouping. At 0-5cm, the eroded banana subplot 8 was clearly lowest in Corg, followed by a group that contained the open grassland area 6 and 7, a subplot previously planted to rootcrops (9), and to middle slope positions under canopy (4 and 5). Subplot 2, 3 and 10 at the upper end of scale came close to the old rainforestation demo site 11. For 7-12cm depth, differences were less pronounced, basically segregating 1-5, 9 and 10 on one hand from 6, 7 and 8 on the other and subplot 11 forming a distinct class. Subplot 8 ranked better than in the top horizon, suggesting that loss of OM was a recent development caused by erosion under the banana plantation. This was affirmed, when a CAB:CAh - ratio was calculated: Values in all subplots were constricted to tight bounds between 0.74 and 0.78 to the exception of subplots 8, where the Ah was supposedly depleted by erosion (0.86), 9, which had been recently disturbed through harvest of root crops (0.82) and 11, an advanced succession characterised by accumulation of OM (0.85). The most relevant driver for Corg balances was land use, determining vegetation type and management intensity. These, in turn, affect canopy cover and litter inputs. Under grassland and banana, Corg was lowest, increasing with more extensive use and standing biomass, e.g. in subplots 3, 10 and 11. Consequently, the 10 year old reforestation showed the highest Corg contents of all subplots. In context with chapter 4, Corg values converted from LoI still overestimated those measured directly by EA. This became obvious from subplot 11 aka Cienda rainforestation plot.

69 They were for 0-5cm, and variance was homogeneous for both depths.

• 121 • 5 Plant growth in an agroforestry system under different small-scale environments

5.1.2 SOM pools derived by physical fractionation

Different fractions of organic matter have different turnover times in soil (BURESH 1999) and reflect litter quality (CADISCH ET AL. 1996). These fractions can be segregated by a combined size-density fractionation. As an order of magnitude, BALESDENT (1996) calculated turnover times of 0.5 years for the OM fraction >2000µm and 63 years for the fraction <50µm in French Eutric Cambisols. Turnover determines release of nutrients but also structural functions of SOM. For modelling, active, slow and passive SOM fractions are often distinguished by physical methods (for an overview, s. PAUSTIAN ET AL. 1997B). The combined sieve and density fractionation used here (based on ANDERSON & INGRAM 1993 and GAISER 1993) is in accordance with requirements for the CENTURY SOM module on which WaNuLCAS is based70. Following the definition of TSBF (ANDERSON & INGRAM 1993), the active pool contains microbial carbon and OM floating on water, the slow pool comprises non-floating OM of 0.25-2mm in size and the passive pool OM of high density (>1.6g/cm3) smaller 0.25mm. In modification, fractions separated here were AGB >2mm, roots >2mm, charcoal >2mm, light organic matter floating on water (not tungstate) <2mm (LOM) and heavy organic matter <2mm (HOM). Each fraction was weighed and analysed for C, N and P. As a control, the unfractionated sample was analysed for C, N and P, too. Twelve samples were bulked to one composite sample per subplot. Organic matter fractions by weight were dominated by over 98% heavy fraction, which contains all mineral compounds (fig.54). The LOM fraction paralleled results for Corg, namely low contents at subplots 6-8 and high values at 2, 3, 10 and 11.

Figure 54: Percentage mass of soil fractions per subplot Absolute contents of C, N and P as distributed over the various fractions are shown in fig. 55-57.

-3 70 Where the active (mainly Cmic), the slow or light fraction (0.25-2mm and <0.9gcm ) and the passive fraction (>1.6gcm-3) are distinguished. As chemical criteria, polyphenolics and lignin contents enter into the WaNuLCAS model.

• 122 • 5 Plant growth in an agroforestry system under different small-scale environments

Figure 55: Absolute C contents per fraction and plot

Figure 56: Absolute N contents per fraction and plot

Figure 57: Absolute P contents per fraction and plot; PLOM highlighted as line

• 123 • 5 Plant growth in an agroforestry system under different small-scale environments

Subplots did not differ much in total C contents, but the very low values at subplot 8, as well as high contents at 2, 10 and 11 are parallel to the findings for Corg. Assuming a bulk -3 -1 density of 0.9g cm and average CLOM of 0.87g kg (see fig.55 for absolute C contents), -1 the sampled top 15cm of the soil contain 1500kg of CLOM ha . As for C, the bulk of N and P was bound to the HOM fraction. Total N and P supply was lowest on subplots 4, 8 and 9. C was by far lowest on 8, as found before (5.1.1). A group of five subplots was generally best supplied: Number 10 with respect to C, N and P, further 1, 2 and 5 for C and N and 11, 12 for C and P. Low P at subplot 4 seems to contradict the good supply of the Ah horizon at nearby profile PN1; as at Marcos site, P seems to be unequally distributed on a small scale. For subplot 10 and 11, 'other' P was suspiciously low / high71 and data need to be interpreted cautiously. Carbon contents in the unfractionated samples are below those obtained by EA in the respective profiles (chapter 3) and by LoI in 5.1.1. The fractions in fig. 54-56 denominated other refer to the control entire – sum of fractions (16-18% in average). These losses include dissolved organic matter (DOM) containing DOC, DON and DOP, microbial biomass and minor amounts of C and N through the wet sieving process, respiration or denitrification during storage in the pails. The rest was due to the procedure and to charcoal, which was not analysed for C, N, P separately. MAGID ET AL. (2002) allotted charcoal to the water-flotable fraction. In terms of amounts, BALESDENT (1996) evaluated several methods of fractionation and quantified C fractions for modelling with a wet sieving procedure similar to the one discussed here. In his experiments, water-soluble C obtained from particle-size fractionation accounted for approximately 1.5% of soil C. Microbial biomass can make up for 2-7% of Corg (INSAM & DOMSCH 1988).

71 possibly an analytical error

• 124 • 5 Plant growth in an agroforestry system under different small-scale environments

C:N and C:P-ratios (fig. 58 and 59) give information on decomposability of substrates, tight ratios indicating better access for microbial decomposers.

Figure 58: C:N ratios of SOM fractions on Cienda subplots

Figure 59: C:P ratios of SOM fractions on Cienda subplots

• 125 • 5 Plant growth in an agroforestry system under different small-scale environments

Generally, lighter (and here also larger) fractions show the wider C:N and C:P ratios (GREGORICH ET AL. 1994). Despite its tight C:N- and C:P-ratio, HOM is the most recalcitrant fraction and most difficult to access for microorganisms due to pore size and chemical protection of OM (LAVELLE & SPAIN 2005, FELLER & BEARE 1997). Range of C:N in entire samples was 9.5 to 12.2, consistent with values found in Leyte profiles (chapter 3). C:P- ratios in Cienda subplots were between 41 and 68, which is tight compared to 'typical' levels for tropical agricultural soils (ZECH ET AL. 1997). Cross-relations between all fractions and subplots cannot be discussed here, but two extremes shall be presented. The tightest C:N and C:P ratios were found on subplot 8 throughout all fractions except for C:P-HOM. If tightening of ratios is an indicator for nutrient accumulation through illuviation (SCHOENAU & BETTANY 1987) or input of litter, this would mean, that easily decomposable banana leaves or stronger leaching or erosion of C compared to P had caused the relative increase. Light organic matter decomposes faster than the heavy organic matter fraction bound to minerals and protected from microbial attack (LAVELLE & SPAIN 2005). The fact, that C:P-HOM as the least changing fraction remained elevated, supports the first assumption of agricultural intensification72. The opposite process took place at number 4 with respect to C:N-ratios. This subplot had been managed extensively during the last decade and C:N-ratios in the large and light fractions were among the widest. C:N-HOM, however was among the tightest, along with 1-3 and 8. GAISER (1993) found for maize grown in Benin, that the effect of LOM on maize yields doubled that of HOM. LOM is the most readily available fraction and would be of most significance for planting, while HOM stands for nutrient reserves stored in the soil, but not necessarily accessible in the short run. In table 12 the share of LOM in the entire sample was calculated.

72 Still, tight C:N or C:P does not imply sufficient supply of a nutrient, as in the case of subplot 8 PBray was comparably low.

• 126 • 5 Plant growth in an agroforestry system under different small-scale environments

Table 12: Shares of CLOM, NLOM and PLOM in element contents of the unfractionated (entire) samples

Subplot Share of Share of Share of

CLOM [%] NLOM [%] PLOM [%] 1 18.1 3.66 1.32 2 21.5 5.03 1.54 3 18.9 3.90 1.37 4 21.3 4.44 1.60 5 17.8 4.79 1.47 6 15.0 3.37 0.93 7 17.1 3.58 1.02 8 10.1 3.23 1.05 9 18.9 4.53 1.17 10 19.3 4.61 1.01 11 19.5 4.95 1.85 12 16.5 3.51 0.98

A ranking of these parameters shows, that subplots, which were advanced in succession and produced most litter (5.1.5), e.g. 4, 1073 and 11, mostly lead with respect to readily available elements. The ones most behind are the eroded banana plot 8, grassland 6&7 and the young bush fallow 12. Contents in C, N and P depend on the plant species, that produce the litter. When subplots are ranked by absolute LOM dry matter, subplots 2 and 8 with their extreme values of %NLOM and %PLOM move to the midfield. For subplot 2 this means, that biomass is relatively rich in N and P, whereas banana biomass is low in the analysed elements.

5.1.3 Soil and basal respiration

5.1.3.1 Simultaneous experiment of soil and basal respiration

Soil respiration refers to CO2 evolution of all biota in the entire soil profile under natural conditions, while basal respiration is usually determined in a sieved sample without roots 2 under standardised laboratory temperature and moisture. Roughly /3 of soil respiration are ascribed to microorganisms, the rest to animals and roots (SCHACHTSCHABEL ET AL. 1992). Absolute magnitudes of CO2 evolved during both experiments cannot be directly related because of the different units (CO2 per g soil for BR vs. CO2 per square centimetre for soil respiration) and organisms involved, but relative magnitudes between subplots can be compared. For a simultaneous soil respiration – BR experiment, water contents of BR samples were not adjusted. No rain fell from three days before until the end of the experiments and temperatures were recorded. Fig.66 (section 5.1.6) shows soil temperatures in 5cm depth under closed canopy and grassland. In fig.60, soil temperatures during the experiment were monitored in and outside the pails used for soil respiration.

73 except P

• 127 • 5 Plant growth in an agroforestry system under different small-scale environments

Figure 60: Soil temperatures (5cm depth) in and outside pails in subplot 1 during the soil respiration experiment Pails heated up relative to ambient temperatures during midday and kept a more balanced temperature, when the soil outside was warmed by oblique sun rays during late afternoon. Under isothermic conditions as in Leyte soils, the major influence on soil and basal repiration was expected to depend on soil moisture (HASHIMOTO ET AL. 2004). As discussed by DILLY (2001), incubation was at room temperature and natural water contents for better comparison with soil respiration values.

Figure 61: Basal and soil respiration at Cienda (25th and 28th -29th of Aug, 2004), BR at natural water contents. Soil respiration shown as means of two independent samples, BR as means of sevenfold technical replicates of one composite sample per subplot Fig.61 shows that trends between subplots in both trials differ substantially for subplots 1, 2 and 6, but generally follow the same tendency. Temperature effects may have led to overestimation of soil respiration in the open plots 6, 7 and 12. Subplots 1&2 and 6&7 as adjacent subplots showed considerable differences in CO2 release for both methods, pointing to high small-scale variability; adjacent 11&13 did not. Higher numbers of replicates might have been necessary to achieve better accuracy, but at the cost of considerable disturbance of the area.

• 128 • 5 Plant growth in an agroforestry system under different small-scale environments

Extrapolating the evaluation curve for soil respiration provided by SCHLICHTING ET AL. (1995; p.254) to an average soil temperature of 29-30°C, all measured values would be classified as moderate. Typical CO2 effluxes found by SCHWENDENMANN (2002) in Costa Rica were in the order of -2 -1 -2 -1 0.88-1.76mg CO2 cm d (or 100-200mg CO2-C m h ). Levels found in this study are lower, but comparable to those findings, especially if the deeper solum of the mesoamerican soils is taken into account. RAICH ET AL. (2003) modelled a yearly average -2 -1 release of CO2-C for Philippine terrestrial soils of >1400gCm a from 1980-1994 -2 -1 -2 -1 compared to about 600Cm a in this experiment, assuming a value of 0.6mgCO2cm d .

5.1.3.2 'Long-term' basal respiration For a basal respiration experiment of 30 days duration, 12 samples per subplot were bulked and repeated in seven-fold technical replicate. Since BR depends strongly on water contents (LUIZÃO ET AL. 1992) samples were adjusted to 55% of water holding capacity (WHC) and monitored during the experiment, but evaporation losses found negligible.

Figure 62: Basal respiration during a 30-day experiment in cumulative depiction All subplots under canopy were concentrated in the upper half of the graph in fig.62. Subplot 3, under canopy, but prone to erosion, ranks first. Lowest respiration rates were observed in the open areas in subplots 6, 7 and 12 as well as the eroded banana subplot 8. On the other hand, BR also depends on the population size of respiring microorganisms. This can be observed for subplot 9, where elevated BR coincides with an exceptionally large microbial population (s. next section). This large microbial pool may have been caused by disturbance (harvest of cassava in 2003) which led to increased aeration, decomposition and build-up of microbial biomass until a new steady state was to be reached (INSAM & DOMSCH 1988). Contrary to this, low basal respiration for subplots 6 - 8 must be seen in the context of their small microbial population. Trends of cumulative respiration remained stable during the entire period. Values of adjacent subplots 1&2 were close and parallel throughout the entire incubation period, 6&7 and 11&13 almost identical.

• 129 • 5 Plant growth in an agroforestry system under different small-scale environments

BR on an hourly basis is shown in fig.63 for better comparability with references from literature.

5.1.4 Microbial carbon

74 As mentioned, the conversion factor suggested by ANDERSON & DOMSCH (1978) for substrate-induced respiration (SIR) is only valid, if the procedure is carried out at 22°C. This requirement could not be met in 2004. In literature, conversion factors given (BECK ET AL. 1997) are for temperatures from 25-28°C or the original factor is used for 25°C (ANDERSSON ET AL. 2004), but not for 32°C, as in the experiment presented here. Consequently, qCO2 and Cmic/Corg ratio were calculated on the basis of released CO2 instead of Cmic. Twenty samples per subplot were bulked and repeated in fivefold technical replicate. Substrate-induced respiration was compared to hourly BR values calculated from the experiment presented in 5.1.3. Basal respiration in fig.63 is displayed on an hourly basis during different periods of a 30-day experiment (e.g. BR 24d is the CO2 released per hour during the period from day 18 to 24).

Figure 63: Basal and substrate-induced respiration and qCO2 on Cienda subplots 2004. All values given on CO2-basis. qCO2 refers to BR 24d. For all subplots, BR started at a high level from day 1 to 3, probably caused by disturbance and aeration of the soil, when samples are mixed and filled into the nylon bags (see MAMILOV & DILLY (2002) about importance of drying-rewetting cycles for microbial respiration). By the time a relative equilibrium was reached, which varied to some degree depending on ambient temperature. The decline of BR in laboratory experiments would continue beyond the 30 days; this is normal and due to depletion of reserves, which are otherwise constantly supplied on site (ANDERSON 1994; FANG ET AL. 2005). SIR is shown as

-1 -1 -1 -1 74 Cmic [mg 100g dry soil] = 20.6 CO2 [mg 100g dry soil h ] (SCHINNER ET AL. 1993). Originally Cmic [mg 100g -1 -1 dry soil] = 40x + 0.37, where x is CO2 [ml 100g dry soil h ].

• 130 • 5 Plant growth in an agroforestry system under different small-scale environments evolution of CO2 to circumvent the conversion factor. Both basal respiration and SIR are well distinguishable between land uses, with subplots 1-5 (canopy) on a higher level than 6-10 (open area and banana) and subplot 11 (old rainforestation demo plot) in between. An exception to this overall trend is the low SIR of subplot 8, likely to be caused by erosion of topsoil, which contains most microbial biomass. Subplot 9 is in the stage of a population build-up after disturbance caused by cassava harvest. qCO2 is given as BR-24d-CO2 per SIR-CO2 instead of the usual BR per unit Cmic. BR 24d - values were used as they represented an equilibrium state of respiration (ANDERSON & DOMSCH 1986), SIR-CO2 due to lack of a calibrated conversion factor. With qCO2 understood as a stressor (s. 4.5), it could be expected, that an undisturbed area under extensive land use like subplot 11, would show low levels and open areas with temporal water stress, 6 and 7, relatively high values. Subplots 8 and 9 are influenced by the factors underlying SIR, namely erosion and post-disturbance. An overview of Cmic and Corg as well as the quotient of both is shown in tab.13. This quotient has been suggested as a sensitive indicator for soil fertility and correlated to crop yields (KAISER ET AL. 1992), which reacts quickly to land use or management changes and varies within a range of 0.27 – 7% typical for European soils (ANDERSON & DOMSCH 1989). MAO ET AL. (1992) found typical values 1.2-1.9 for reforested plots and of 1.0 – 1.2 for rainforest in tropical China. For the sampled soils, the original formula by ANDERSON & DOMSCH (1978) was used as an approximation to convert SIR-CO2 into Cmic. Values for Cmic and Cmic/Corg are meant as relative, not as absolute figures to be compared to literature values. Table 13: Contents of Corg, Cmic per subplot and Cmic as percentage of Corg

Subplot Cmic Corg Cmic / Corg A&D75 (Ø 0-5 and 7-12cm)

-1 [mg g soil] [%] [% of Corg] 1 1.01 4.73 2.13 2 1.15 4.96 2.33 3 1.05 5.04 2.08 4 1.10 4.67 2.36 5 1.15 4.67 2.46 6 0.72 4.57 1.56 7 0.73 4.57 1.60 8 0.53 4.34 1.21 9 1.33 4.82 2.76 10 0.76 4.92 1.54 11 0.80 5.36 1.50

A clear divide between Cmic/Corg of subplots 1-5 and 6-10 is owed to the higher Cmic contents in the first cluster. The recent disturbance of subplot 9 (harvest of cassava in 2003) was reflected by the exceptionally high Cmic contents, that overcompensate high Corg. Despite its low Corg and due to the even lower Cmic, subplot 8 ranks last in Cmic/Corg. As Cmic precedes Corg in its reaction to land use change, a comparatively high Cmic/Corg

-1 -1 -1 75 After ANDERSON & DOMSCH (1978): 1mgCO2g soil h = 20.6mgCmicg soil, if the respiration coefficient is assumed to be 1 (used for 22°C).

• 131 • 5 Plant growth in an agroforestry system under different small-scale environments quotient is a sign of microbial biomass build-up in progress as in subplot 9 and vice versa for subplot 8. According to DINESH ET AL. (2004), the quotient Cmic/Corg reflects substrate availability, with low values indicating recalcitrant substrate. For the examples shown here, Cmic/Corg reflects the unfavourable conditions at subplots 6-8 and the land use change at 9 well. This agrees with findings by INSAM & DOMSCH (1988), that Cmic precedes Corg in its reaction to land use changes. Following this logic, subplot 11 can be interpreted as advanced succession characterised by humus accumulation and adaptation of the microbial biomass to (unfavourable because acidic etc.) ambient conditions.

5.1.5 Litter production Litter production was monitored in order to characterise plot variability in canopy cover and organic matter cycling. While the latter is generally desirable for plant production, the first can stand for protection, but also competition for light or indicator of nutrient or water competition. Three fractions of aboveground litter were distinguished, namely leaves, flowers & fruit (F&F) and bark & branches (B&B). Leaves was an interesting category with respect to shedding during dry season and changing understorey light regime. Lignified B&B fraction was considered relevant because of slower decomposition rates and F&F in context with reproductive phenology. Litter of all fractions by subplot is shown in fig.64 for the period Jan 2005 – Jan 2006. Leaf litter was identified as the fraction with least dispersion, while F&F and B&B data were distorted by outliers for some subplots. Adding up leaf litter inputs of the entire observation period (Jan 05 – Jan 06), tenure and past land use can be clearly distinguished: Subplots 1-5 under closed canopy differ from 6 and 7 under low-growing Imperata sp., Saccharum spontaneum and ferns and subplot 9 (as 6 and 7 plus creepers and scattered small bushes). Subplot 8 is mid-slope banana land and influenced by the more forest-like area around subplot 10 at the upper slope. Figure 64: Litter fractions (oven-dry) per subplot in Cienda. Total AGB litter follows the Samples collected from Jan 29th, 2005, to Jan 1st, 2006 same tendencies to the exception of subplot 9, where one coconut stalk distorted the otherwise low overall amounts of B&B. The presented amounts cover litterfall during a span of 337 days. Extrapolating this to one year time, total aboveground litter is 471 to 9,340kg ha-1a-1, in the upper third of what -1 -1 KELLMAN (1970) calculated for secondary forests in Mindanao (2,000-12,000kg litter ha a ). -1 -1 ASIO (1996) cites literature data by CUEVAS & SAJISE (1978) of 13.48t litterfall ha a for a rainforest on volcanic soil at Mt. Makiling, Luzon.

• 132 • 5 Plant growth in an agroforestry system under different small-scale environments

Figure 65: Total leaf and fruit litter time series Cienda

Looking at the temporal distribution of total leaf and F&F litter quantities (fig.65), a maximum of leaf fall was observed for all subplots except 6, 7 and 9 in March 2005 following a dry February. A smaller peak occurred in November, caused by subplots 1 and 4. Between these two peaks a maximum in F&F litter occurred at the end of June to the end of August, which could be observed in most subplots (plotwise data not shown). A background 'noise' of coconut flowers was found throughout the year for all plots including 6, 7 and 9. Outliers on subplots 1 and 4 were caused by aborted coconuts.

5.1.6 Litter decomposition According to a hierarchical model of factors steering decomposition, soil temperature and moisture are the most powerful drivers, usually outweighing influences exercised by the type of clay minerals, substrate quality and predation of bacteria by protozoa and other organisms (LAVELLE ET AL. 1993). Soil temperature and water regime differed between subplots depending on canopy cover. As an example, maximum soil temperature (in 5cm depth) during the leaf decomposition experiment in Cienda was >38°C in open areas like subplots 6 and 7 and <31°C for subplots 1, 2 and the rainforestation demo plot installed in 1996 (fig.66). Clay minerals were not expected to differ strongly, but clay contents did (see sections 3.2.1-4): For subplots 4 and 8 (middle slope) 35-40%, for 6 (footslope) clay contents of 45-50% and for the rainforestation plot ('demo site') of 65-70% were found in the topsoil. Litter quality was controlled by a standard substrate and the role of arthropods and other mesofauna for litter decomposition (directly or through predation on microorganisms) was assessed by net tissue of different mesh width (0.1 and 4mm) allowing access to the litter sample for different decomposer groups.

• 133 • 5 Plant growth in an agroforestry system under different small-scale environments

Figure 66: Soil temperature in 5cm depth under selected plots during leaf decomposition experiments (June 4th to 19th, 2004)

The experiment was designed as a comparative short-term assessment of subplots rather than sequential deinstallation to estimate the temporal course of decomposition. Decomposition rates are given as mass loss after certain periods, not as yearly k-rates, as these differ between seasons. During an exploratory experiment with fresh Leucosyke sp. leaves76 on extreme spots in Cienda (fig.67), the comparatively highest decomposition rates in subplot 6 show, that extreme temperatures did not hamper biological activity, at least as long as soil moisture was sufficient. Rainfall during the exposition period (June 4th- 20th, 2004) amounted to 202mm and was evenly distributed. Under these optimum conditions, easily decomposable substrate, high and balanced temperatures and moisture, decomposition ranged between 46 and 75% within 16 days. Where significant differences existed, decomposition of Leucosyke sp. in the 4mm mesh capsules was higher than for 0.1mm. A ranking for fine and thick mesh minicontainers gave an almost identical sequence and a correlation coefficient Figure 67: Decomposition of leaves and Cocos fine roots in of 0.83. Generally, only Cienda 2004 (note that exposition periods were not the same) subplot 6 differed

76 An easily decomposable substrate

• 134 • 5 Plant growth in an agroforestry system under different small-scale environments significantly from the others. During a second experiment with Cocos fine roots (Aug 21st – Sep 4th), rainfall was not more than 80mm most of which fell during two events and it can be assumed, that water stress was relevant at least on the open subplot 6. Ranking differed more between mesh widths and the correlation was only 0.61. Especially for the demo site, the role of the mesofauna is evident. Subplot 6 shows by far the lowest decomposition rates for both decomposer groups. On the other hand, similarly low decomposition rates would have been expected for the slope plot, an exposed SW-slope with grass vegetation. Another root experiment including all ten subplots of the new plantation was set up for a more detailed insight in decomposition dynamics (fig.68).

Figure 68: Decomposition of Cocos fine roots in Cienda, exposition

Decomposition of coconut fine roots was slower than expected and exposition period still too short for clear trends to develop. To cover all subplots, numbers of replicates had to be reduced at the cost of higher coefficients of variation, in average 53% for the 0.1mm mesh and 88% for 4mm mesh. The greater statistical spread for 4mm can be ascribed to higher mobility and food consumption of the mesofauna compared to microorganisms: Decomposition of 100% was found more oftenly in the 4mm capsules than in the fine mesh and young earthworms lived in some of the 4mm capsules. This and the velocity of the process indicate that decomposition of lignified material depended to a lesser extent on saprophytic fungi. Decomposition rates of adjacent subplots 1&2 as well as 6&7 were almost identical. With respect to the relatively high root decomposition rates for 4mm mesh in the least disturbed and most forest-like demo plot, DAUB (2002) found, that certain arthropods play an important role with ongoing succession. Another explanation for the larger contribution of the mesofauna could be the higher clay contents of the demo plot topsoil, which can inhibit microbial activity (LAVELLE & SPAIN 2005) and decomposition. The same applies for pedogenic (Fe-, Al-) oxides, which form organo-mineral complexes inhibiting decomposition (VELDKAMP 1994). Both oxalate- and dithionite- extractable Al and Fe were found in higher concentrations in the demo site soil, but also in profile PN3, corresponding to subplots 6 and 7.

• 135 • 5 Plant growth in an agroforestry system under different small-scale environments

5.1.7 Root length and weight density Roots can make up 20-50% of all carbon inputs in forest soils and especially in the tropics help to stabilise OM rather than easily decomposable leaf litter does (ZECH ET AL. 1997). Fine roots, generally defined as <2mm in diameter are often concentrated in the topsoil (e.g. CUENCA 1983), decisive for nutrient uptake and root litter, but do not contribute significantly to root biomass in old-growth forests (CLARK ET AL. 2001). Distribution and density of roots were interpreted in addition to data on PAR and litter production as all reflect different aspects of stand density. While litter provides the seedlings with nutrients and improves soil climate, the expected effects of PAR were protection from as well as competition for radiation, and the role of roots was mainly competitive with respect to water and nutrients, especially during the dry season in dense stands like subplot 11.

Figure 69: RWD of Cienda subplots. For the bar chart n = 1. The curve indicates composites of 12 samples per subplot as taken for OM fractions analysis. Note that for OM samples threshold diameter was 2mm and for RWD samples 1.5mm One sample per subplot was analysed for root length density and root volume. These exploratory experiments gave a rough insight on distribution of roots in 0-15 and 15-30cm depth, ordered by root diameter classes. Correlation coefficient was 0.69 for total litter production vs. RLD 0-15cm, supporting the observation of GAISER (1993), that root growth depends on the existence and thickness of a litter layer. However, for a statistical analysis, root weight density (RWD) data, as root weight per soil volume [mg cm-3], were preferred. Apart from one sample per subplot taken from 0-15 and 15-30cm and segregated into > and < 1.5mm diameter fractions, another more representative dataset was available. These were composite samples of 12 single samples per subplot each, which had been collected for SOM fractionation. Only roots from 0-15cm depth and of > 2mm diameter were included.The red curve in fig.69 stems from the SOM fractions' sample collective and

• 136 • 5 Plant growth in an agroforestry system under different small-scale environments gives statistically more solid values as each sample is composed of 12 subsamples. Bars show RWD of only one sample, namely the B series of RLD and root volume (not shown). SOM-RWD values correspond to B series-RWD less than to RLD and root volume (correlation coefficients 0.44; 0.51; 0.57) in 0-15cm depth. The most 'forest-like' subplots by appearance, 1-5 and 11, are those with the highest root biomass. RWD 15-30cm shows the relatively higher proportion of the closed-canopy subplots 1-5 with respect to coarse roots and also corrects the very low values per subplot for RLD and volumes. Subplot 11 is clearly higher in RWD 15-30 than 12, an adjacent fallowed field. Subplots 6 and 7, Imperata / fern, are as low as expected in 15-30cm. RWD values are in the same range as such found for one to ten-year old cocoa and Gliricidia sepium systems in Sulawesi by SMILEY (2006). As general trends over all samples, subplots 3-5 and 11 had the highest values for all observed root parameters. Subplot 8 was by far the lowest in all categories. Subplot 9 showed surprisingly high values even for the lower depth (possibly due to loosening of the soil during planting and harvesting of cassava in 2003). As in Marcos, where Gmelina trees were parametrised, roots often extended widely in a lateral direction, especially on clayey horizons in the subsoil (see also AKINNIFESI ET AL.1999).

5.1.8 PAR measurements Photosynthetic active radiation (PAR) is measured in W m-2 or µE m-2 s-1, but expressed here as % of a reference PAR above canopy. Measurements were conducted under clear skies during midday hours. Values in open areas exceeding the reference were set 100%.

Figure 70: Photosynthetic active radiation as percentage of PAR above canopy

• 137 • 5 Plant growth in an agroforestry system under different small-scale environments

Fig.70 shows a map of PAR over the entire plot; note that the upper parts of lines A-D and V-X are outside the plot boundaries and were not planted. These areas appear black. Light regime differed clearly between subplots. Numbers 1-5, under extensive management and denser canopy cover, were generally less exposed to sunlight, numbers 6, 7 and 9 most. An indirect effect of light intensity can be observed in the strip from K-J 12 upwards. The overstorey in this area was relatively open and the undergrowth covered by a thick mat of Pueraria phaseoloides, which strongly affected plant survival. Averages for PAR per subplot are given in table 14. For correlations between PAR and growth parameters (see 5.1.9), values for expanded zones around the subplots were used to increase numbers of plant individuals in each plot and to cover the whole area. Plot margins and corridors between subplots were excluded from these zones. PAR values of the subplots sensu strictu and the expanded zones are listed in table 14, were low percentage indicates dense canopy. Table 14: Average PAR per subplot and expanded area around the subplot

Subplot # Mean PAR Mean PAR subplot [%] zone [%] 1 28 37 2 18 27 3 21 27 4 38 20 5 28 23 6 69 75 7 62 60 8 39 46 9 59 63 10 56 37

In both cases, the less intensive land use under canopy can be ranked apart from the more open area under different tenure. However, subplots 8 (subplot) and 10 (expanded zone) do not differ significantly from 1-5.

5.1.9 Synopsis of environmental parameters Principal component analysis (PCA) aggregates parameters with similar trends to components. This method was used to • facilitate an overview of the numerous parameters and, if possible, reveal underlying influences, • explore the variance between subplots explained by the different components, • reduce the set of parameters for discussion and later to predict plant growth in a multiple regression (5.3.2). The superior explanatory power of PCA to single factors has been underlined for soil microbial parameters BR, SIR and phosphatase by SVENSSON & PELL (2000). DINESH ET AL. (2004) used this approach for a similar experimental question. PCA reduces multidimensional single parameters into bundles or principal components, which explain a maximum of variance in the dataset. A coordinate system is projected in such a way, that the first axis underlies the most relevant component. In case of the applied varimax rotation, orthogonality of each following component to the previous one is required to accentuate contrasts between components.

• 138 • 5 Plant growth in an agroforestry system under different small-scale environments

Parameter groups forming the components are shaded in table 15. Among the various SOM fractions, the proportion of C, N and P in the light organic matter (C, N and PLOM) were preselected, because they were expected to be most relevant for plant growth (GAISER 1993) and microbial communities (ALVAREZ ET AL. 1998). Single and cumulative percentage of observed variance explained by each component are included in the table. The first five among the identified components explain 89% of all observed variance. Table 15: Principal components analysis (varimax rotation with Kaiser normalisation applied) for environmental parameters in Cienda Component Matrix Component Varimax Rotation 1 2 3 4 5 6 PAR -0.817 -0.407 0.057 -0.204 -0.073 -0.066 Leaf litter 0.821 0.390 -0.132 0.065 0.053 -0.061 BR1 0.838 0.130 0.355 0.379 0.018 0.033 BR3 0.764 0.375 0.118 0.174 0.373 0.228 BR6 0.826 -0.100 0.291 0.092 0.242 0.304 BR12 0.909 0.064 0.189 0.255 0.146 0.077 BR18 0.669 0.147 -0.043 0.117 0.661 -0.028 BR24 0.693 0.053 0.228 0.266 0.534 -0.085 BR30 0.573 0.184 0.026 0.420 0.587 -0.153

PLOM 0.777 0.361 0.454 0.044 -0.008 0.174 RWD 0-15cm <2mm 0.166 0.854 -0.033 0.204 0.412 -0.095 RWD 0-15cm >2mm 0.144 0.865 0.037 -0.068 0.357 0.064 RWD15-30cm <2mm 0.179 0.821 0.352 -0.005 -0.176 -0.102 RWD15-30cm >2mm 0.314 0.885 0.212 -0.073 -0.110 0.133

Cmic 0.425 0.112 0.787 0.159 0.390 -0.028

Cmic / Corg 0.394 0.164 0.813 0.049 0.375 -0.040 qCO2 0.121 -0.152 -0.935 -0.120 0.006 -0.184

CorgLoI 0-5cm 0.273 -0.072 0.236 0.847 0.247 0.272

CorgLoI 7-12cm 0.454 -0.287 0.407 0.529 0.243 -0.071

Corg-EA mean 0.348 -0.117 0.274 0.824 0.312 0.089

CLOM 0.213 0.338 0.595 0.648 0.151 0.145 Root decomp 0.1mm -0.090 -0.103 0.116 -0.965 0.009 -0.065

NLOM 0.275 0.259 0.438 0.309 0.582 0.170 Root decomp 4mm -0.050 -0.058 -0.348 -0.155 -0.833 -0.192 Soil resp 0.182 0.015 0.120 0.251 0.099 0.908 Explained variance 50.78 14.10 10.28 7.36 6.47 4.18 Cumulative expl. var. 50.78 64.88 75.17 82.53 89.00 93.18

A first component was related to soil microbial activity as reflected by basal respiration, which is known to be strongly dependent on even-tempered microclimatic conditions (MARTIUS ET AL. 2004). Consequently, basal respiration is hampered by intense solar radiation (PAR, negative coefficient), which causes extreme temperatures and water regimes. On the other hand a closed canopy, as reflected by leaf litter, creates a more equilibrated temperature and water balance (see 5.1.6, fig.60, soil temperature, and 3.2.5, tensiometers) as well as substrate. Light fraction P was also part of component 1. Root weight density of fine and coarse roots at different depths as a second factor was highly self-correlated and confirmed, that the method gave consistent results. Besides, it did not reveal much additional information on underlying interrelations between parameters.

• 139 • 5 Plant growth in an agroforestry system under different small-scale environments

A third component was determined by microbial biomass, alone and as quotients Cmic/Corg and qCO2 = BR/Cmic. As an indicator of stress, qCO2 was oriented opposedly to both other parameters. It was remarkable that microbial respiration and biomass were distinguished by the PCA procedure. Fourth, all parameters related to soil organic carbon, including the light fraction, formed a factor reflecting substrate availability. Microbial decomposition (0.1mm mesh) of recalcitrant coconut roots belonged to the same group, possibly as an alternative substrate in case of lacking easily decomposable material (negative coefficient). Component 5, like number 3, pointed to substrate, integrating labile N and decomposition, in this case of microbes plus mesofauna (4mm mesh). Interestingly, the long-term part of basal respiration (days 18 to 30, in italics), has a similar orientation as this group. This shows, that BR at the beginning and end of the 30- day experiment differed in tendencies. The most interesting components are 1 and 4, as they reveal coherences between parameters, that had not been obvious before. The clear correlation between light-related and basal respiration parameters suggests a causal relationship and so does the correlation between BR and readily available P presented by PLOM. A joint component for Corg as a reservoir and available P has been found in a PCA by JOERGENSEN & CASTILLO (2001) for young volcanic soils in Nicaragua. The case presented here points to P as a limiting substrate for microbial activity. Component 4 showed coherence between total Corg and CLOM, which confirmed the results by independent analyses. A restriction to stability of this analysis was, that plot means were used, so that the number of observations was limited. On the other hand, leaving out single parameters did not result in substantial changes of factor formation. For example, pH was not considered for PCA due to an incomplete data set covering only seven of the ten subplots; still components were the same and coefficients similar to a separate PCA including pH. The applied varimax rotation seeks orthogonality of axes in order to make grouping factors best visible. This procedure identifies the most contrasting factors, but may confine more subtle findings as e.g. a further subdivision of component 1 (CODY & SMITH 1997). For this reason a promax rotation, which does not postulate orthogonality, was run on the same data set. However, the results were very similar in absolute numbers and identical with respect to factor grouping. For further analysis sampling might be reduced to a minimum dataset, which would allow for conclusions about other parameters. For example, CLOM can represent the Corg – group and BR after one day as the most frequently correlated parameter or after 24 days as a more stable measurement are sufficient to represent the entire BR time series. An observation in the crude correlation table had been, that BR correlations decreased in significance from day 1 to day 30. The most important significant correlations between selected parameters are compiled in table 16.

• 140 Table 16: Correlation coefficients between relevant environmental parameters at Cienda. Figures marked * are significant at α = 0.05, ** at α = 0.01 level (two- tailed test).

PAR Leaf litter BR 1d BR 24d PLOM RWD fine Cmic Cmic / Corg qCO2 CorgLoI CorgEA CLOM Root NLOM Soil resp. 0-15cm 0-5cm decomp. 0.1mm PAR 1 -0.916** -0.791** -0.575 -0.830** -0.564 -0.377 -0.348 -0.019 -0.394 -0.401 -0.482 0.303 -0.545 -0.208 Leaf litter -0.916** 1 0.717* 0.558 0.733* 0.565 0.310 0.303 0.091 0.252 0.249 0.299 -0.214 0.347 0.023 BR 1d -0.791** 0.717* 1 0.767** 0.877** 0.328 0.716* 0.662* -0.302 0.643* 0.700* 0.683* -0.409 0.564 0.321 BR 24d -0.575 0.558 0.767** 1 0.613 0.413 0.763* 0.715* -0.124 0.562 0.678* 0.502 -0.325 0.544 0.252

PLOM -0.830** 0.733* 0.877** 0.613 1 0.414 0.712* 0.712* -0.435 0.376 0.394 0.635* -0.096 0.607 0.346 RWD 0-15cm <2mm -0.564 0.565 0.328 0.413 0.414 1 0.333 0.346 -0.109 0.242 0.235 0.490 -0.299 0.554 0.016

Cmic -0.377 0.310 0.716* 0.763* 0.712* 0.333 1 0.991** -0.707* 0.517 0.602 0.740* -0.118 0.719* 0.241

Cmic / Corg -0.348 0.303 0.662* 0.715* 0.712* 0.346 0.991* 1 -0.733* 0.412 0.494 0.692* -0.016 0.685* 0.196

qCO2 -0.019 0.091 -0.302 -0.124 -0.435 -0.109 -0.707* -0.733* 1 -0.345 -0.298 -0.701* 0.039 -0.499 -0.246

CorgLoI 0-5cm -0.394 0.252 0.643* 0.562 0.376 0.242 0.517 0.412 -0.345 1 0.966** 0.802* -0.829** 0.616 0.542

CorgEA -0.401 0.249 0.700* 0.678* 0.394 0.235 0.602 0.494 -0.298 0.966** 1 0.782 -0.778** 0.634* 0.423

CLOM -0.482 0.299 0.683* 0.502 0.635* 0.490 0.740* 0.692* -0.701* 0.802** 0.782** 1 -0.609 0.775** 0.398 Root decomp. 0.1mm 0.303 -0.214 -0.409 -0.325 -0.096 -0.299 -0.118 -0.016 0.039 -0.829** -0.778** -0.609 1 -0.241 -0.304

NLOM -0.545 0.347 0.564 0.544 0.607 0.554 0.719* 0.685* -0.499 0.616 0.634* 0.775* -0.241 1 0.336 Soil resp. -0.208 0.023 0.321 0.252 0.346 0.016 0.241 0.196 -0.246 0.542 0.423 0.398 -0.304 0.336 1 • 5 Plant growth in an agroforestry system under different small-scale environments

Due to the narrow range (5.44 – 6.01), influence of pH was less powerful than for the paired plots discussed in chapter 4. Increasing pH was related to lower SOM and CLOM. As the latter are substrates for microorganisms, pH was also weakly negatively correlated to microbial activity (BR) and biomass (Cmic), and significantly to higher specific respiration (or qCO2 = BR/Cmic) indicating lower metabolic efficacy. Microbial litter decomposition in fine mesh minicontainers and pH were highly correlated. Photosynthetically active radiation (PAR) is part of and represents solar radiation which is attenuated by canopy. PAR measured in the understorey represented density of the canopy cover and was negatively correlated to root weight density and leaf litter production (fig. 71). Intense solar radiation causes extreme surface temperatures and reduces topsoil moisture, which is crucial for microbial activity. Thus PAR can indirectly indicate stress for plants as Figure 71: Correlation of leaf litter production and PAR for 2 well as microorganisms. expanded Cienda subplots. Coefficient of determination r = 0.84 Consequently, PAR was significantly negatively correlated to microbial respiration (fig.72). Correlations of PAR with Cmic, Corg, Cmic / Corg and C, N and P in LOM were also negative. As high PAR seems to coincide with unfavourable conditions not only for microorganisms, the respective subplots will deserve special attention in context with the growth of abaca (5.2). In contrast to PAR, leaf litter stands for the improvement of microclimate – conservation of soil moisture and dampening of temperature extremes – and the provision of substrate for microorganisms and plants. Positive correlations to Cmic have been reported by MAO ET AL. (1992), while in this study correlation with BR was more significant. In conclusion, leaf litter was a more integrated measure for canopy density than a snap-shot captured by a light sensor, but at the cost of higher fuzziness77. Figure 72: Correlation of PAR and BR at Cienda subplots

77 especially, when different tree species are involved.

• 142 • 5 Plant growth in an agroforestry system under different small-scale environments

SOM as source of CLOM, NLOM and PLOM determines the size of the microbial population (Cmic) and respiration as indicated by the significant positive correlations of these with Corg. Positive correlations between Cmic and BR on one hand and LOM on the other have been found by GREGORICH ET AL. (1994). The correlations presented here also agree with findings of SVENSSON & PELL (2000), that Cmic-SIR is controlled by availability of C and N, whereas BR depends more on quantity and quality of carbon, here represented by CLOM. The mutual interrelationships between PAR, leaf litter and microbial activity as well as the context between Corg and CLOM, as explained before, are reproduced in the component clusters of the PCA, which is based on linear correlations. Relevance of the aggregated parameters for plant growth will be assessed later. In summary, tendencies of all C-related parameters including basal respiration and CLOM as well as NLOM and root weight density were concordant over subplots. Reflecting present biomass, shade and management, subplot 6-8 appeared less suitable for tree planting than 1-5 and 10. Subplots could be grouped, the latter accordingly to high contents of the most important parameters Corg, Cmic, CLOM, NLOM, PLOM, BR and leaf litter and low values for PAR and qCO2. Unweighted ranking for each parameter and summing up scores would lead to the conclusion, that subplots 2 and 3 are best for plant growth, followed by the other subplots under canopy. Among 6-10, the latter come close to the first group and 6-8 would be least suitable to grow plants. The following section will show, in how far the measured environmental parameters were reflected by plant survival and growth.

5.2 Plant performance

Data on survival rates and growth of abaca and trees are presented in this section. Data on plant growth will be used later (chapter 6) as a reference for model calibration and validation.

5.2.1 Planted abaca

5.2.1.1 Survival rates Abaca was included into the plot design as early-yielding component and to sequester most carbon during the first years after planting. Harvesting age is reached at 18-24 months, and continuous growth of lateral shoots (suckers) will lead to a relatively steady stock of biomass even after harvest of mature pseudostems at any one time. In practice, mortality was high and growth fulfilled expectations only in few parts of the plot. An important reason for the high mortality rates was inappropriate planting material. Plants had been propagated through tissue culture and were still weak when transplanted. While plants propagated by cormus usually have enough reserves to resprout from the subterranean parts after dry spells or damage, this was not the case for the planting material used on this plot. The majority of losses (45%) occurred during the rainy season 2004/5, so that drought as direct cause of mortality can be ruled out even though two dry weeks after planting might have weakened plants. During the second year another 34% of plants died in the relatively short span between May 5th and July 4th, 2005 (fig.68 biomass), an exceptionally dry period (s. fig.8). Another important factor influencing plant survival was maintenance by the supposed farmer cooperators. This could be observed in the upper part of lines K – N, where Pueraria was growing vigorously and even suffocated some trees as well as in subplots 6 and 7. In the period between July 4th, 2005, and the inventory of Feb 14th, 2006 (not shown in fig.74), the more active owner of subplots 6-10 (lines L-X) retired from the

• 143 • 5 Plant growth in an agroforestry system under different small-scale environments farmers' cooperative as a consequence of the group's lacking commitment. Generally, the left half in fig.74 received more attention and care than the other side. Survival rates of abaca, as recorded at each subplot during inventories are displayed in fig.73.

Figure 73: Survival rates of abaca plants per subplot from July 2004 to April 2006 Overall survival rates dropped from almost 100% (first inventory July 14th, 2004) to 54.9% at May 5th, 2005, to 21.3% on July 4th of the same year and 15.3% on April 30th, 2006. Within these averages, however, variations were considerable.

5.2.1.2 Growth rates Growth was measured as height of the pseudostem from ground level to the point of bifurcation between stalks of the youngest leaves. Height was then converted into dry matter by the use of an empiric regression. Inventories were carried out once during the first year and twice (before and after dry season) in the second and third year (see fig. 74, inventory of Feb 14th, 2006 not shown).

• 144 • 5 Plant growth in an agroforestry system under different small-scale environments

Figure 74: Growth of abaca plants at different inventory dates from 2004-6. Note different scales

• 145 • 5 Plant growth in an agroforestry system under different small-scale environments

The blank patches at the first inventory in fig.74 are inside subplots 1 and 6. These had not been planted as controls in order to monitor effects of abaca. The space M-X 1-8 is open area around subplots 6 and 7, in uphill direction subplots 8 (banana) and 10 (advanced fallow - secondary forest) with best overall growth rates follow. For the right half (subplots 1 to 5), growth was more uniform. For abaca, most regressions in literature (e.g. KELLMAN 1970) are based on diameter as sole predictor for biomass, whereas for the small plants in Cienda height was more meaningful to be measured. HAIRIAH ET AL. (2001) give a diameter-height and diameter- weight regression for banana, but referring to diameters between 7 and 27cm and measured at 135cm height78. Moreover, plantlets bred from tissue culture showed a different habitus during their early development stages compared to abaca or banana grown from corms. Abaca aboveground biomass at Cienda was calculated from destructive measurements of height and dry weight (n=3 over the entire range of sizes) as

B = 0.0772 H1.828, where B is biomass dry matter measured in [g] and H height in [cm]. Coefficient of determination for the empiric equation was r2 = 0.974. Abaca biomass at the consecutive inventories is shown in fig.75 as means of initially planted individuals, with dead plants considered as biomass zero79.

Figure 75: Mean abaca biomass per subplot at different inventory dates

78 Height [m] = 0.7071x0.6835, where x = stem diameter [cm] at 135cm height (r2 = 0.8143). Biomass [kg] = 0.0303x2.1345, where x = stem diameter [cm] at 135cm height (r2 = 0.9887). 79 Means of living plants would have distorted the overall image and comparison of total biomass was not appropriate because of different n.

• 146 • 5 Plant growth in an agroforestry system under different small-scale environments

Applying a Mann-Whitney test to detect significant differences between mean plant biomass per subplot, these were generally more different at the first inventory (approx. 3 months after planting) and then became more uniform towards the second inventory. Exceptions were subplots 8 and 10, where plants started soon to grow comparably better. From inventory 2 to 3, mainly subplot 6 began to contrast more strongly as plants grew slower than in other subplots and finally died. Generally, abaca on subplots 10 and 8 developed best, followed by 7, 2 and 3. For the open area around subplot 6, no live plant remained at the last inventory. Growth was more uniform on subplots 1 to 5, under canopy. Table 17: Differences in abaca biomass between Cienda subplots at the last inventory, April 30th, 2006. Combinations marked with one asterisk are significantly different at α = 0.05, with two asterisks at α = 0.01 level (Mann-Whitney test). P S1 S2 S3 S4 S5 S6 S7 S8 S9 S2 0.374 S3 0.103 0.764 S4 0.167 0.905 0.777 S5 0.438 0.152 0.036* 0.057 S6 0.010* 0.002** 0.001** 0.001** 0.046* S7 0.618 0.237 0.081 0.119 0.814 0.033* S8 0.006** 0.260 0.158 0.124 0.003** 0.000** 0.009** S9 0.029* 0.006** 0.001** 0.002** 0.148 0.340 0.107 0.000** S10 0.000** 0.001** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000**

The large differences between subplot 8 and the following subplots (see tab.17) were not always statistically significant due to the high standard deviations.

5.2.2 Planted trees

5.2.2.1 Survival rates Survival rates were for all tree species higher than for Musa textilis (fig.76). As documented by QUIMIO ET AL. (1998) for the Cienda rainforestation site, even some dipterocarp trees are able to grow well under full sunlight. Percentage survival rates per species at the last inventory are given in the legend of fig. 76. Durian and white lauan were severely affected by the dry summer 2005 (period between the inventories of April and June). In addition breakage during coconut harvest and cutting through the owners were further major causes for mortality. Pests affected mainly kalantas and Figure 76: Survival rates for ten planted tree species from rambutan (inventory data not 2004-2006. Survival rates at last inventory in brackets shown) and abaca (periodic

• 147 • 5 Plant growth in an agroforestry system under different small-scale environments appearance of leaf hoppers). Intercostal chlorosis was observed for most lansones. As for abaca, survival depended to a large extent on planting material. Especially white lauan seedlings had been uprooted just before planting and many plants dried soon. Marang wildlings were also weak, but planted later than other species, so that they were not affected by drought during the first year. Damage and mortality of durian were mainly due to breakage and cutting; another important reason was leaf-shedding as a consequence of the dry summer 2005. Some mangosteen were initially struck by the abrupt transfer from a shade-bed into open areas, but recovered soon under artificial shade of palm fronds. Among the planted timber species S. palosapis was the most pioneer-like, least susceptible to drought and performing best in open areas.

5.2.2.2 Tree growth

Figure 77: Flow chart for tree biomass measurements. All biomass data refer to dry matter. Inventory dates and methods as well as classical biomass measurements to obtain allometric equations have been described in 2.6.4.2 -3, an overview is given in fig.77. This classical determination of plant biomass was used later to calibrate and validate WaNuLCAS data, which are based on habitus, branching pattern (FBA), C, N, P contents, specific leaf area (SLA), leaf weight ratio (LWR), lignin and polyphenolics, among others.

• 148 • 5 Plant growth in an agroforestry system under different small-scale environments

Wood density was mostly determined experimentally, but taken from literature where wood samples were not available. Carbon contents were determined by Walkley-Black method for leaves, twigs, branches, stems and roots of every planted species, Cocos sp. and undergrowth. Specific densities and carbon contents are listed in Annex 5.2.2.2. Leaf length and width were determined for 15 to 20 leaves of different sizes per species and average length-width ratios calculated. For 3 to 10 leaves per species, leaf area was analysed using a scan or xerox and image analysis software. A combined factor of length x width proved as best predictor for leaf area. Linear or quadratic equations were of the form

2 A = y0 + b l w or A = y0 + b1 l w + b2 (l w) where A is leaf area, l is leaf length and w stands for leaf width. Equations passing through the origin were preferred, as they better reflect morphology over the entire range of leaf sizes (for equations see Annex). Coefficients of determination were 0.99 to one. Total woody and leaf biomass were summed up for each species and allometric equations deducted. A first approach was an empiric formula used for forests of the Humid Tropics 2.53 80 independently of site and tree species (B = 0.118D ; BROWN 1997) . This was compared to an approach by V. NOORDWIJK ET AL. (2002), using a site-specific exponent c. Biomass is calculated as B = 0.11DBH2+c where  is wood density and 0.11 an empiric constant and c is derived from H = a Dc which describes the proportions (height to diameter) of a tree.

A generic exponent (c = 0.62) was used as suggested by the authors and in addition species-specific c-factors and their site-specific average (c = 0.934) were tested as modifications. These were compared to the values obtained by own non-destructive measurements as presented above. In modification to the proposed method, diameters at 5cm height replaced DBH as numerous trees were still smaller than 135cm. All models overestimated measured values for AGB (see fig.78). Using the default value c = 0.62 reduced this error (species- specific c ranged from 0.64 to 1.12), and also followed the trend of measured values best81. As a third approach, an empiric equation based on Figure 78: Measured AGB as compared to a generic and a species- specific model predicting AGB from stem diameter 80 Empiric regressions derived by foresters oftenly refer to merchantable height or volumes, not to overall biomass (WEIDELT & BANAAG 1982). 81 Note that different species are plotted in fig.75, so that equal diameters can result in different biomass.

• 149 • 5 Plant growth in an agroforestry system under different small-scale environments the abovementioned geometric concept of squared diameter by height (0.11hD2) as suggested by KETTERINGS ET AL. (2001) was adapted to the measured values. This empiric formula was

B = a (0.11 h D2)b, as shown in fig.79.

As the predictor is a mixed factor of most influencing components, the exponent is almost 1 and the curve close to linearity. On the other hand, constant variation was not met due to height, so that diameter as predictor was preferred. Alternatively, empiric regressions were formulated for each species as shown in tab.18. Figure 79: Empiric allometric equation based on height and diameter after KETTERINGS ET AL. 2001 For these, high coefficients of determination are due to the low number of sampled trees. Still, a generic equation for all ten species gave a similarly high r2.

Table 18: Species-specific allometric equations for planted trees. B = biomass [g DM], D = diameter [cm].

Species Allometric equation Adjusted r2 Dipterocarpus validus B = 19.0824 D3.0651 0.98 Shorea contorta B = 36.5806 D2.7693 0.99 Shorea palosapis B = 2.6216 D 5.9158 0.99 Toona calantas B = 67.2904 D2.0594 0.99 Nephelium lappaceum B = 0.0019 D13.4364 0.99 Durio zibethinus B = 18.9759 D3.4375 0.99 Garcinia mangostana B = 22.8971 D4.3455 0.99 Lansium domesticum B = 34.9073 D2.2848 1 Artocarpus heterophyllus B = 48.4267D2.5907 0.99 Artocarpus odoratissimus B = 32.6774 D2.8576 0.99 Generic all species B = 34.9456 D2.6742 0.95

V. NOORDWIJK ET AL. (2002) emphasise the advantage of species-specific over generic

• 150 • 5 Plant growth in an agroforestry system under different small-scale environments regressions traditionally used in forestry. As few individuals of known species were tested in Cienda, this could easily be realised, but the generic regression presented in fig.80 rendered a surprisingly high coefficient of determination and had the advantage of higher n relevant for statistical accuracy. In the case of rambutan, diameters at stem base did not cover a sufficient range to produce a reasonable calibration curve (exponent >13). For this reason the empiric generic equation of all 10 species was chosen to represent rambutan growth. For allometric equations, the general form B = aDb makes sense biologically as it leads through the origin, contrary to some linear equations which gave better coefficients of determination. The exponential regression Figure 80: Measured and predicted biomass for a generic type has been suggested by equation based on diameter of all ten planted tree species various authors. SMITH & BRAND (1983) compiled regressions for 98 shrubs and small trees including a formula to transform dbh to diameters measured at 15cm height. Their equations covered similar magnitudes as the ones used for Cienda, with a values of 4-80 and exponents between 1.4 and 2.5, exceptionally 3.5. Low exponents would be important for extrapolations into the low and high range, as otherwise biomass of small plants would be under- and of tall plants overestimated. For Cienda this was not relevant, as the smallest and tallest individuals had been used for calibration.

Biomass timber trees Biomass fruit trees

Figure 81: Mean tree biomass based on empiric species-specific equations.

• 151 • 5 Plant growth in an agroforestry system under different small-scale environments

Growth of the different species between inventories is shown in fig.81. These values are based on the empiric equations of the form B = aDb as mentioned before (tab. 18). Mangosteen and lansones are known to be slow-growing trees, while the two Artocarpus- species are fast-growing. A. odoratissimus was planted as uprooted wildlings and trees were smallest at the beginning, but caught up with all timber species. For mangosteen, some small trees beneath a mat of weeds were overlooked during the inventory in Dec 2005, so that the average for this date was overestimated (see arrow in fig.81, right side). Deviations from means are large, because growth is influenced by a number of interacting factors. These data have been adjusted in so far, as biomass was calculated on the basis of diameter, not height, so that mechanical breakage of tips is not accounted for.

5.3 Environmental conditions for abaca growth

Given the environmental data presented before, it was of interest to relate these to survival and growth of plants. It was assumed, that factors influencing both might not be the same or of unequal importance. Abaca was selected to evaluate growth conditions, because it contained the highest number of individuals and because differences between subplots were most pronounced.

5.3.1 Survival rates of abaca In order to estimate the impact of factors for survival / mortality of abaca plants, logistic regressions were formulated to predict the probability of survival as depending on environmental parameters. A logit transformation is used to topple results of the predictor equation into categorical values, either 0 (here: dead) or 1 (here: survival). Thus, the odds of a chosen reference event (here: survival) can be predicted at a certain probability level (CODY & SMITH 1997). Regressions were produced for the abaca inventories on May 5th and July 4th, 2005. These were the dates with the most balanced numbers of events and non-events. In a first ranking, the most significant single factors for the predictor equation were identified and then integrated into the logistic model as long as they were significant82, contributed to better fit of the model, and increased either concordance or sensitivity / specificity of the model. th The selected logistic model for May 5 , 2005, contained the independent parameters Cmic / Corg, BR, Corg 0-5cm, leaf litter production, slope position and decomposition. Leaf litter production as an integrated descriptor of light, soil temperature and moisture was given preference over PAR as a predictor for mortality, because the latter reduced model fit below acceptable thresholds. The parameter slope is a positive equidistant index starting with 0 from the footslope. It was entered because of its numeric, not causal explanatory power, as it represents an aggregated item of not well-defined components. Erosion processes, as shown in chapter 3, may influence this factor. Other implicit factors are management and land use history, which differed in the lower and upper part of subplots 6-10. Coefficients and p-values of the variables are presented in tab.19. Negative coefficients indicate negative influence of the respective parameter on the odds of the reference event survival to occur; this is also expressed by odds ratios < 1. All factors were significant at the 95% level (see Pr > 2).

82 If variables are intercorrelated, they will influence each others' significance when entered into the regression.

• 152 • 5 Plant growth in an agroforestry system under different small-scale environments

Table 19: Selected parameters for logistic regression with their coefficients and p-values. Dependent variable was survival on May 5th, 2005.

Parameter Coefficient Odds ratio Pr > 2 Intercept -23.1622 0.0018

Corg 0-5cm 3.9790 53.462 0.0012 Decomposition roots 0.1mm 0.5213 1.684 0.0026 Leaf litter production 0.0260 1.026 0.0009 Slope 0.0203 1.020 0.0121

Cmic / Corg -0.7766 0.460 0.0014 BR 6d -1.1798 0.307 0.0089

Pearson (P = 0.5098) and deviance (P = 0.2025) tests for goodness of fit showed that the model predicted the dependent data well. Concordance of expected and predicted events was 61.1% (at 1.6% ties), meaning that in three out of five cases events occurred as predicted. The equation to predict odds of survival reads as follows: -1 -1 Log (odds) = -23.1622 + 3.9790Corg [%] + 0.5213decomp. [%] + 0.026 leaf litter [kg ha a ] -1 -1 + 0.0203 slope [ ] - 0.07766Cmic/Corg [%] - 1.1798BR [µg CO2 g h ]. Table 20 shows the number of events and non-events and their proportion, their correct or non-correct prediction and the deducted sensitivity and specificity of the model at selected probability levels. Sensitivity or true positive rate characterises the proportion of correctly predicted reference events, specificity or the true negative rate the share of correctly predicted non-events. To predict the non-event 'mortality' on the basis of the given parameters, a high specificity would be required; this would be met at a probability level of 0.82, but at the cost of no sensitivity. To make a statement of balanced specificity and sensitivity, a cut-off at 0.620 could be chosen, at 59.2 and 53%, respectively.

Table 20: Classification table for the logistic regression of inventory 050505.

Correct Incorrect Correct Sensiti- Specifi- False False vity city positive negative Probability Event Non- Event Non- % level event event 0.440 519 0 313 0 62.4 100.0 0.0 37.6 . 0.500 464 59 254 55 62.9 89.4 18.8 35.4 48.2 0.600 331 149 164 188 57.7 63.8 47.6 33.1 55.8 0.620 307 166 147 212 56.9 59.2 53.0 32.4 56.1 0.700 90 289 24 429 45.6 17.3 92.3 21.1 59.7 0.800 46 298 15 473 41.3 8.9 95.2 24.6 61.3 0.820 0 313 0 519 37.6 0.0 100.0 . 62.4

Among all tested logistic regressions for the inventory of July 4th, 2005, the same selected variables (see table 21) as included for May 5th rendered the best combination for goodness of fit, significance, concordance and sensitivity / specificity.

• 153 • 5 Plant growth in an agroforestry system under different small-scale environments

Table 21: Selected parameters for logistic regression with their coefficients and p-values, ordered by odds ratios. Dependent variable was survival July 4th, 2005.

Parameter Coefficient Odds ratio Pr > 2 Intercept -23.3842 0.0103

Corg 0-5cm 3.9578 52.341 0.0072 Decomposition roots 0.1mm 0.4504 1.569 0.0395 Slope 0.0510 1.052 <.0001 Leaf litter production 0.0341 1.035 0.0027

Cmic / Corg -1.3541 0.258 <.0001 BR 6d -1.6191 0.198 0.0021

Coefficients of all variables were of the same algebraic sign and, if compared to each other, of similar magnitudes for both inventories. Between inventories, however, odds ratios for BR and Cmic/Corg decreased to 64 and 56% of their respective values at the May inventory, giving them even stronger negative impact on the event survival to occur. Concordance reached 74.8% (at 1.5% ties), but at clearly lower probability levels, with a balanced cut-off for sensitivity and specificity at 0.260 (table 22). Table 22: Classification table for the logistic regression of inventory 040705.

Correct Incorrect Correct Sensiti- Specifi- False False vity city positive negative Probability Event Non- Event Non- % level event event 0.040 214 0 618 0 25.7 100.0 0.0 74.3 - 0.100 203 122 496 11 39.1 94.9 19.7 71.0 8.3 0.200 158 365 253 56 62.9 73.8 59.1 61.6 13.3 0.260 146 426 192 68 68.8 68.2 68.9 56.8 13.8 0.300 126 469 149 88 71.5 58.9 75.9 54.2 15.8 0.400 73 577 41 141 78.1 34.1 93.4 36.0 19.6 0.500 73 577 41 141 78.1 34.1 93.4 36.0 19.6 0.600 73 577 41 141 78.1 34.1 93.4 36.0 19.6 0.700 0 618 0 214 74.3 0.0 100.0 - 25.7

When a full logistic model, including all parameters, was formed, results would hardly improve compared to the reduced models above. This was mainly due to the exclusion of variables which were linear combinations of other variables. Most additional parameters were of low significance. PAR as such did not contribute significantly to the explanation of survival / mortality rates (P > 0.75). On the other hand leaf litter and BR, which were highly correlated to PAR, did. This is remarkable, as PAR values were available in high density and leaf litter data as subplot means. Yet, considering that the effect of light on plants follows an optimum curve, it is difficult to encounter any linear or logistic relationship between PAR and survival. From a common sense perspective it appeared obvious from an early stage, that drought was the main causal factor for mortality of abaca plants. The species has been described as typical for successional stages after selective logging (KELLMAN 1970) like banana in secondary forests (MILZ 2001), which implies protection through a closed canopy. ECKSTEIN & ROBINSON (1996) found, that water stress during few consecutive days can result in

• 154 • 5 Plant growth in an agroforestry system under different small-scale environments massive reductions of photosynthesis in banana83. As mentioned before, mortality was high within a short period of time, indicating that it was rather caused by stressors than lack of nutrients. Consequently, available N and P from the LOM fraction did not play any role for survival in the regressions. Large leaf area and low mutual shading of leaves make Musaceae susceptible to transpiration stress and temperatures of 38 – 40°C can halt leaf growth (RODRIGO ET AL. 1997). Although drought delimits nutrient availability and uptake, dehydration due to high transpiration would be the more probable reason for mortality. Another indicator pointing to influence of water was the importance of Corg. Soil organic matter serves as a reservoir for soil water and interrupts capillary rise from the subsoil. If the nutrient aspect had been the crucial component of SOM, then CLOM, NLOM and PLOM would have been of major importance, which was not the case. One factor that changed clearly during the experiment was management of the plots. While this consisted of uniform weeding after installation and declined to zero towards the second year, the owner of subplots 6 - 10 started then to engage more on the upper part of his field, subplots 8 and 10. In a single-factor logistic model based on management as a categorical variable, concordance rose from 0 to 29, then 46 and remained at 44% for the subsequent inventory dates. This indicates, that management played an increasing role for survival from the second year on. The relevance of planting material and management could be observed at a comparable abaca field planted under full sun, but from corms, and being weeded continuously. The plantation showed survival rates of nearly 100% and uniform growth, comparable to some of the best plants at subplot ten in Cienda. The plantation had been installed in August 2004 at LSU (the plot has been described in chapter 4 as LSU annuals paired plot) and showed Corg levels below most Cienda subplots. With respect to the planting material, it was supposedly the small size of the plants rather then tissue culture techniques per se, which led to high susceptibility to drought stress. ECKSTEIN & ROBINSON (1996) remarked that tissue culture was becoming more popular in the 1990s, because banana plants produce up to 60% more root DM and 100% more leaf area during the first five months compared to suckers. Under water stress a stronger increase in leaf area than root biomass may be fatal. As a rule of thumb for South African banana growers, ECKSTEIN & ROBINSON recommended, that soil water potential should not drop below -15 to -20kPa for young plants from tissue culture. This corresponds to pF 2.3. As shown under 3.2.5, potentials as low as 2.8 or -60kPa were reached at subplot 6 during dry spells. PAR artificially reduced to 31% of open area values led to a decrease in photosynthesis of banana to 73%, but transpiration decreased even more sharply to 62% (ECKSTEIN, ROBINSON & FRASER 1997). Total cropping cycle (planting – harvest) was elongated by 11 weeks and the time span from planting to flower, corresponding to harvest for abaca, by almost 6 of 53.3 weeks.

5.3.2 Abaca growth Multiple regression procedures were used to predict abaca growth from the parameters presented before. As dependent variables, two types of plant data were used: Biomass calculated from heights at each inventory and growth as difference between two inventories. This was done to reduce covariance caused by potentially unequal initial biomass at planting. On the basis of subplot means, linear regressions were formed. In a first approach, a reduced set of independent variables was selected using a maximum r2 procedure. Among the best predictors, the backward elimination procedure was applied. Strongly

83 Under South African conditions, 12 days of water stress caused photosynthesis to drop by 79%.

• 155 • 5 Plant growth in an agroforestry system under different small-scale environments correlated or similar parameters were reduced to as few representative variables as possible in order to avoid overfitting. For the same reasoning, it was aimed at employing PCA components integrating various single parameters. In addition C, N and PLOM as percentages (C, N and PLOM%) in the entire sample were replaced by absolute concentrations in [mg kg-1] after some preliminary tests; these parameters are denominated C, N and PLOMabs. Equations listed in table 23 with their respective coefficients of determination and P-values were selected to best predict biomass and growth. Table 23: Multiple regressions for prediction of abaca aboveground biomass and growth.

Dependent variable Equation r2 P

th Biomass July 24 , 2004 y = -112 + 0.13PSOM + 486.4NLOMabs -0.21qCO2 -15.7BR30 + 0.9665 0.0047 20.81 Corg 7-12

th Biomass May 5 , 2005 y = -493.9 + 106.8 Corg 0-5 -2.9SIR -192.3CLOMabs + 9512.4 0.9682 0.0006 NLOMabs

th Biomass July 4 , 2005* y = 42.4 -78.9Cmic/Corg + 5100.9 NLOMabs + 31.7PCA4 0.8884 0.0029

th Biomass April 30 , 2006* y = 69.6 + 39782 NLOMabs+ 145.8 PCA 4 -9.7SIR - 757.9CLOMabs 0.9769 0.0003

Growth inventory 1 – 2* y = 22.8 +8421.5 NLOMabs+ 27.2 PCA4 -1.8SIR -167.6CLOMabs 0.9706 0.0005

Growth inventory 2 – 3 y = 347.4 -56.4Corg 0-5 + 13.9 NLOM% -19.6PLOM% -11.1 0.9712 0.0005 decomp0.1 *Intercept not significant

All variables were significant at α = 0.05. The regression formulated for the first inventory date was exceptional in number and type of parameters. Corg – in the form of the single parameter or integrated in PCA component 4 - was represented in all regressions, and so was NLOM. Parameter Cmic – as SIR-CO2-rate, qCO2 or Cmic/Corg – entered in all but the exceptional last regression, where average growth was mostly negative. Predicted and observed biomass and growth values are shown in fig.82 for biomass at the inventory of July 4th, 2005 as an example. Coefficient of determination was close to 1 for all biomass and growth data. The graph illustrates, that a disproportionate part of the trend is owed to the good plant performance on subplot 10, but also, that most other points fit very well into this tendency. For the inventory of July 4th, 2005, and growth between the first two inventory dates, a sensitivity analysis was carried out to assess the impact of the dominant parameters (fig.83). All except one predictor were kept stable at the mean value of all subplots and the remaining variable was varied from 10 – 190% of its average value. Figure 82: Predicted vs. observed biomass for the 3rd inventory

• 156 • 5 Plant growth in an agroforestry system under different small-scale environments

Growth1 Biomass 04/07/05 Figure 83: Sensitivity analysis of variables for the regressions on growth from July 24th, 2004 to May 5th, 2005, and on biomass at the 3rd inventory (July 4th, 2005).

As recognisable from the steepness of slopes, NLOMabs had the strongest influence on growth, followed by CLOMabs and SIR; PCA 4 was almost horizontal and did not affect growth. At the third inventory, Cmic/Corg and NLOMabs had similar though opposed impact on biomass, whereas PCA 4 again was not influential. Removing PCA 4, however, led to a drastic reduction of r-square, implying intercorrelations between variables. On the other hand, PCA 4 could be replaced by Corg0-5cm without much decrease in fit. It is known, that banana plants require fertile sites rich in organic matter and of slightly acidic to neutral pH (REHM & ESPIG 1996) and the same has been assumed for abaca as a close relative (KELLMAN 1970). Usually advanced fallows or sites under secondary forest are selected for abaca plantations. For this reason, factors connected to carbon and nitrogen as well as microbial carbon contributed to the fit of equations whereas light- related parameters did not and P only for two of six regressions. The relevance of the LOM fraction for N availability for plants has been highlighted by ZECH ET AL. (1997). BARRIOS ET AL. (1997) state, that trees have the potential to increase availability of N compared to annuals and the explanatory power of N for the fast-growing abaca plants was confirmed for both sets of equations. On the other hand, litter and soil organic matter are important for the water holding capacity of soils, so that an effect of water balance cannot be ruled out for any equation. Each multiple regression presented here is only one of several possibilities to predict biomass/growth from a set of preselected parameters, which is then reduced by the elimination procedure due to the explanatory power of the single factors. In a first approach, some regressions had been formulated for PAR, among others, showing a negative impact of increasing PAR on biomass and growth. Relationship of PAR and biomass / growth was negative at all times, implying that radiation in the observed range was connected to a stress factor. This negative effect of PAR has been discussed before. With respect to the dependent variable growth 2, it could be assumed, that stress-related variables would have had a statistical impact on the negative rates. Still, parameters PAR and qCO2 did not meet the required significance level to remain in the model. PAR as a single parameter was not coherent with growth. Measurements conducted at

• 157 • 5 Plant growth in an agroforestry system under different small-scale environments tree positions (s. 5.1.8) and intrapolated to the corresponding abaca locations were related to the height measurements at each of the five abaca inventories. Direct interrelations, mainly optimum curves, between abaca growth and PAR were calculated, but never of r2 > 0.20. As shown before, biomass and growth were clearly highest on subplot ten. If subplots are ranked by soil parameters, this subplot was above average for CLoI, CLOM and NLOM. Subplot eight, with the second highest growth rates, ranked last for CLoI, CLOM, NLOM and Cmic and first for pH and qCO2. With respect to qCO2, subplot 10 was second. With respect to PLOM, both are low in contents compared to the other subplots. This is remarkable as P is frequently assumed to be the most limiting factor for plants on volcanic soils in Leyte. Subplot 2, the supposedly best subplot for planting, if judging from the environmental parameters discussed in 5.1.9, was far behind 8 and 10. Interestingly, banana had been traditionally planted to greater extent only on the two subplots, that proved to be best for growth, namely 8 and 10. It would be worthwhile to investigate farmers' criteria and site indicators for planting banana and abaca. Two contrasting possibilities are that, a.) farmers have an empiric knowledge on additional factors which were not considered in this study and which benefit abaca/banana growth especially on sites with the lowest fertility, or b.) banana has always been planted on these spots and will be grown there as long as soil fertility allows it. The good performance at some spots would then be attributable to more regular weeding on the traditional banana land.

• 158 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

6 Modelling growth and carbon sequestration of agroforestry systems in Leyte Plant growth at various sites was modelled using WaNuLCAS, a model for Water, Nutrients, Light and Carbon in Agroforestry Systems (V.NOORDWIJK, LUSIANA & KHASANAH 2004). The model runs in a STELLA® environment, which is combined with an Excel spreadsheet as front end. For this study, WaNuLCAS version 3.1 was used. Several steps need to be taken, before scenarios can be run in the model. In fig.84 the procedure is shown for crops, while for trees, a parametrisation software called WanFBA is used instead of WanHelp. Soil and weather data are directly entered into the spreadsheet or Stella file.

Figure 84: Flow chart crop parametrisation

Optional help modules such as WOFOST or WanHelp can be employed to derive input parameters, which are not at hand, from parameters, which are easier to measure. For tree parametrisation, a tree survey including semi-quantitative questions to farmers is part of the WanFBA and Tree Parametrisation modules. Entering field or literature data into the system (parametrisation) and calibrating, i.e. fine-tuning the model using field data, precede validation, i.e. running the calibrated model on additional measured datasets. Finally, hypothetical scenarios can be run on the calibrated and validated model. Mature plants were used for parametrisation of the model, which was then calibrated and validated at the Cienda subplots 6, 7, 8 and 10 on young plants as well as at the Cienda demo site and LSU on >10year-old trees. On this basis, different land use systems were compared as regards to their carbon sequestration potential over a period of up to 20 years. These land use systems are grassland, Gmelina pure stands and rainforestation as presented in chapter 4. For rainforestation, different options were explored to optimise the system.

6.1 Parametrisation

Practically, entering data into WaNuLCAS consists of three steps (s. fig. 84): Optional help programmes like WanFBA for tree parametrisation or WOFOST and WanHelp for crop parametrisation are fed with measured biomass values and produce plant physiological data, which are then entered into a spreadsheet denominated Wanulcas.xls. Direct inputs

• 159 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte on soil, weather, profitability and management activities – planting, weeding, harvesting etc. at certain times – are also specified in this file. All soil, weather, plant and economic data fit into the spreadsheet are internally processed and transferred into an .stm (Stella) file, the actual model. The stella file allows for extensive fine tuning of nutrient, SOM and litter pools, soil physical properties, erosion, management, pests or maintenance respiration, among others.

6.1.1 Crop parametrisation Technically, the parametrisation procedure has been outlined in section 2.7 for crops and trees. There was only one crop, abaca, to be parametrised. Instead of WOFOST, data obtained as interpolations of biomass inventories were directly entered into WanHelp. These were dry weight of leaf, stem and storage organs as well as SLA at different phenological stages. Growth stages in WOFOST and WanHelp range from 0, germination, to 2, maturity, where 1, flower, is defined as the point of transition from the vegetative into the generative phase. Abaca is harvested before flowering, so that the generative phase is never reached. Another characteristic is, that the first vegetative phase until fibre harvest lasts 21 months, while subsequent harvests take place every six months. Due to the existing rootstock, growth is faster than during the initial phase. To account for this, it was assumed, that after harvesting the two tallest stems (of 3m), another two, of 2m and 1.5m height, are left while all other suckers are cut. For parametrisation this meant, that growth of the second and following suckers was defined as generative phases and that the plant would return to stage 1 after rattooning. N and P concentrations in tissues of both stages were maintained at the same level. Specific leaf area (SLA) was measured after >1 year on grown-up plants. For the initial stage, half leaf thickness was assumed and decreased linearly. Thus SLA decreased from 200 to 100% of the final value in phase 0-1, and during phase 1-2 a decrease from 150 to 100% was assumed. For leaf weight ratio (LWR), the relative weight of different plant organs, changing proportions during development were interpolated from own measurements as leaf:stem ratio = -0.0122+0.0011dap (r2=0.98). The cycle from phase 0-1 was then also applied for 1-2. For harvest allocation, two approaches were compared: In one case it was set zero, as fruits or storage organs are not developed; in the other values derived from WanHelp were adopted; after extensive testing over various environment, the second option was preferred. For root allocation, standard values from other species parametrised for WaNuLCAS were used as these are very similar across crops; in contrast to those species, root allocation was maintained at a constant level throughout the entire life cycle as is for the other crops during phase 0-1. Light use efficiency (LUE) was calculated through WanHelp as growth during the respective phase divided by maximum growth at the given relative light capture. Literature regarding light use efficiency (LUE) of different Musaceae was taken into account for the calculations (RODRIGO ET AL. 1997 for banana; TSEGAYE ET AL. (2003) for Ensete ventricosum and TURNER (1972a) and STOVER (1982), cited by TSEGAYE ET AL., for banana). RODRIGO ET AL. (1997) found average LAI of 1.33 from 8 to 28 months after planting in a comparable banana system84 in Sri Lanka, compared to 1.79 calculated by WanHelp for abaca after 24 months for Cienda under optimum conditions. Observations by TSEGAYE ET AL. (2003) on maximum LAI of enset were between 3.2 and 4.6, compared to 4.8 reported by TURNER (1972b) and 3.2-4.3 by STOVER (1982) for banana. For crop parametrisation in WaNuLCAS an intermediate value of 3 was chosen. The light extinction coefficient k, in WaNuLCAS Cq_klight, was estimated 0.58, after

84 wet lowlands, soil pH 4.84; 1500 plants ha-1, interplanted with rubber.

• 160 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte measurements carried out on enset by TSEGAYE ET AL. (2003), namely 0.56- 0.62, and a range of 0.46-0.75 reported by TURNER (1990, as cited by TSEGAYE ET AL.) for banana. For parametrisation of abaca roots, own measurements of banana root length density (Cienda subplot 8) were used (WaNulcas parameter Rt_ACType = 0). A root system extending to 40cm depth and 125cm in radius as reported by MURTHY & IYENGAR (1997) for several banana varieties may develop during several years of growth; for the two-year old abaca at Cienda, only few individual roots would extend beyond 1m of lateral distance from the stem. Abaca polyphenol contents were analysed in Hohenheim and are used in table 24. Regarding lignin, literature values were adopted: Abaca fibres examined by MORENO ET AL. (2005) contained 6.43-9.54% for the most preferred varieties (selected for low contents), while DEL RIO & GUTIERREZ (2006) found 13.2%.

6.1.2 Tree parametrisation Among the ten planted tree species, Shorea contorta, Dipterocarpus validus, Durio zibethinus and Artocarpus heterophyllus were selected for modelling due to economic potential, solidity of datasets and performance in the field. For the reference land use Gmelina arborea was also parametrised.

6.1.2.1 Aboveground architecture General data on habitus and phenology were gathered through expert interviews and literature studies. FBA procedure as specified under 2.7 was conducted for mature individuals as suggested by the model developers (MULIA ET AL. 2001; MULIA 2001). The advanced stage of seedlings at planting is considered as initial stage in the spreadsheet and in the tree planting section of the stm-file, where field data for initial stem height and biomass are entered. All inputs entered into the tree parametrisation spreadsheet are listed in the Annex.

6.1.2.2 Belowground Roots of D. validus, S. contorta, G. arborea and A. heterophyllus were studied by a modified FBA procedure (MULIA 2001), but measurements did not match the criteria set for WaNulCAS in any case. Usually, a thin turn-off from a proximal root would blast the required proportionality factors p and q. In this respect, setting a minimum diameter for measurements might have been appropriate, but in turn would have led to more extensive excavations and damages in the dense plantations. Instead, potential root length density was estimated for each planting zone (in Stella: Rt_TType=0) departing from the measurements described in chapter 5. Root length density was assumed to decrease exponentially from initially 3cm cm-3, the measured average RLD at Cienda demo site (2.93cm cm-3 at a maximum distance of 1.1m from the trees85). A species-specific decrease could then be estimated assuming an elliptically skewed gradient as for WaNuLCAS root type 1. As observed by MASRI (1991) for durian, root biomass would decrease to 15% of its initial value towards the canopy edge. Depth distribution of roots was estimated from knowledge of effective rooting space (chapter 3) and experiences gained during root excavations. Root length density and distribution of Gmelina roots were estimated based on data of MERCADO ET AL. (2005) and RUHIGWA ET AL. (1992). Both data sets are based on fine roots <2mm diameter. This diameter class accounted for 89.3% of all roots at 0-15cm depth and 86.3% at 15-30cm across subplots 1-13 at Cienda (coefficients of variation 5% and 85 Planting distance of 2x1m

• 161 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

11%, respectively)86. Lateral extension of Gmelina roots at Marcos site clearly exceeded 200cm from the stem and was not homogeneous as reported by RUHIGWA ET AL. (1992). This can be ascribed to the physiologically shallow soil (lateral water flow on a clay horizon).

6.1.2.3 Polyphenolics contents of plant tissues Polyphenolics and lignin are classes of substances, that retard litter decomposition and thus influence turnover rates and nutrient cycling. Certain polyphenols reduce growth and activity of decomposers, while lignin as a major component of leaves is hard to metabolise if compared to other structural molecules such as cellulose (ISAAC & NAIR 2005). GAISER (1993) found humification rates of mulch to be correlated to polyphenol but not lignin contents. Data are required for both above and belowground parametrisation and polyphenols were analysed for leaves and fine roots (table 24). Table 24: Total extractable polyphenolics (TEP) contents of fresh leaves and fine roots

TEP [%] Leaves Fine roots D. validus 6.65 8.46 S. contorta 9.75 13.1 S. palosapis 6.33 n.d. T. calantas 9.24 n.d. G. arborea 1.56 2.14 N. lappaceum 9.80 9.17 G. mangostana 5.07 n.d. D. zibethinus 1.36 7.31 A. heterophyllus 3.70 n.d. A. odoratissima 6.48 5.74 L. domesticum 3.20 4.14 M. textilis 1.09 2.51 C. nucifera 9.26 2.56

Values for leaves and fine roots of the same species were similar with exception to durian and coconut. Gmelina contained clearly less TEP compared to the native timber species and was similar to abaca tissues. Ranges were comparable to those stated by various literature sources: ISAAC & NAIR (2005) found 2% of polyphenols in leaves of A. heterophyllus and PALM ET AL. (2000) report median values for polyphenols in fresh leaves of 3%, though at considerable dispersion with many species >6%; fine roots were generally lower in polyphenols with a median of <2% TEP. For lignin contents, default values provided by the model were used. Where literature data or own observations justified this, modifications were made. As an example, ISAAC & NAIR (2005) measured 15.2% and JAMALUDHEEN & MOHAN KUMAR (1999) 17.9% lignin for leaves of A. heterophyllus.

6.1.3 Site data

6.1.3.1 Weather WaNuLCAS uses an elegant approach to simulate climate, based on the three

86 From the relative decrease of roots with depth, a factor denominated Rt_TDecDepth in Stella of 2 and 2.89 was calculated. This factor could have been employed alternatively to Rt_TType=1 or 2, in contrast to type 0, which was used here.

• 162 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte parameters rainfall, soil temperature and potential evapotranspiration. These can be randomly generated or read from monthly or daily observations. For this study, daily rainfall and potential (pan) evaporation recorded by PAGASA at LSU served as direct meteorological input data. Measured pan evaporation was significantly correlated to calculated FAO Penman-Monteith ETP87 (Pearson correlation r2 = 0.705 at α = 0.01), derived from general geographical data, so that few missing values caused by pan overflow could be substituted by theoretical ones. Soil temperatures were calculated from a quadratic regression of mean air temperature at PAGASA-LSU on measured soil temperatures at Cienda site between 2004-6; minima of 9 outliers (of 4380) were set 19.0°C. Four years (Jan 1st, 2003 – Dec 31st, 2006) of calculated soil temperatures, measured rainfall and ETP were used as a loop in Stella, coinciding with real time weather and extended back- and forwards for the simulations.

6.1.3.2 Soils Analytical data as presented in the previous chapters were entered for parametrisation of the following parameters: Corg, texture, bulk density, estimates for field capacity and rough estimates of saturated hydraulic conductivity (Ksat) were used to derive pedotransfer functions and PBray values inside the WaNuLCAS spreadsheet file. Further, stone contents and horizon thickness, pH KCl, C, NT, C:N and N:P ratios of different soil organic matter pools were fed directly into the respective Stella files.

6.1.4 Management and profitability Management was reduced to a minimum and no external inputs, burning, grazing or other special treatments were applied except pruning (for one later scenario). Model input data concerning labour costs and demand as well as commodity prices were obtained from the planting experience at Cienda, interviews with farmers, negotiations with tree nursery owners and a cost-benefit analysis evaluating rainforestation in comparison to coconut + abaca and Acacia mangium + abaca land uses in Leyte (AHRENS ET AL. 2004). Rentability data resulting from this module are seen only as a first approximation since detailed data, e.g. on transport costs, were not collected or updated. Results will not be discussed in the context of this study.

6.2 Model calibration

Having entered all soil, weather and plant data as well as management and others in the .stm model, additional adjustments were needed to adapt the simulation results to biomass data measured in the field. This is an iterative procedure based on test simulations. Key input parameters, which influence the respective outputs have to be identified and modifications effected must not deviate too much from measured physiological and architectural characteristics of the species. Calibration was carried out for subplots 6 and 7 including as few modifications as possible to the parametrisation data. Goodness of fit was evaluated by various statistical terms: Coefficient of determination (R2) expresses explained variance, the sum of squared deviations of observed data from their mean, as ratio of total variance, the sum of squared deviations between predictions and observations. However, this parameter is sensitive to trends but not orders of magnitude, so that constant over- or underestimation would not necessarily be indicated

87 Potential evapotranspiration for a standardised surface at PAGASA-LSU at 7m asl, P = 101.2kPa. N10°44' = 0.1873rad.

• 163 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte by low R2. This is detected by modelling efficiency (EF), the squared difference of explained and total variance divided by explained variance. For both R2 and EF, a value of one stands for best fit of prediction and measurement; EF can become a negative value. The root mean square error (RMSE) gives a percental error of the predictions from the mean of observations. Best fit would be associated with an RMSE value of 0. Formulae for all model fit statistics can be found under 2.8. For abaca, the reference unit to measured AGB was C_CanBiom, the sum of stem, leaves, storage organs and, theoretically, fruits. Modelled values were compared to the average of all surviving plants at the respective inventory, not the totality including dead plants. As measured soil parameters were available at sufficient detail, only adjustments affecting plant physiology were made during Figure 85: Model calibration for M. textilis on subplot 7, Cienda (open calibration: In a first area) instance, relative light intensity at which shading affects growth rates (T_RelLightMaxGr) was adjusted; this is the 'light switch' in WaNuLCAS, a threshold at which plants can still achieve maximum growth. Daily maximum dry matter production was also modified, as it limited biomass production below levels observed on site (fig. 85). For tree calibration, basically the same procedure was applied with light and maximum growth as adjusting screws. The model was calibrated for S. contorta, D. validus, A. heterophyllus and D. zibethinus (fig.86-89).

• 164 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Figure 86: Model calibration for S. contorta on Figure 87: Model calibration for D. validus on subplot 7, Cienda (open area) subplot 7, Cienda (open area)

Figure 88: Model calibration for D. zibethinus on Figure 89: Model calibration for A. heterophyllus subplots 6&7 (average, open area) on subplots 6&7 (average, open area)

The slow overall growth in abaca biomass was represented well – plant growth stagnated after a small initial increase – but the high mortality under field conditions was not sufficiently reflected as negative growth in the simulation. Still, coefficient of determination (R2) remained below values obtained for timber tree calibration. Root mean squared error (RMSE) was satisfactory for S. contorta, less for D. validus and M. textilis. For durian and jackfruit, good fit with respect to correlation coefficient and R2 does not reflect the constant underestimation revealed by modelling efficiency and RMSE. A calibration for Gmelina was not carried out, because the species had not been planted in Cienda in 2004 and not included in inventories. Anyway, heights of ten year old trees at Marcos were known from FBA and could be used to estimate plausibility of modelling. Generally, the numerous parameters in WaNuLCAS would allow for a very precise calibration; on the other hand, it cannot be the intention to deviate too much from measured values and create artefacts. In addition, these would be very site-specific and most likely not match data during validation. In small datasets, the used statistical terms are sensitive to outliers (LOAGUE & GREEN 1991). Especially for young plants, which are strongly susceptible to environmental factors (here: water stress), these environment- related outliers can occur even across means of many plants.

• 165 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

6.3 Validation

All validation was carried out on aboveground basis88, which can be displayed separately by the model. This was considered more accurate than total biomass as available aboveground field data were more precise than root biomass data. Four different data sets on biomass were used to evaluate the fit of simulations. • Growth of abaca under canopy on subplots 8 and 10 in Cienda, varying mainly in light conditions and to some extent in SOM and water. These subplots represent optimum growth. • Growth of the trees planted under canopy at Cienda in 2004 (also subplots 8 and 10). • Height of trees planted in 1997 at the Cienda demo site and • at the LSU rainforestation plot installed from 1992 onwards. Tree inventories by KOLB (carried out in 2003) served as a basis for the two latter datasets. For LSU, soil data by ASIO (1996) were available to initialise the model; for Cienda these had to be estimated.

6.3.1 Abaca and tree seedlings in Cienda Two different characteristic subplots at Cienda were tested for abaca and tree growth. These were the banana-dominated subplot 8 and subplot 10 under tree canopy, which had yielded the highest abaca growth rates. For both areas soil data were available from measurements presented in chapter 5. Profile data from PN1 and 2 (see chapter 3) were used to complement these. Both subplots were under canopy when the plantation was installed in 2004. Thus, the challenge was to simulate reduced light availability and competition with trees present before plot installation. Two options were tested: • To grow trees well before 2004, which would reach the density, height and lower canopy boundary observed on site before planting. Following this approach and using plant physiological values as calibrated for the open area PN3, competition for light was extreme. Almost independently of planting dates, one tree species grew vigorously and shaded out the second species. A minimal change in light demand (T_RelLightMaxGr) could make D. validus overgrow S. contorta, although the latter had been planted 6 years in advance. Secondly, the present canopy was composed of different species, which the model had not been calibrated for. N and P were only minor limitations to the growth of either species, which led to the second approach: • To plant all trees at one time, assuming the influence of the canopy to be restricted to reduced availability of sunlight and reduced transpiration leading to less water stress. In a simplified approach, this was done by setting Run_WaterLim? to 0.5, in other words to reduce water stress. Biomass was then calibrated setting a fixed shade tolerance (T_RelLightMaxGr) and initial amount of growth reserves (T_GroResInit, in stella) and successively varying species-specific potential growth T_MaxGro. This procedure can be justified as competition for N or P was not limiting at the given seedling stage and planting distances. Simulation run time was 1460 days from Jan 1st, 2003, onwards with all species planted after 16 months on May 1st, 2004. 20% slope and one forested plot uphill were assumed for both subplots, extension of all agroforestry zones (AF_ZoneTot) was 10m, equally shared among S. contorta, M. textilis, M. textilis and D. validus. Planting density for the trees was 200 per hectare and species. Measured and predicted data of aboveground tree biomass were compared on inventory days 544 (58 days after planting), 678, 834, 910,

88 T_BiomAG for trees and C_CanBiom as sum of leaf, stem, fruit and storage organ biomass, for crops.

• 166 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

1065 and 1218 and on days 570, 855, 915 and 1215 for abaca. Plant physiological parameters as calibrated for the open area were used first. Even without water and nutrient competition through big trees, plant growth stagnated after small initial growth under these settings, so that a new plant calibration was necessary for the plots under canopy; this was realised for subplot 8. The adjusted physiological settings are listed in the Annex and the goodness of fit statistics are shown in fig. Figure 90: Model calibration for M. textilis on 90-94. subplot 8, Cienda (area under canopy)

Figure 91: Model calibration for S. contorta on Figure 92: Model calibration for D. validus on subplot 8, Cienda (area under canopy). subplot 8, Cienda (area under canopy).

Figure 93: Model calibration for D. zibethinus on Figure 94: Model calibration for A. heterophyllus subplot 8, Cienda (area under canopy) on subplot 8, Cienda (area under canopy)

In most cases for subplots 8 and 10, maximum biomass, corresponding to the last inventory date, was much higher than values of the previous dates. This can be explained by the relatively high growth rates on these subplots once plants were well established; on

• 167 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

subplots 6 and 7, in contrast, biomass remained low even at later development stages. Despite very good R2, calibration for abaca and jackfruit were not satisfactory with respect to modelling efficiency and RMSE. For durian, EF was also not satisfactory, while R2 was very good and RMSE was still acceptable. Biomass of both timber species was underestimated when simulating the last but one inventory date, but still rendered very satisfactory modelling efficiency and 2 Figure 95: Model validation for M. textilis on R . Data obtained for subplot 8 were then subplot 10, Cienda (area under canopy). used to test the validity of the calibrated model to predict the dataset of subplot 10 (fig.95-99).

Figure 96: Model validation for S. contorta on Figure 98: Model validation for D. validus on subplot 10, Cienda (area under canopy). subplot 10, Cienda (area under canopy).

Figure 97: Model validation for D. zibethinus on Figure 99: Model validation for A. heterophyllus subplot 10, Cienda (area under canopy) on subplot 10, Cienda (area under canopy)

• 168 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

This time, abaca and D. validus were predicted well and so was durian, while the linear regression for S. contorta was constantly below the 45° line and thus biomass was underestimated. Jackfruit was massively overestimated when predicting the last measurement, but underestimated for the previous measurements. This points to a general tendency, which can be also observed for S. contorta (fig. 96): The last but one inventory was carried out without supervision and several, especially smaller, plants were overlooked; this impression was verified during the last inventory, when supposedly dead plants appeared again. So to speak, growth was not underpredicted, but measurement averages were inflated. Effects of the exceptionally dry period in 2005 were best visible as limiting factor for abaca growth. Nutrients were not limiting for plant growth, if a short but steep drop-down in P-supply is neglected. This sudden descent always occurred directly after planting and lasted only 1-2 days before returning to 100% or optimum supply. This seemed to be an artefact rather than a physiological effect.

6.3.2 Trees planted 1996 at Cienda site

This validation builds on data collected by KOLB (2003) from May - Jul 2001, five years after installation of the demo plot at Cienda had started. KOLB's soil data for pH KCl, C:N ratio and clay contents were almost identical to those obtained during this study, but Corg was higher (3.6 vs. 2.77%) in 2001 compared to 2004. CEC in 2001 doubled and BS tripled that found in 2004, while Al saturation was also remarkably higher in 2001. This is likely to be due to different sampling points rather than to changes in time. Simulating weather data, open area temperatures were used as the plot had been planted to annuals before the tree plantation was installed. A four-year loop was used for rainfall, temperature and soil evaporation. Run time was 2005 days, from Jan 1st, 1996 until Jun 30th, 2001, planting time being on day 90 in year 1996. Planting scheme was dense, 2x2m as initially planted (in reality this was intensified to 2x1m later, but at high mortality), resulting in 2500 trees per ha and species. Plant parameters as obtained from the calibration of PN3 were employed as the plantation was set up under open sky. Comparison of observations and predictions were based on tree height; when this was not indicated in the reference data, formulae provided by Kolb to convert diameter into height data were applied. These are d 3 H=1.3[ ] 0.87190.39239d lnH=0.723640.79444lnd−0.03803lnd2 for D. validus for S. contorta with H = height and d = diameter at breast height. For simulation of understorey trees, it was assumed, that these had been planted later. Thus, one species was given an advance of 2 years before the second species was planted. This corresponds to the extended planting period mentioned by Kolb and obtained during interviews with resource persons.

Table 25: Measured and predicted tree height for various species at the Cienda demo plot

Species (all Average height Height [m] planted 1996) [m] observed predicted Overstorey S. contorta 12.62 12.52 Understorey S. contorta 3.41 10.56 D. validus 3.78 6.17

• 169 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Height of S. contorta was predicted very well for the main stand (see tab. 25), which may be attributed to the use of original weather data. On the other hand height of understorey trees was subject to the difficulties mentioned before and thus over- or underestimated. Realistically, simulated height increase of seedlings planted under tutor trees was accelerated if compared to plants of the same species growing under full sunlight. Further, understorey plants stretched out towards the sunlight as can be observed in dense stands. On the other hand, denser spacing of even-aged stands led to less height increase, which is contrary to observations in plantations, unless other factors than light are more limiting to plant growth.

6.3.3 Trees planted at LSU in 1991-3 Reference measurements for validation were again obtained from Kolb's inventory in 2001 as given in table 27. Observations by QUIMIO ET AL. (1998) 5 years after planting are shown for comparison in table 26. Kolb based his inventories on a calculated original planting density of about 10,000 trees per hectare and subsequent mortality of 50%. For the part of the area under study, density was clearly lower in 2004, so that 2500 trees ha -1, shared among two species, were assumed for simulations. For weather data, a 4-year loop was run on soil temperatures starting under open sky. Simulation run time was 3500 days. Profile data from ASIO (1996) were evaluated to obtain an estimate of the original state of the soil. Horizonation was almost identical, if Asio's Bt1 and 2 are merged into one horizon and roughly 5cm deviation in topsoil thickness are considered due to different profile positions and soil erosion. Texture, pH and even organic matter contents mentioned by Asio (approximately 4% OM, corresponding to 2.3%C) were found unchanged89 compared to those found in 2005.

Table 26: Observed (QUIMIO ET AL. 1998) tree heights at LSU site 1998 Species Number of Average As for Cienda, measurements were observations height [m] differentiated into two groups, main observed stand and understorey. While QUIMIO ET AL. (1998) simulation of the main stand was Plot # 1 straight forward, conditions were simulated for the understorey by letting S. contorta 18 5.17 one set of trees grow from year 0 D. validus 19 5.05 onward and plant the second set in A. heterophyllus 3 5.33 year 5 (as sequential planting at LSU Plot # 2 took from 1991-1996; KOLB 2003). For S. contorta 9 3.75 this case, planting density was assumed to be 1650 trees per species D. zibethinus 8 3.48 and hectare. A spacing of 2m between Plot # 3 and 1.5m in lines was implied for S. contorta 8 5.26 undergrowth simulation. Especially D. validus 16 5.88 planting distances of D. validus as overstorey species were crucial for the D. zibethinus 5 5.20 development of S. contorta. A. heterophyllus 1 7.30

89 or recovered

• 170 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Under these circumstances and with no adjustments effectuated in the spreadsheet or stella file, growth of S. contorta was overestimated, while D. validus was predicted well for both strata (tab. 27). Modelling jackfruit also gave a realistic picture, but understorey durian was overpredicted. Table 27: Observed (KOLB 2003) and predicted tree heights at LSU site 2003

Species N of observations Average height Height [m] KOLB (2003) [m] observed predicted Overstorey S. contorta 17 12.1 18.12 D. validus 46 10.1 12.82 A. heterophyllus 8 11.9 10.46 Understorey S. contorta 31 5.5 11.27 D. validus 38 6.3 5.24 D. zibethinus 9 4.6 12.98

D. validus was more susceptible to water stress than S. contorta, but turned into the stronger competitor once P limitations were switched off. Shifting the planting date from dry into wet season did not significantly affect biomass development, while a dry spell during any later year could still have drastic effects. This has been documented by KOLB (2003) for the Cienda plot in 2000, when drought supposedly killed the majority of a pioneer species (Melia dubia) more than four years after planting. This needs to be considered since original weather data were only used for the last years of the model run and copied backwards in time. Causes of overestimation of S. contorta were assessed through different approaches. Running a sensitivity analysis across a wide range of Corg contents (more precisely, of Mc_CNRatInitMetab, i.e. different C:N ratios at stable N contents), did not lead to any response in aboveground biomass of S. contorta. Water, N and P limitations affected potential growth of D. validus, but for S. contorta limitations were probably caused by other factors or intrinsic to plant parametrisation. Gmelina as parametrised on ten year old trees in Marcos was tested for LSU site and gave plausible heights of 11.64m after 3500 days. The dry season from approximately day 2600 onwards delivered very realistic biomass loss due to leaf shedding. At the same time, N and P availability dropped suddenly as would be assumed as an effect of drought. In summary, validation showed, that biomass and height could principally be predicted for small as well as taller trees on the basis of the same calibration set-up. Calibrations could be used across different sites, but not transferred from plots under full sun to such under canopy. These experiences are true for simulations of even-aged plantations, while predicting growth of understorey trees was not possible to the desired accuracy under the given circumstances. Simultaneously growing S. contorta and D. validus complicated simulations of undergrowth species and did not always deliver plausible results. Kolb's documentation does not inform, whether understorey trees had been planted later than others, e.g. as replacement, or overtaken in growth and shaded by others. Apart from this, planting density at LSU could not be exactly reconstructed, because seedlings had been planted randomly with respect to species and position. Growth rates often oscillated between two extremes, which were caused by competition and could be inverted by small changes in planting density.

• 171 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

6.4 Modelling land use scenarios

The main goal of modelling was a comparison of various land uses under the aspect of carbon sequestration in context with potentials for CDM projects. In this context, the following hypotheses were formulated: ● Within 20 years after planting, an agroforestry system as planted in 2004 at Cienda will: a.) act as a sink for CO2 and b.) sequester more carbon than the traditional grasslands but less than a dense Gmelina pure stand. c.) This is also true for the soil carbon balance of the systems. ● In addition to its marketable yield, which is not evaluated here, the abaca component will – due to the rapid biomass build-up – contribute significantly to carbon sequestration in the system. With respect to phosphorus as most limiting nutrient, another hypothesis was tested with WaNuLCAS : ● Limiting P resources can be mobilised from the subsoil by trees and via litterfall made available to plants with a superficial rooting system.

6.4.1 Scenarios To test the hypotheses, the following scenarios were based on some common assumptions: Soil pedotransfer functions and nutrient contents for all scenarios were those of subplot 10, combined with profile data for PN1. This plot is representative for slope lands under extensive use. It was assumed that the plots had been cleared for the plantations, which were installed under full sunlight using the respective weather data in a 4-year loop from 2003-6. On the basis of the available calibration, scenarios focused on the time frame covered by the reforestation projects, roughly ten years. ● Scenario 0 (CO2 baseline): Grassland as a widespread land use in Leyte was chosen as the baseline scenario for carbon sequestration of improved systems. Grassland plots often contain scattered coconut trees, which were not taken into account for any scenario. It was assumed, that they would be maintained under agroforestry use, so that costs, benefits and carbon sequestration concerning this component would generally not differ between the systems. On the other hand, cultivation of cocos palms is usually a pre-requisite for grassland as it implies the necessity to brush the land before every coconut harvest, which prevents significant biomass to build up in the form of natural forest regeneration. For the dominating Imperata cylindrica, default values provided in the model were used. The reference scenario does not include burning, overestimating the average carbon balance of all grasslands in Leyte. ● Scenario 1a: S. contorta + D. zibethinus + M. textilis, 20% slope, no pruning: This basically represents the system installed in 2004 at Cienda site. Planting distances are 10x10m for timber trees, 5x5m for fruit trees and 2.5x2.5m for abaca (lay-out see chapter 2), resulting in 100 timber and 300 fruit trees and 1200 abaca plants per hectare. The lines were arranged as timber-abaca-abaca-fruit, zones being distributed equidistantly over 10m. No external inputs were imported and maintenance was reduced to a minimum as is the case for many upland plots in Leyte, which are located distant from their owners' homes. Weeds were included in the scenario (at 0.5 level); their contribution to the carbon balance can be distinguished and listed separately in WaNuLCAS. Anyway, undergrowth vegetation is shaded out in the model once maintenance respiration leads to a negative

• 172 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

CO2-balance for a plant. Besides weeding, pruning was also not included in this scenario. This was later contrasted with a pruning scenario to assess effects on abaca growth. Tree and abaca parameters were based on the calibrations for subplot 10, but slightly modified due to the presence of weeds and absence of canopy. In the case of S. contorta this meant, that T_RelLightMaxGr was increased from 0.15 (under canopy) to 0.35, and that T_MaxGro was slightly reduced from 0.0409 to 0.049. Inputs for plant growth are documented in the attached CD. For abaca, seed weight was increased as is the case for corms compared to tissue culture; in addition, planting was shifted to September (Julian day 244), the onset of the rainy season. Plant biomass and carbon as well as soil carbon were then evaluated against those of the baseline. ● Scenario 1b: S. contorta + D. zibethinus + M. textilis, 20% slope, pruning An option of automatic pruning as soon as the trees surpassed an LAI of 2.5 was chosen to increase available light for the understorey abaca. An additional effect of this management on SOM was expected as the pruned biomass was left on the field. Parameters evaluated were abaca biomass and SOM carbon as compared to 1a. ● Scenario 1c: S. contorta + D. zibethinus + M. textilis, no slope, no pruning This scenario was chosen to assess the influence of soil erosion on soil carbon stocks. Settings were the same as for 1a except slope, which was set from 20% to 0%. Soil carbon as influenced by erosion was compared to contents of scenarios 0 and 1a. In the same context, systems with different litter and mulch inputs were evaluated. ● Scenario 2: S. contorta + D. zibethinus. This scenario was chosen to quantify the contribution of abaca to the system in terms of carbon sequestration. In WaNuLCAS, biomass is listed separately for each component or planting zone of the system. However, a simple subtraction of the abaca biomass calculated in scenario 1a may not have accounted for interactions between crop and trees. For this reason, an extra scenario was run. Planting distances are the same as for the previous scenarios, as otherwise trees would strongly interfere during later years. Biomass and soil carbon were compared to those of scenario 1a. ● Scenario 3: Subsoil P acquired through tree roots Subsoil phosphorus contents were one important reason to choose PN1 for modelling. P is a limiting factor for plant growth in many volcanic soils in Leyte (ZIKELI 1998, ZÖFEL 2004), but was found in higher concentrations in the PN1 subsoil. It was of interest, if trees would be able to tap this pool and make P available to the system, including shallow-rooting abaca (CANNELL ET AL. 1996). Gmelina as a fast-growing competitive plant was used for this test. In a first simulation run, a.), Gmelina was grown under the conditions found at PN1, i.e. at 25mgP kg-1 in the subsoil. For the comparative run b.), subsoil P was reduced to 1mgP kg-1, a typical concentration found in other Leyte soils. c.) Doubled root length density in the deepest soil layer. All other parameters were left unchanged and both scenarios were evaluated with respect to P contents in tree tissues and litterfall. Apart, carbon balances of 3a were included in the systems comparison. Run time for scenarios 1-3 was 3500 days, less than one rotation of a timber-oriented system, but long enough for the fruit trees to enter in production. ● Scenario 4: Comparing long-term carbon balances An exploratory scenario over 7300 days was added to estimate long-term total carbon balances of a.) the grassland, b.) rainforestation and c.) the Gmelina system. Environmental conditions were left unchanged to those of scenarios 0, 1a and 3a.

• 173 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

6.5 Modelling outputs

6.5.1 Total carbon balance compared across land uses In a comparison of the grassland baseline (0), the rainforestation system and the dense and fast-growing Gmelina pure stand (3a), balances of total carbon during 7300 days were not as presumed for hypothesis 1b (fig.100): Although both captured more carbon than the Imperata-grassland, the expected ranking between tree-based systems was inverted. Gmelina as a pioneer species grew ahead of the mixed system during the first years, but was soon overtaken. This was due to the surprisingly fast growth of S. contorta, which overestimated observed growth at Cienda demo site (after 5 years) and LSU (after 10 years). At the same time, modelled height increased faster than cross-checking at LSU or Marcos would have suggested. Tendencies of the simulation are valid even though the absolute magnitude of S. contorta was not predicted correctly. Magnitudes of C stocks and rates will be discussed in depth under 6.6.2. Comparing predictions to literature data will show, that the simulation is not too far from real conditions.

Figure 100: Carbon stocks under different land uses during 7300 days of simulation

An intrinsic limitation in the plant parametrisation file was not the reason for the slightly decreasing tendency in Gmelina biomass; this is fully attributable to environmental conditions. Oscillations in Gmelina biomass can be explained by leaf shedding due to water stress: A factor governing the degree of leaf shedding as a reaction to water stress (T_LifallThreshWstr) is defined for each species. This value was estimated to be 0.5 for Gmelina, 0.7 for durian and 0.9 for S. contorta, which explains the more constant progression of the latter. In conclusion, while slow-growing durian was well-predicted, S. contorta is likely to be overestimated as had been the case for the LSU validation. The common sense impression of Gmelina as a fast-growing species is true with respect to tree height; looking at biomass, this is partly rectified through the low wood density and losses in leaf litter.

• 174 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Hypothesis 1a, too, could be discarded. A break-even for carbon stocks (intersection with the dashed line in fig. 100) to the state before planting was not reached within the simulated time of 7300 days. Under a more realistic growth of S. contorta the break-even point would be postponed even further. Still, carbon balances are calculated against a business-as-usual baseline, not against previous levels, so that for CDM purposes the scenario would be considered as a relative sink. In the scenario, carbon stocks departed from a relatively high level characteristic for recently cleared fallows, but not degraded lands. A more typical land use history would be fallow → clearing → annual crops → land rehabilitation through agroforestry. In that case, soil carbon stocks at planting time of the agroforestry system would have started from a lower level and a positive balance been reached earlier. The warming potential – expressed as CO2 set free from the plot and -2 inverse to carbon sequestration (both in gCO2 equivalent m ) – would then be lower. Total carbon balance during the first years was almost entirely determined by changes in soil carbon stocks as plants still did not contribute substantial amounts of C to the system. Absolute magnitudes of soil carbon depletion during the first years appear extreme, especially as no tillage was involved.

6.5.2 Biomass distribution between plants and agroforestry zones Fig. 101 gives an impression of the different components' contributions to total biomass. It shows the dominant influence of S. contorta, which is realistic if compared to older plots as in Marcos or (s. KOLB 2003) at LSU, even though the absolute magnitude was overestimated (see previous section). This did not change, if the plant settings from Cienda DS validation – which had given an excellent fit – were used instead of the slightly modified version used here. These validated settings differed from the ones used here with respect to maximum growth rate (which was actually even higher for the validated settings) and extinction light coefficient (which was reduced from 0.78 to 0.5). When the maximum growth rate was maintained and extinction light coefficient reduced, S. contorta did not grow beyond its planting stage. Durian reached a relatively stable level after approximately 5 years, which is not due to the transition from the vegetative into the generative phase, which per definitionem did not start before day 2920. It is rather owed to an intrinsic limit set in the spreadsheet for canopy height above stem (assuming an improved variety with limited growth). Oscillations from this stable tendency are due to leaf shedding as a consequence of drought. Phases of decreasing biomass coincide with water stress (limiting factors for durian will be discussed later). As measurements confirmed, abaca contributed a significant C input during the first years in relation to the tree component, but not if sequestration during the entire cycle is considered. Hypothesis 2 needs to be answered with some differentiation; it can be accepted, if the first two years are considered but is not valid over a longer period. As a consequence of light competition, abaca biomass in zone 2, next to S. contorta, constantly remained below that in zone 3, next to the less competitive D. zibethinus. Abaca in zone 2 was shaded out around day 900 (i.e. biomass decreased to values near zero), more than two years before abaca in zone 3. WaNuLCAS allows to display factors limiting plant growth: Departing from potential plant growth under optimum conditions, limiting factors water, light, N and P can be quantified as a percentage of amounts/concentrations required for maximum growth. Using this option, light already became a limiting factor before a dry spell hit the plot around day 900. Abaca growth in zone 3 was constrained to a lesser extent by lack of nitrogen from day 900-1700.

• 175 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

At this point, abaca in zone 2 had already been phased out. In context with competition, an advance in planting of 154 days for the trees has to be taken into account. As abaca grows fast under favourable conditions, simultaneous planting could have retarded the take-over of trees for several months or in some places protected trees from too much sunlight.

Figure 102: Contribution of pools to the agroforestry C balance

When the abaca component was omitted, but planting pattern and density for trees were maintained, the total carbon balance was lower than in the scenario that included abaca (fig. 102).

Figure 101: Contribution of M. textilis to the system C balance

• 176 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

This is not obvious as trees partly made use of the freed resources and could have (over)compensated for the missing element. A relatively higher biomass production of durian indeed compensated for some of the resources desoccupied (grey hairline in fig. 102). S. contorta biomass remained the same between scenarios during the first years, but was slightly lower in the system without abaca. This cannot be explained as a direct effect of abaca, but as competition with a fortified durian, which during the first years benefited from more light in the middle stratum and less belowground competition. Apart from initial carbon sequestration, increasing demand of abaca fibre could be a justification to patronise this element. The importance of abaca as of any annual crop in agroforestry lies in the timely return of investments, which otherwise confines acceptance of tree-based systems among farmers with limited capital. In scenario 1b, the effects of tree pruning on abaca biomass were assessed. WaNuLCAS offers an automatic pruning option, which allows to determine a percentage of biomass reduction as soon as the leaf area index of a tree surpasses a defined threshold. In the case of scenario 1b, this was set to LAI 2.5 which induced pruning. Fig. 103 illustrates the effect of such pruning on abaca biomass. For zone 2, pruning was minimal as WanFBA determined a maximum LAI of 2.66 for S. contorta.

Figure 103: Effect of tree pruning on abaca growth

• 177 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Consequently, abaca growth was equal with and without pruning. In zone 3, the reduced canopy of durian led to longer permanence of abaca in the system, even though at decreasing biomass levels. The same effect could be obtained in zone 2 without pruning, when the original Cienda demo validation settings were used for S. contorta. These include a lower extinction light coefficient (0.5 instead of 0.78), which allowed more light to reach the abaca canopy. Limitations for durian growth were also System with abaca explored under different management options. While pruning of timber branches is beneficial to obtain a straight bole and self-pruning species may be preferred, yields of fruit trees may decrease under heavy pruning, depending on planting distances and variety. On the other hand, pruning will increase yields and facilitate harvest, if carried out correctly. Generally, if space is limited, it will be ceded to the more profitable crop, which is in most cases the fruit tree. The scenario assumes improved varieties of durian with limited canopy volume, for which mild pruning is sufficient even at dense System without abaca planting. In practice, this corresponds to the grafted early-yielding varieties planted at Cienda. Fig. 104 shows the modelled response of durian potential growth to different species composition and pruning. As mentioned before, depressions in durian growth around days 900, 1600, 2400 and 3100 were related to water stress additionally causing shortages in P supply.

System with abaca; durian pruned

Figure 104: Durian biomass as limited by water, N, P and light as a consequence of planting system and management

• 178 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

The main difference between the systems with and without abaca consists in increased light availability for durian in the mixed system until approx. day 875. Competition for water occurs mainly during the initial 1.5 years, but also during single events as from days 2240-60. Effects of P competition are observable during >5 years. Looking at the pruned system (which includes abaca), reduced light use efficiency of durian after each pruning event is the most obvious reaction. This is always accompanied by limitations in P availability. A beneficial effect of pruning on durian water supply is best visible around days 2000, 2250 or 2700. This may be an effect of reduced transpiration due to cut leaves or reduced soil evaporation as a consequence of mulching.

6.5.3 Soil conservation The importance of soil carbon stocks for the global carbon budget has often been highlighted, e.g. by POWLSON (2005). On plot level, litter and SOM, initially containing more than 300MgC ha-1, represented by far the most important C pool (see fig. 105), which was drastically depleted after clearing the land and reached less than 120MgC ha-1 after 3500 days (fig. 106). Consequently, apart from the carbon sequestered in plant biomass, a decisive role of trees to be planted would be their contribution to litter and SOM pools.

Figure 105: Contribution of pools to the agroforestry C balance As hypothesised, soil carbon contents under Gmelina exceeded those under agroforestry. Regarding grasslands, the hypothesis was too optimistic in favour of the agroforestry system, which remained below the Imperata plot, although differences were decreasing during the simulated period. This could have been anticipated from sources as V. NOORDWIJK ET AL. (1997), who state, that soil carbon does not necessarily decrease through land use change from forest to well-managed (!) pasture and from results presented in chapter 4, where Corg contents under rainforestation were mostly below those under reference land uses. Due to the importance of soil carbon, different scenarios were tested in order to improve SOM under agroforestry. Apart from the previous scenarios with and without abaca and pruning, erosion control was included as an additional option. In an unrealistic but most effective theoretical approach, terrain was 'tilted' from 20 to 0% slope.

• 179 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Again, the influence of abaca as additional system component on the carbon balance was minimal. Concerning erosion, carbon stocks of agroforestry with abaca on flat land were lower than in the same system on slopes. No effect would have been a plausible modelling output as water infiltration was excellent. Lower stocks would only be realistic for footslopes. Higher contents on slope lands can only be explained by run-on from upper slope positions; this option, however, had been deselected in the model. The effect of pruning significantly improved the soil carbon balance. A trend line in fig. 100 would give very similar magnitudes for agroforestry and Gmelina, even at a slightly steeper slope for the first option. The stronger oscillation of the Gmelina curve represents the lower frequency of naturally induced litterfall compared to pruning, but possibly also the better decomposability of Gmelina litter. Similar tendencies for soil carbon have been reported by LASCO ET AL. (2005a): In a simulation over 100 years, SOM decreased by about the same amount at which forest biomass C was built up. Given the restrictions for S. contorta, the Cienda scenario, however, is more optimistic, marking a turning point from total C depletion to accumulation within the first 20 years.

6.5.4 Nutrient supply – acquisition of subsoil Phosphorus Tree growth was, as assumed, influenced by P in the C horizon, meaning that roots were able to access reserves from the subsoil. In fig. 106, average P in plant tissues and in litterfall is shown for Gmelina grown on PN1 with medium P contents (25mgP kg-1) and a virtual low-P (1mgP kg-1) subsoil under a 2x2m Gmelina plantation. After 3500 days, Gmelina biomass (dry weight) on the plot with low P in the subsoil reached hardly 80% of the biomass at PN1 profile; on a plant P basis, the ratio was even lower.

Figure 106: Phosphorus in Gmelina and in litterfall on sites with medium and low subsoil P contents

PN1 is still a moderate example compared to Marcos site with >300mgP kg-1 where effects might have been even more significant; on the other hand it is clear, that plants at Marcos were not able to make full use of subsoil P reserves: Clay contents are moderate at PN1

• 180 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte and rootability is better than at Marcos with up to 72% clay and stagnic properties. This is supported by the modelled result, that P leaching (data not shown) was considerable at PN1, especially if compared to the low P version. V. NOORDWIJK ET AL. (2004) modelled a similar situation with respect to safety net functions of roots and nitrogen. They found, that safety net efficiency, expressed as the fraction of N uptake to leached N, varied seasonally depending on plant demand. Looking at different P pools, the ratio between low and high P treatment after 3500 days was most distinct for plant P of the entire plant (T_Biom[P, SpX]; 75%), followed by P contents of leaves and twigs (T_LfTwigConcAct[P,SpX]; 94%) and consequently litterfall (T_LifallConcAct[P,Sp1]; 94%). It appears, that the tree, after covering its own needs, would still store some P in the stem and roots. Equal P contents in live leaves / twigs and in leaf litter indicate low relocation as may be the case for drought-induced leaf shedding. On site, this result was supported by elevated topsoil PBray contents in Marcos. Doubling root length density (red curve) led to significantly higher P uptake into the plant, while P in litter did not increase at the same ratio. P mobilisation by mycorrhizae was neglected in the model, which most likely led to underestimation of P supply in the plants. Especially Dipterocarpaceae are known for their association with ectomycorrhizae (YOUNG 1997). In the standard agroforestry scenario including S. contorta and D. zibethinus, the first species was limited through P during most days.

6.6 General discussion

6.6.1 Evaluation of modelling assumptions Generally, identifying key parameters and developing a feeling for their impact were crucial for the entire modelling process. Continuous work and experience with the model were needed for an efficient calibration as well as reference data and common sense for the evaluation of outputs. In the following sections, some assumptions taken for the different steps in modelling are evaluated. For calibration and validation as well as in the tested scenarios at the new Cienda plot, differences in soil between subplots did not play a major role for plant growth. This became evident, when the respective parameters were varied and had been expected for soil chemical parameters, which were not yet limiting for young plants. On the older plots at Cienda demo site and LSU, P and N deficiencies affected growth. With respect to water availability, better supply at the footslope subplots 6 and 7 had been expected due to lateral flow and capillary rise in the clayey profile. However, plants did not benefit from supposedly higher soil water contents, probably due to a larger portion of unavailable water in fine pores or due to their still shallow root systems. In this context, it would be of interest, how plants on the very well to excessively drained and clayey calcareous soils, e.g. at Punta, cope with drought. Erosion is an important factor on forest lands in Leyte, but this was not reflected in the soil carbon balance when slope was varied. This may be due to the high infiltration rates mentioned in chapter 3 and would then change for clayey soils like Punta or Marcos. The full set of options available in the model was not explored for erosion scenarios. Especially settings for regeneration of soil structure through soil fauna were kept static for the sake of simplicity. Contrasting to the real conditions at Cienda, a protecting canopy was not modelled for the scenarios, so that rain had its full impact on the ground. Among weather parameters, potential evapotranspiration is an influential factor, especially for plants like abaca, which is not well adapted to water stress. Soil temperature, in

• 181 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte contrast, was not expected to play a major role for plants. The effect of rainfall could be most clearly observed, in the model as well as on the field. Especially for Gmelina, dry spells led to a reduction of biomass through leaf-shedding, which was reflected in the model. Farmers confirmed, that water stress could lead to tree mortality even several years after planting. If weather data are generated by WaNuLCAS, attention should be paid to variability, e.g. El Niño patterns. During calibration and validation it is advisable to use measured real time data. This is especially true for sensitive and young plants. Among the multitude of plant-related input parameters, many of which are not measured but indirectly derived or estimated, those governing potential growth rates and light use efficiency were the most influential. The importance of light as a limiting factor became obvious, when other potential constraints such as P, N and water were excluded and T_RelLightMaxGr was varied. Below certain discrete threshold light levels, other factors did not have any influence on plant growth anymore as light was limiting. Beyond, biomass depended almost exclusively on plant-intrinsic parameters, as external constraints were not relevant. To obtain any growth under canopy, T_RelLightMaxGr was often set to values such as 0.15 or 0.2, which may appear very low. On the other hand it has been shown in section 5.1.8, that PAR as low as 20% of above-canopy levels was still sufficient for good plant growth on subplots 2-5. Key intrinsic factors governing tree growth were T_GroMax, the maximum growth rate, and T_GroResInit, the growth reserves stored in the young plant. T_GroMax could lead to a logistic curve, where at a certain critical point in a sensitivity analysis minimal changes had tremendous effects on tree biomass, overriding most other parameters. T_GroResInit rules initial development of the plant; when this parameter was chosen too low, biomass kept increasing steadily, while tree height stagnated and then suddenly increased up to tenfold (e.g. from 30cm to 3m) within one day. On the other hand, setting T_GroResInit too high, e.g. at the default level, eliminated the influence of most other factors. Both parameters cannot be measured on site, so that field experience would not accelerate the calibrating procedure. Correspondingly for crop growth, C_GroMax, the potential growth rate, seed weight (Cq_GSeed) and shading (Cq_RelLightMaxGr) were critical factors. Depending on seasons, litterfall can significantly reduce aboveground biomass, especially of young trees, when lignified parts still make up for a low share. This can also be concluded from litter inputs shown in fig. 107. Since for most of the planted trees these temporal dynamics were not exactly known, a periodicity of leaf-shedding was deselected, so that dropping of leaves depended solely on water-stress. To avoid litter-dependent fluctuations, growth could have been evaluated on the basis of tree height as was the case for Cienda demo and LSU, where biomass data were not available. For the small plants this approach was not chosen, because biomass can increase even at stagnating height – which was correctly reproduced by the model – and due to the aforementioned strong influence of T_GroResInit. Another impact factor with respect to plant growth was spacing in and between lines. This needs to be considered for calibration of taller trees, where interactions play a major role. For the case of Cienda 1-10 this was not yet relevant, but for the randomly planted LSU plot, distances were important and had to be estimated. Related difficulties arose in context with literature review, where values are often given as biomass per hectare, but planting schemes are not described in detail. Environmental conditions exercise their strongest influence on plants during the critical initial phase after planting, when small variations of a single parameter can lead to either exponential growth or mortality. Considering this, it was remarkable, that a calibration performed for seedlings during the first two years gave satisfactory results for validation even of grown-up trees, as was the case in the main stand of LSU and Cienda demo site.

• 182 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Still, further validation of big trees, including more species as well as individuals beyond 10 years of age, would be desirable to extend simulations. The modelling exercise presented here can be understood as a first step to depict multi-storey systems, but only includes a small part of plant-plant interactions in complex systems. An approach such as rainforestation, which aims at maximum diversity, will remain a challenge for modelling. For the given research question, mainly biomass and C balances, the modelling approach appears appropriate, especially when compared to the standard approaches used for CDM evaluation.

6.6.2 Magnitudes of stocks Carbon stocks and mean annual increments (MAI) for different land uses in the Philippines have been compiled, mainly based on data collected by LASCO (for quotations see tab.28) and co-workers in Leyte, on a site located 50km north of the research sites for this study. Following their ranking, Philippine old growth dipterocarp forests as natural climax vegetation can store >200MgC ha-1, followed by secondary forests, tree plantations, agroforestry, brushland and grassland, while mean annual increment is lowest in old growth forest and highest in tree plantations. As a rule of thumb, LASCO & PULHIN (2004) propose that agroforestry systems contain about half of the carbon of natural forests, but still substantially more than grasslands and pastures. On the other hand, these data are based on different approaches and stand age as well as spacing were not always determined. This becomes obvious from the broad range of Gmelina data or references by ALBRECHT & KANDJI (2003), who estimate C sequestration potentials of agroforestry systems in Southeast Asian humid ecosystems to be 12 - 228MgC ha-1 50a-1. Complementary to these references, a data base provided by the Intergovernmental Panel on Climatic Change was used to estimate LUC-induced carbon sequestration for a Philippine wet climate and high activity clay during a 20 years period90 (see table 28). Compared to the literature review presented, annual sequestration rates based on IPCC were more conservative. In context to what has been mentioned in previous sections, it is of importance that all MAI values cited do not include changes in soil carbon. As examples from South America, estimates of standing C stocks for 5-6 year old fallows by BROWN & LUGO (1990), UHL (1987) and TREXLER & HAUGEN (1991) as cited from SCHROEDER -1 (1994) range from 7-12 MgC ha , similar to the values by LASCO & SUSON (1999). In -1 -1 contrast, UHL & JORDAN (1984) found MAI of 10Mg biomass ha a for natural regeneration after forest disturbance at the Upper Rio Negro, a nutrient-poor blackwater river in Amazonia.

90 Settings for the IPCC tool: Philippines, tropical wet climate, high activity clay mineral (existing C stocks 44MgC ha-1). From moderately degraded grassland (management factor 0.97, land use factor 1, input factor 1, predicted C stock 42.7MgC ha-1) to native ecosystem / nominal management as best fit (management factor 1.23 for no-tillage, input factor 1, land use factor 1, predicted C stock 44MgC ha-1). Change 0.1MgC ha-1 a-1 over 20 years. For severely degraded grassland (predicted C stock 30.8MgC ha-1) the improvement would be 0.7MgC ha-1 a-1.

• 183 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

Table 28: Carbon stocks and mean annual increase in biomass carbon under different land use systems in the Philippines; literature review. All stocks and MAI refer to aboveground biomass.

Land use, location Pools Stock [MgC ha-1] Mean annual Reference increment [MgC ha-1 a-1]

Rain forest humid AGB > 20Mg dry matter YOUNG 1997 tropics (not C!)

Dipterocarp forest 80 AGB 126-241 LASCO ET AL. 2004 years

Secondary forest Leyte AGB plus 199.4 0.9 LASCO ET AL. 2002 necromass

Natural forest Leyte AGB 188 0.9 LASCO ET AL. 2002

Gmelina 6a, N. Ecija AGB 3.47-7.75 0.58-1.29 LASCO 2001

Gmelina AGB 32 7.9 LASCO 2001

Gmelina AGB 55.8 8.2 LASCO ET AL. 1999

Gmelina + Cacao, AGB 113.4 LASCO ET AL. 2001 Makiling

Mixed species AGB, not C 48.77 dry matter, LASCO ET AL. 2005a plantation not C

Tree plantation AGB 10.1 LASCO ET AL. 2002

Fallow, Cebu AGB 14.4 5.3 LASCO & SUSON 1999

Brushland AGB 29 4.3 LASCO ET AL. 2002

Saccharum 15.2 LASCO ET AL. 1999 spontaneum

Imperata grassland 8.9 LASCO ET AL. 1999

Grassland AGB 17.15 LASCO ET AL. 2005a

Undergrowth AGB 0.08-0.42 LASCO ET AL. 2005a Moderately degraded AGB From 42.7 to 44 0.1 IPCC tool grassland to natural within 20 years vegetation, 20a Severely degraded AGB From 30.8 to 44 0.7 IPCC tool grassland to natural within 20 years vegetation, 20a

In relation to orders of magnitude from the Philippine references, the biomass produced in scenario 1a appears more realistic than in relation to the field observations in Baybay: S. contorta biomass for subplot 10 was 0.01kg m-2 at the last inventory date (two years after planting), while the modelled value was 2.1kg at a corresponding time. This was surprising in so far as T_MaxGro as well as T_RelLightGroMax had been modified in a direction that would slow down growth, especially during the later stages, if compared to the initial calibration. Further increasing T_RelLightGroMax would immediately have led to zero growth. Under the given circumstances, biomass for S. contorta was modelled 8.65kg m-2 at a time of 5 years after planting. At the end of the simulation, S. contorta had reached approx. 18 and durian about 1.8kg m-2. This would be equivalent91 to -1 approximately 99MgC ha , half of the level given by LASCO ET AL. (2002) for secondary forest in Leyte, but achieved within only ten years and only 400 trees, 100 S. contorta and

91 assuming C contents to be 50% of plant dry matter

• 184 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

300 D. zibethinus, per ha. Even if a sigmoidal growth curve, i.e. slower growth during old- age, and better resource use of the plantation due to two storeys, are considered, the predicted values still appear too high. Dividing the carbon stocks by numbers of years results in a mean annual increase of 0.33MgC ha-1 a-1 for the S. contorta component under the plot settings of PN1. This is well within the range given by the literature data for plantations (10.1MgC ha-1 a-1) cited above. When the 30-fold higher literature MAI would be applied to the Cienda plot neglecting plant interactions, this would correspond to a planting density of 2x2 instead of 10x10m. From this perspective, modelled predictions may just slightly overestimate observations, but all in all be very realistic. Transforming S. contorta diameters from Kolb's study into biomass following the allometric equation presented in chapter 592 gave 10.19kg per tree, or 0.1019kg m-2 of dry matter for Cienda after 5 years and 6.60kg per tree, corresponding to 0.66kg m-2, for LSU after 10 years. Dividing predicted by observed values for S. contorta, this quotient decreases from 210 after 2 years through 85 after 5 years to 30 after 10 years, indicating that growth was most severely overestimated during the first years. When tree height of S. contorta is regarded instead of biomass, the simulation still clearly overestimated growth after 10 years compared to the Cienda demo or LSU plots and even more when compared to the new plot at Cienda. Modelling durian gave a similar picture: Measured 0.008kg m-2 after two years were -2 2 modelled as 0.44kg m . For LSU, KOLB (2003) measured 0.12m basal area per ha, shared between 10 trees. This can be transformed into 12cm diameter and then converted into biomass using the allometric equation from chapter 5. This results in 97.25kg dry matter per plant and, at a planting density of 300 trees per hectare, in 2.9kg dry matter per square metre at year ten after planting. This would roughly correspond to 14MgC ha-1. In summary, predictions for S. contorta strongly overestimated observed stocks at Cienda and LSU, with decreasing tendency during later years, while predictions compared to literature values, especially annual increase, are realistic. This may point to factors responsible for suboptimal growth at Cienda site, which the model settings did not account for. Examples could be pests and diseases (set zero), weeds (set 0.5 on a scale of 0-1) and plant physiological estimates. CEC and base saturation cannot be manipulated in the model; although these were not limiting if judging by profile data, intercostal chlorosis observed on L. domesticum plants on the upper part of the plot may point to Mg deficiencies (see discussion in chapter 4). LASCO ET AL. (2005a) mention in a case study on reforestation of a watershed in Luzon, that carbon sequestration rates are generally low under the harsh and sub-marginal conditions in grassland areas. Abaca, however, gave quite accurate values of 0.228kg m-2, while 0.249kg m-2 were calculated from field measurements after 21 months. -1 Gmelina biomass was below the 29Mg ha measured by MERCADO ET AL. (2006) after 2 years in Mindanao. Apart from different planting systems (hedgerows in Mindanao, approx. 2x3m in Marcos) an important difference between Mercado's plot and the ones in -1 Leyte were soil P contents: In Mindanao, 4-5mg kg PBray were found in the upper soil layers from 0-30cm. Under these elevated P conditions Gmelina biomass would correspond well with the increased root length density scenario shown in fig. 107. Generally, it is difficult to compare tree plantations, if planting densities and age are not known exactly. It is even harder to realistically estimate mixed stands or biomass estimates that have been calculated from allometric equations, another potential source of

92 B = 19.0824 D3.0651 for S. contorta, B = 18.9759 D3.4375 for D. zibethinus

• 185 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte error. Despite this, it can be stated, that biomass of S. contorta as predicted by the model overestimated biomass measured on the Baybay plots. This contrasts the general obsevation by CAREY ET AL. (2001), that old-growth forest biomass is oftenly underpredicted due to a multi-storey structure, which is not considered by the respective models. Predictions for D. zibethinus were much more realistic and such for M. textilis, which could be best verified in the field, were excellent. These differences in accuracy are due to the number of individuals used for calibration and validation as well as the range, over which measurements were available. Further, predicting growth of young plants depends strongly on exact coincidence of observation dates and weather, whose influence on biomass varies within short intervals. Soil carbon was the most important C pool in all systems, pointing to the relevance of -1 erosion control and soil cover. Based on the profile data, 122MgCorg ha were calculated for PN193, compared to roughly 300MgC ha-1 in the simulation. Both calculations refer to the same depth, not taking subsoil carbon below 1m into account. Average C values for fine earth of moist climates in tropical Asia given by BROWN ET AL. (1993) estimate 116MgC -1 ha . Soil below 1m depth can contain about 30% of soil carbon (SCHWENDENMANN (2002) for more deeply weathered Mesoamerican soils), which mostly belongs to the passive pool with turnover times of millennia. This portion is not affected by land use change. The first calculation does not include aboveground nor root litter, which are included in WaNuLCAS. Root biomass typically amounts to 20-50% of aboveground biomass and orders of magnitude of root litter can be roughly deducted from mortality of Gliricidia -1 hedgerow fine roots, about 1Mg ha , assuming annual renewal (YOUNG 1997). Both litter pools, however, are initially small in the scenarios which started from cleared land, so that all litter inputs necessarily come from the planted seedlings and weed. In reality, decomposing roots would provide an important pool of carbon and nutrients after clearing a fallowed plot. A literature review conducted by MURTY ET AL. (2002) showed, that conversion from forest to pasture does not necessarily imply reduction in soil C contents. v. NOORDWIJK ET AL. (1997) studied soil carbon stocks as affected by land use change in Sumatra. They compared data from a broad range of soils and concluded that conversion of forests into well- managed pastures does not necessarily reduce carbon levels in soils. GUO & GIFFORD (2002), in a meta study on soil carbon balances after land use change from 74 publications, concluded that even a change from native forest to pasture would result in 8% increase in soil carbon stocks. Installing a plantation after pasture would lead to a loss of 10% soil C. For such a scenario, it may be justified, that the IPCC methodology does not take belowground carbon into account for land use changes over a 20 year horizon as additionality is expected to be low. For the scenarios described above, the IPCC tool drew a too optimistic picture since soil carbon clearly dropped during the first years and did not fully recover to its original level within 20 years. From a theoretical point of view, loss of soil carbon, if caused by the plantation, should enter as leakage into calculations. In practice, it is not clear how far these fluxes can be evaluated during the verification of CDM projects.

6.6.3 Lessons learnt from modelling with WaNuLCAS Modelling agroforestry systems with WaNuLCAS was intended to assess the prospects of agroforestry under the aspect of a C balance in CDM projects. Conclusions on the modelling process itself and on implications for such systems will be discussed here. WaNuLCAS is based on Stella and can be easily explored by less experienced users but

93 From Corg contents and bulk density, considering stone contents

• 186 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte also allows experts to go into details and modify interactions and algorithms on a graphical and command line basis. The input parameters for basic soil function sheets (pedotransfer and P) are confined to the necessary minimum and were found straight forward to be measured. So are weather data, although soil temperature could not be obtained from a meteorological station and had to be derived through a regression. In the case of soil organic matter, the different carbon pools were determined following the TSBF fractionation method and microbial biomass was determined separately. Still, total soil carbon deviated from what was calculated on the basis of total Corg measurements. A direct input of total Corg fraction and proposed segregation into different pools could have provided a valuable hint for cross-checking data. Tree inputs, however, were not always easy to collect. While aboveground FBA analysis was found a very useful procedure, that gave reproducible results, root FBA proved to be impracticable specifically for clayey and dry soils. Fine roots can be easily cut and then overlooked in heavy soils; to avoid this, excavations sometimes followed archaeological rather than agricultural principles. Factors p and q usually disqualified root systems to be estimated by FBA because of fine roots branching off main roots. Setting minimum diameters on the other hand would have implied the necessity of wider excavations. Root systems in nutrient-poor or shallow soils can be extensive, which made sampling labour- intensive and, in dense plantations, destructive. Alternatively, extracting soil cores to destructively measure root length density required sufficient numbers of replicates to obtain reliable results (BÖHM 1978). These were difficult to collect in the proximity of the stem and from deeper soil horizons. Subsequently, labour-intensive procedures were needed to quantify root length density from the extracted cores. Especially for crop input parameters, some plant physiological parameters like root and harvest allocation, root water potential, hydraulic conductivity and light use efficiency, among others, were difficult to determine. This was complicated in the very specific case of abaca due to the phenological characteristics mentioned before. For trees, the tree survey questionnaire greatly helped to semi-quantitatively estimate inputs which could not be measured exactly. When parametrising systems, the most challenging issue was to simulate sequential planting since interactions in the relatively dense systems were crucial. On some plots, the system to be simulated was enrichment planting rather than agroforestry. It was not possible to simulate this at a reasonable accuracy, because the existing overstorey was diverse in species and age and could not be entered into the model for calibration purposes. Generally, it was not possible to simply apply the same plant settings under changing canopy. A mere replacement of soil functions was possible where a similar canopy existed (subplots 8 and 10). Another critical point for accuracy was validation. Own inventories of old stands would have gone beyond the capacities available for this study. The data collected by KOLB (2003) were helpful but additional datasets for other rainforestation sites would definitely have contributed to better representativity. Under the given circumstances, WaNuLCAS gave satisfactory results for the targeted purposes: ● The most relevant carbon pools could be identified for each of the tested land use systems ● Changes in magnitudes of C pools could be traced back to planting and management options, new alternatives could be tested and their effects on the system be separated (BAYALA ET AL. 2004) and evaluated. Availability of nutrients as depending on root length density and other factors were plausibly represented and nutrient cycles followed from the soil through the plant into the litter component. ● A systems comparison with respect to carbon stocks was possible and could give

• 187 • 6 Modelling growth and carbon sequestration of agroforestry systems in Leyte

valuable indications for land use optimisation. Statistical evaluation of the calibrated and validated model showed that there was still potential for improvement. Coefficient of determination was good in most cases, while other statistical parameters were not always as desired. Among these, RMSE is widely agreed on for modelling purposes, while modelling efficiency is less frequently used. Mean absolute error, forecasting coefficient and others have also been suggested (s. YANG ET AL. 2000), but were not used here. Determining exact magnitudes could not be expected, because most of the planted species had not been used for modelling before, so that few references were available. In addition, these species are, if planted, mostly grown by smallholders in mixed systems, so that the few existing reference data are often not well- defined with respect to growing conditions. At any case, the importance of soil carbon became very clear as did the relatively modest contribution of abaca over the entire rotation. The greater potential of S. contorta over a longer term as compared to Gmelina was also shown. The presented results were achieved after time-consuming measurements, parametrisation and calibration processes of the model. As long as the model is not widely used and unless tested species data bases are available for different sites, WaNuLCAS cannot be employed for rapid exploratory assessments, like the IPCC tool. Anyway, considering the numerous options and flexibility the model provides, superficial use would waste the full capacities provided. More specialised models for single compartments or processes exist, like CO2Fix for carbon sequestration, CABALA (BATTAGLIA 2004) for forestry or RothC for soil carbon dynamics. The strength of WaNuLCAS is its ability to simulate interactions of different plants on the same plot. Additional expertise in the form of modules like CENTURY for the SOM compartment or WOFOST as a pre-processing tool has been integrated into the model and WaNuLCAS is being continuously refined and expanded. As any model, WaNuLCAS will become more and more relevant and applicable with an increasing user community.

• 188 • 7 Conclusions and outlook

7 Conclusions and outlook The aim of this study was to illuminate the rainforestation system in comparison to other land uses from different angles, mainly its original rehabilitation approach, biomass production and carbon sequestration aspects. Parameter selection followed the rainforestation leitmotiv of land rehabilitation through accelerated succession, leading to a forest-like system of high diversity and enhanced resilience against environmental extremes. Consequently, the focus was set on parameters determining and concerning organic matter turnover and such expected to change within a relatively short time under rainforestation, i.e. 10-15 years. Methods had to be applicable under basic and low-budget conditions and comparable to those applied in other studies, an approach inspired by TSBF. At the selected sites, differences were substantial. Soil analyses at the selected plots gave valuable insights into typical time-, rock- and relief-related processes. Major groups of soils – obviously calcareous and volcanic, but also slope, footslope and plateau positions as well as autochthonous and allochthonous types were distinguished and implications of these groups on nutrient and water balance identified. Assets and constraints of soils for agriculture were evaluated. Data from these sites also provided orders of magnitude as orientation to judge land use-specific changes. An important finding in this context was the elevated level of P in two subsoils, since P is usually the most limiting element in volcanic soils in Leyte. The ability of trees to bring nutrients from the subsoil to litter and humic horizons under the given conditions was later tested in a modelling scenario. Environmental changes as supposedly caused by the rainforestation systems were assessed through a paired-plot approach across 7 existing rainforestation sites, which fulfilled the criteria for a paired plot approach, mainly minimal area and an adjacent reference plot. As in many tropical forest ecosystems, succession goes along with acidification and net-export of nutrients from the soil. Part of this is owed to leaching caused by torrential rains once the oftenly excellent soil water holding capacity is over its limits (leaching of N and P has been indicated by the modelling approach). Loss of bases is intensified by the decreasing pH. In addition, nutrient uptake into plants can play an important role in soil depletion of certain elements (NYKVIST 1997), so that concentrations especially of macronutrients in soils can decrease over time under reforestation. This may not meet expectations from such systems, if restoration of soil fertility is evaluated merely from a productivity perspective. Gross productivity rates of ecosystems usually decrease as succession proceeds (ODUM 1969). However, nutrients are not completely exported from the ecosystem, but rather stored in the plant biomass as in a bank account. They can be set free, when the rotation is finished or, in natural forests, when a gap opens. This corresponds to the function of fallows under numerous nomadic indigenous systems (examples in POSEY & BALÉE 1989). Looking at ecological functions of soils, tree canopy mitigates microclimatic extremes and, through addition of litter, increases organic matter contents and improves soil structure. As an example, the relatively open banana-dominated subplot 8 showed clear signs of erosion such as a thinner Ah horizon and higher topsoil bulk density than the surrounding subplots under tree canopy. Reduced erodibility is crucial for the steep slopes on the Cordillera Central in Leyte94. From this point of view, the role of trees as providers of environmental services lies in their provision of shade, formation of biomass, contribution to SOM and reduction of erosion. Under the mentioned aspects of productivity and environmental services, systems

94 While landslides occur frequently even in old growth forests and the load of trees superimposed on the soil may in some cases increase the risk of landslides.

• 189 • 7 Conclusions and outlook targeting the latter may be less profitable in the short run, but more sustainable. Late- successional ecosystems tend to stability, while pioneer stages aim at rapid biomass build-up (ODUM 1969); the latter also applies for agricultural systems, which are arrested in a pioneer stage (EWEL 1999). Where financial liquidity is critical, sustainability can be a side effect but will not be a priority95. The aim of any successful agroforestry system must be to integrate both ends. Short-term rentability can be achieved through the integration of early-yielding species, which motivate farmers to maintain the plot during the first years. Continuous revenues from carbon sequestration can be another option to create income at an early stage of the system. As commodities, CER should not be considered as subsidies, but income from environmental services. In contrast, provision of all necessary inputs plus 'livelihood' subsidies by a project may hamper motivation of project 'beneficiaries'. Regarding the first option, integrating early-yielding species has not been emphasised on the early rainforestation plots, but is now more accepted. Diversification in early-yielding species does not necessarily restrict short-term profitability, especially under low-capital smallholder conditions. In this context, an imbalanced focus on one particular cash crop fostered by project activities or even real demand could be fatal in case of pest or disease outbreaks96 or a collapsing market. Regarding income generation through CER, the modelling approach showed, that fast- growing pioneers quickly add large amounts of C to the system, but at a later stage stagnate in biomass production, unless late-successional species have been interplanted to take the lead in biomass production. Recalcitrant litter such as provided by many dipterocarp species would be more useful for SOM accumulation than easily decomposable Gmelina leaves. Considering preference of native species, pioneer trees such as S. palosapis or fast-growing species like A. odoratissimus could be an equivalent substitute fulfilling similar ecological functions as Gmelina. Constructing ecosystem functions with a minimum number of plant species (LANGI in GEROLD ET AL. 2001), surely falls short of imitating an intact natural ecosystem. Too many factors and interrelationships are not known, so that apart from obvious measurable functions such as biomass or nutrient contents, ecosystems can hardly be engineered. Instead of trying to determine and reduce 'functional redundancy', pursuing maximum diversity appears more appropriate. On the other hand, the argument of native species should not be taken too far since many of the most popular fruits cultivated in the Philippines are not native. Unless pure stands are installed, a few exotic species should be accepted within the system. The integration of annuals or early yielding species can play a role for carbon balances during the first years, but then loses importance. Under the aspect of soil conservation, any crop or soil cover is of advantage in order to reduce erosion and to supply biomass for humification. Improved microclimatic conditions under cover crops may cause more intensive litter decomposition, humification and formation of stable aggregates, but not necessarily mineralisation. When CDM projects are planned, the influential role of soils has to be considered. Small changes in soil C balances can have large impact on the total C stocks (POWLSON 2005). Apart from the soil-conserving factors mentioned before, design and management such as planting on contour lines and mulching can be other promising ways of soil conservation facilitated by trees (MERCADO 2007, PhD thesis, unpublished). The effect of mulching on

95 Philosophies based on sustainability may thus be considered a luxury phenomenon as they require investments that do not yield immediate returns. Not coincidentally, the fundaments of sustainability in forestry were laid by a government official (Oberberghauptmann V. CARLOWITZ 1713) and extended to a broader concept by the Brundtlandt commission in 1987, and not by smallholders. 96 Such as Sigatoka on abaca plantations in Southern Leyte.

• 190 • 7 Conclusions and outlook

SOM was well reproduced with WaNuLCAS, but an effect on erosion could not be deducted from the simulations. In this context, the relative accuracy of the simulations must be addressed again. WaNuLCAS worked well as a scientific model (PASSIOURA 1996) in order to explore hypotheses and tendencies, which were then cross-checked with field observations: For example, elevated P contents in the topsoil over a subsoil with relatively high P; the topsoil P was supposedly accumulated at the surface through tree leaf litter. For its use as an engineering model, i.e. to quantitatively predict growth or carbon pools, interpretation needs to be exercised with caution. Relative biomass magnitudes of the various agroforestry components were realistic when compared to measurements on site, but absolute dimensions would require more inventory data. Values fit well into literature data, but overestimated the measured dipterocarp growth. In practice, the growth experiment at Cienda site showed that, apart from planting time, good planting material is critical for the successful installation of any system. The most sensitive abaca was used as a bioindicator for site conditions. In this context, reserves are important for the plants to recover after stress phases. The issue of tissue culture vs. corms can probably be reduced to reserves, which are greater in corms. Key factors for abaca survival identified through logistic regression were mostly water-related while nutrient deficiency did not play a major role. If plantlets are of sufficient size, transplanted after dry season and management is carried out properly, even plantations of M. textilis can prosper under full sun as was observed at LSU in 2005. Leaf pruning after transplanting may be helpful to avoid water stress as long as roots are not yet fully developed. For growth, different factors were critical. Unlike for survival, the most influential factor according to a multiple regression across subplots was N supply. In the modelling approach, N supply was the second ranking constraint after light (in a scenario under canopy). A combination of abaca with leguminous creepers like Pueraria phaseoloides (kudzú), Desmodium ovalifolium, Canavalia ensiformis or Arachis pintoi can reduce evaporation and at the same time improve N supply. The first species, however, is competitive and can only be sown, if proper maintenance is ensured. Losses of plants in the upper centre part of Cienda plot were mainly attributable to kudzú, which suffocated even small trees.

After a long series of failures, reforestation activities in the Philippines became more successful in the 1990s, but often donor projects had a high and probably unsustainable economic cost (CHOKKALINGAM ET AL. 2006). This might have been a reason, why the rainforestation approach in Leyte has not left the stage of demonstration plots. Once the solvent donors retreated, ownership among farmers was missing. Under these conditions of lacking incentives and revenues from the plot, which were also observed at Cienda, labour may be a major constraint to the adoption of complex agroforestry systems (CRASWELL 1997). FRANZEL ET AL. (2004) assessed key factors for successfully implementing agroforestry on a wider scale. They found that for all case studies, one of them conducted in the Philippines, marketing, among others, was not sufficiently addressed by projects. POLINAR (2004) comes to the same conclusion, poor marketing outlets, adding delayed return of investments to the shortcomings of present initiatives. Thus, additional sources of income for rainforestation farms are urgently needed, if implementation shall be successful and last longer than a project cycle. CDM projects are expected to provide such additional income from the beginning and at a justifiable effort. However, this still needs to be proved since no CDM A/R project has been implemented in the Philippines by now. Principally, agroforestry projects in the Philippines are eligible under the umbrella of

• 191 • 7 Conclusions and outlook afforestation or reforestation. Special procedures have been introduced for small-scale projects <2180MgC addressing low-income individuals or communities. In the light of the upcoming dramatic IPCC assessment report 2007, KANNINEN'S (2004) argument that the international community cannot afford not to make use of agroforestry becomes ever more valid. LUC measures can globally sequester an annual 1 to 2 GtC, if all available options are made use of. This would include agroforestry, although such systems grow comparably slower than e.g. tree plantations. LEHMANN & GAUNT (2004) state that agroforestry due to its share of annuals is not an appropriate measure to increase long-term carbon pools in the soil unless special attention is paid to carbon conservation. This could be achieved by incorporation of charred lignified biomass, which is not extracted for use. The comparative advantage of agroforestry in turn is the potential of earlier income generation and thus accessibility and acceptance by small farmers (PENA 2004). LASCO ET AL. (2004) have estimated potential agroforestry land areas on the basis of existing degraded uplands (grassland and brushland areas) to cover more than 3.5 million ha in the Philippines. Adding other farmland in need of stabilisation or soil conservation, 5.7 million ha could potentially be planted to agroforestry systems. For CDM projects, these figures might need to be corrected downward depending on the national forest definition: If a low threshold level is defined as minimum canopy coverage for a forest, then many degraded grasslands under coconut would be considered forests and could not be reforested anymore. These extensive areas would consequently not be eligible for CDM A/R projects. Once eligibility is given, bureaucratic procedures such as the formulation of a Project Design Document (PDD)97 will probably require expert assistance, e.g. by an NGO. Apart from CDM eligibility and bureaucracy, land tenure can be a limiting factor for implementation of agroforestry plots (POLINAR 2004). Uncertain property titulation can keep people from investing in trees. A large portion of forest land is not private and among alienable and disposable (A&D) lands, claims are not always indisputable. Further, harvesting timber requires the approval of the Dept. for Environment and Natural Resources (DENR). The administrative act can discourage farmers, who do not dispose of resources for travelling and feel uncomfortable when dealing with authorities. Another factor discouraging any initiative is harvesting on other peoples' land, which is common and often informally accepted in Leyte. Finally, small plot size (POLINAR 2004) and low income lead to allocation of areas to the most profitable land use under a short-term perspective. Agroforestry is only implemented, if on a farmer's own initiative at all, in home gardens or as abaca underplanted in state-owned natural forests. An economic evaluation of the scenarios was not undertaken since more detailed and precise data would have been necessary. Still, some general considerations can be discussed here. In comparison to an agroforestry system not participating in CDM, additional costs consist in formulation of the PDD including follow-up, and in certification. Both items are difficult to estimate since experiences do not yet exist in the Philippines; facilitation of the process could be an input provided by local NGOs or development agencies. Investment into planting material and labour force cannot be expected from low- income groups as long as they are not well-organised or advised by professionals. Mostly, willingness to work depends on so-called livelihood support, meaning that farmers are paid, directly or in form of food, for work on their own land. Apart from monetary constraints, production or purchase and transport of planting material are obstacles for the successful implementation of agroforestry systems. Table 29 resumes the main advantages and disadvantages of the rainforestation

97 As published in Annex 1 of the Report of the First Meeting of the Afforestation and Reforestation Working Group, July 2004, Bonn.

• 192 • 7 Conclusions and outlook approach as detailed above, subdivided into internal and external factors. Arguments specifically relevant in context with CDM projects are printed in italics.

Table 29: SWOT analysis of rainforestation with respect to C sequestration in CDM projects

Strengths Weaknesses Based on native plants, adapted to local conditions Investment required Simple technology, easy to replicate Knowledge required Soil conservation Difficulties to supply seedlings No additional inputs required for CDM Maintenance costs Additional products, risk minimisation Late yields, if additional elements are not introduced Uncertainty regarding procedures

Opportunities Threats Land rehabilitation, sustainable land use Availability of areas (tenure, distance) Conservation of species and genetic diversity (within Demotivation through projects, lacking ownership species) Abaca boom and encroachment into natural forests Additional income generation Bureaucratic efforts for accreditation Better acceptance and dissemination through added Instability of CO2 prices value

The idea to minimise transaction costs and accurately meet demands by means of market-based mechanisms has been tested before the ratification of the Kyoto protocol. The World Bank's Prototype Carbon Fund, the Singapur ASEAN Carbon Fund and other voluntary initiatives of oil or travel companies are examples, which showed that emission trade technically works and attracts interest among stakeholders (s. DE CONNINCK & V.D. LINDEN 2003). A more principal question regarding CDM projects in developing countries is, if they lead to more sustainability. Industrialised countries have been responsible for 85% of anthropogenic atmospheric CO2 and accused to act ruthlessly and aggressively in not doing their ecological homework (K. Töpfer, former director of UNEP98). According to the definition by BASS ET AL. (2000), leakage occurs, when a project's activities and outputs create incentives to increase GHG emissions from processes taking place elsewhere. From a holistic standpoint, this is exactly what happens, when reforestation projects deflect attention from the roots of the problem and protract patterns of mass consumerism and myths of unlimited growth in industrialised countries. In public debate, however, the reasoning that CDM would be counter-productive in the long run as it does not tackle problems in Annex I countries, was outweighed by the cost-effectiveness argument. Agroforestry can only be one part of a global mitigation strategy (ALBRECHT & KANDJI 2003, CHAMBERS ET AL. 2001), which needs to concentrate on reduction of fossil fuel burning. In turn, following the principle of risk minimisation in agroforestry, the reliance on carbon projects should not be overemphasised. As all commodities, prices of emission rights can drop drastically, as happened in April 2006 in Germany99. Lobbyism of polluters and exaggerated estimates of certificate demand has led to issuance of too many certificates and collapse of the market (German Advisory Council on the Environment 2006100). In summary, CDM appears a promising option to generate income from agroforestry, if paperwork for project application can be limited to a reasonable extent. On the other hand,

98 in www.tagesschau.de, download 25/09/06 99 tageszeitung, 29/04/06 100 National Implementation of the EU Emissions Trading Scheme: Market-based climate change mitigation or the continuation of energy subsidies by other means? www.umweltrat.de

• 193 • 7 Conclusions and outlook

CO2 certificates should be seen as only one component of a farmer's portfolio diversified as much as possible. Future research Although inventory data from the 'mother of rainforestation' at LSU were available, these did not cover biomass data of more than 10 years (from 1991 until Kolb's inventory 2001). Extrapolations beyond 10 years of age remained speculation. Meanwhile, a range of 16 years could be covered, if an inventory would be carried out again. This would make simulations more reliable, especially with respect to the growth of native timber trees like S. contorta and D. validus. Other interesting species like the pioneer Melia dubia or high value timber like molave (Vitex parviflora), Hopea spp. and others could be subject of future scenarios. Erosion as an important constraint for upland soils in Leyte could be assessed to more extent using WaNuLCAS. Erosion has been modelled with good results by SAMANIEGO (2006) for soils in Thailand. The added value of modelling would mainly consist in the partial replacement of labour-intensive measurements such as aggregate stability and erosion. Several management options could be assessed within short time and the most promising ones tested in the field. Another important field for more detailed research is related to socio-economic factors, as profitability, availability of labour and acceptance are still critical factors for dissemination of the approach and continuance of the plots. An economic evaluation was not included in this study as it would have required more detailed field data. Basic data on rentability of rainforestation systems exist (AHRENS ET AL. 2004, DIRKSMEYER 1998) but need to be updated and complemented. On the present basis an evaluation of demand and market saturation, transport costs and other parameters could not be estimated accurately. Increased opportunity costs due to necessary owner's presence on site at harvesting time are also difficult to estimate. This presence would be required to protect harvestible products against theft rather than for the harvest itself. Furthermore, costs of certification for CDM projects are still difficult to determine as only few practical experiences exist for reforestation CDM projects (none in the Philippines). As mentioned before, prices for carbon certificates have been strongly fluctuating. In the EU, distribution of certificates has by far exceeded the actual demand, so that prices decayed drastically. Integration of further socio-economic parameters into WaNuLCAS is intended (V. NOORDWIJK 2004) and cooperation with the model developers could be part of future research. C in durable wood/fibre products: GPG-LULUCF provides accounting for harvested wood products from A/R areas, but the evaluation methodologies are still under development. In this case study, abaca fibres could play an important role for the carbon balance of a CDM project. The fibres are not woody, but their durability easily exceeds that of paper pulp given with 1.8 to 2 years (GPG-LULUCF Annex 3A.1), especially, when used as component of mould board in a matrix of synthetic resin. On the other hand, excessive export of biomass from the plot could lead to nutrient imbalances, which would have to be addressed. IPCC (2003) stresses the knowledge gap of downscaling from landscape-data to plot- level, permanence, saturation in fractions/compartments, stability (under changed climate) and increasing net biomass productivity. This study is a step towards generation of such plot-related data during the initial growth phase, but more data covering later development stages are still required.

• 194 • 8 Abstract

8 Abstract This study aimed at investigating rainforestation systems in Leyte, Philippines, under different aspects:

● Characterisation of relevant soils in Leyte with respect to physical, chemical and biological parameters relevant for tree growth, ● possible contributions of rainforestation to restoring soil fertility, ● performance of a recently planted rainforestation system under different microclimatic and soil conditions, ● potential of the rainforestation approach for projects under the umbrella of the Clean Development Mechanism (CDM).

In relation to growth conditions and as a basis for the assessment of rainforestation impact on soil characteristics, a thorough soil survey was carried out on 10 profiles across 8 rainforestation plots. Soils can be grouped by parent material into a volcanic and a calcareous category. The latter were formed on coralline limestone and are thus high in pH and available Ca2+ and Mg2+. Contents of organic matter are high while concentrations of plant available PBray are low. Soils of the volcanic group are characterised by low pH and concentrations of basic cations as well as extremely low PBray contents. Organic matter levels are below those of the calcareous soils but still moderate. Volcanic soils could further be subdivided into a colluvial class of yellowish colour and moderately low bases and soils formed in situ. The latter are strongly weathered, reddish in colour and lower in pH and bases than the yellow group. Mixtures occur, where colluvial horizons have been superimposed on older soils. In any of the analysed soils, N would not pose a limiting factor for tree growth. Pore volume and water infiltration were propitious for all sites, which is relevant in the context of erosion. However, lateral water flow on the clay horizons of Acrisols, Luvisols or Cambisols can lead to temporarily anaerobic conditions. For calcareous soils, drought and reduced rootability due to clayey subsoil posed the most relevant constraints.

The frequently claimed role of rainforestation in the rehabilitation of degraded soils was assessed in a paired plot approach. Chemical and biological soil parameters under 10 year old rainforestation were contrasted with such under adjacent fallow or Gmelina sp. plots. The most clear tendencies across all seven sampled sites were lower available Mg2+ and pH under rainforestation use. Other differences were less distinct; even when there were trends in the majority of sites, differences were often not statistically significant at the chosen level. Generally, a depletion of soil reserves e.g. in basic cations can be explained by uptake into the plants. A feed-back of these elements to the topsoil via leaf litter, however, could be observed only for available P at Marcos site: On the P-rich subsoil, concentrations in the topsoil (3.10mg/kg) were also elevated as compared to the underlying AB/Bt horizon (0.47mg/kg) and topsoils at other sites (0.7 to 1.98mg/kg)101. In conclusion, plant uptake of single elements can reach orders of magnitude, which can even lead to a clear reduction of soil stocks (e.g. significantly for Mg2+ in 6 of 10 cases). At the same time, generally lower pH under rainforestation (7 of 10 cases, four of them significant) may have contributed to elevated losses, especially of basic cations. A general improvement of the sampled soils in terms of chemical or biological characteristics through rainforestation could not be observed. In this sense rainforestation serves as a

101 In comparison of land uses per site, burning played a role for elevated topsoil P (under Marcos and Patag grassland).

• 195 • 8 Abstract bank for nutrients rather than immediate soil improver. Capital in form of nutrients can be withdrawn totally (clear cut) or in rates (by pruning and mulching).

To evaluate the performance of different promising tree species, a mixed-species tree system was installed on a 1ha plot. Six timber and four fruit species were planted, most of them native to the region. The concept of rainforestation, commonly understood as high- density closed canopy system was modified in so far as the usually dense tree spacing was changed into a 5x5m grid, interplanted with Musa textilis (abaca). The plot varied strongly on a small scale due to heterogeneous canopy closure and relief. Methodologically, the entire area was divided into 10 subplots in representative positions to be sampled. Soil physical and chemical properties, microbial activity, PAR and root density were determined and correlated to plant survival and growth at consecutive inventories. Due to the multitude of measured parameters a principal component analysis was employed first to identify groups of similar parameters. For Musa textilis, the most sensitive species, which was used as an indicator, logistic regressions were calculated to determine the influence of all relevant parameters on survival rates. Subplot characteristics had a strong effect on abaca performance, the most important predictors for survival being organic matter contents, parameters related to biological activity and leaf litter production, which resembled canopy closure and thus indirectly light intensity, but also reflect soil moisture. To assess growth, multiple regressions were formulated. This was done for biomass at five inventories and growth between these dates. Corg and NLOM were the most relevant variables determining the regressions used for biomass and growth of abaca. Survival rates of tree species varied between 48 and 98%. Apparently, size and condition of the seedling at the time of transplanting played the most important role for performance.

Assessing the potential of rainforestation for CDM measures, the following observations were made: Smallholder agroforestry projects are generally eligible under the Clean Development Mechanism in the rubric of Afforestation/Reforestation102. In the context of this study, amounts of sequestered CO2 during 10 and 20 years, respectively, were estimated under different management options using the WaNuLCAS model. Despite all given uncertainty associated with modelling, one very obvious finding was the dominant role of soil carbon for the plot balance: Appropriate soil management, especially during land preparation (e.g. clearing vs. enrichment planting) is of paramount importance. The carbon balance would turn out even more unfavourable in case of burning, which was not part of the modelling exercise. Erosion, on the other hand, did not affect carbon stocks significantly in an additional scenario. This is in accordance with the low erodibility found in the soil surveys, but may differ on a small scale, depending on vegetation cover, soil management or compaction, among others. Looking at the modelled contribution of various tree species to the carbon balance, Musa textilis had a significant influence only during the very first years; later on, the principal share of carbon was bound in the tree component. Here, Gmelina arborea built up biomass more quickly than a rainforestation plot composed of Shorea contorta and Durio zibethinus, but was then overtaken. In absolute quantities of CO2 sequestration, magnitudes matched inventory and modelled data given in various literature sources for Leyte and the Philippines. Relative to inventory data by KOLB (2003) from two of the existing rainforestation sites, modelled values overestimated growth. This may have been caused by unfavourable weather conditions

102 under certain country-specific conditions and if the plot was not forested before 1990. See chapter 1 for more details.

• 196 • 8 Abstract

(instead of complete real time data a three-year loop was used) or neglect of maintenance as well as inaccuracy and error propagation in context with the inventories.

• 197 • 9 Kurzfassung

9 Kurzfassung Auf Leyte, Philippinen, wurde in den 1990er Jahren der Rainforestation-Ansatz zur Inwertsetzung degradierter Flächen durch Aufforstung mit einheimischen Baumarten entwickelt. Im Rahmen der vorliegenden Arbeit wurden auf bestehenden Rainforestation- Flächen folgende Aspekte untersucht:

● Standortskundliche Charakterisierung typischer Böden in Hanglagen im Hinblick auf physikalische, chemische und biologische Parameter; ● Beitrag der Rainforestation-Systeme zur Bodenrehabilitation im Vergleich mit traditionellen Landnutzungen; ● Identifizierung geeigneter Wuchsbedingungen einer Neupflanzung unter kleinräumig unterschiedlichen Boden- und Klimabedingungen; ● mittel- bis langfristige Potenziale zur CO2-Sequestrierung einer Rainforestation- Pflanzung im Rahmen von Projekten des Clean Development Mechanism (CDM) mit Hilfe eines Wachstumsmodells.

Zunächst wurden Bodenprofile an 8 Rainforestation-Standorten gegraben und zur Charakterisierung der Bodeneigenschaften horizontweise beprobt. Die Böden in den Hanglagen Leytes können zunächst nach ihrer Entstehung aus Korallenkalk oder Vulkangestein unterschieden werden. Die Kalkböden (Cambisol und Leptosol) sind neben typischerweise alkalischen pH-Werten und hohen Gehalten an verfügbaren Ca2+- und Mg2+-Ionen durch niedrige Gehalte an pflanzenverfügbarem Phosphor gekennzeichnet. Die Vulkanböden (Luvisols, Cambisols, Nitisol) weisen dagegen in aller Regel pH-Werte zwischen 4 und 5, relativ geringe effektive KAK und extrem geringe P-Gehalte auf. Eine weitere Unterteilung der vulkanischen Standorte in Kolluvien und in situ entstandene Böden ist möglich. Von den untersuchten Standorten waren die wenigsten hinsichtlich ihrer Humus- und Stickstoffgehalte unterversorgt. Auch Porenvolumen und Wasserinfiltration, beide relevant für Bodenerosion, wurden durchgehend als günstig eingestuft. Allerdings kann Hangzugwasser über BT-Tonhorizonten wechselfeuchte Bedingungen hervorrufen. Besonders auf den Kalkböden kann das Pflanzenwachstum zudem durch temporäre Trockenheit und Verhärtung der tonreichen Unterböden eingeschränkt werden.

Im Zusammenhang mit der Durchführung von Rainforestation-Projekten wurde in der Vergangenheit oft die Vorzüglichkeit dieser Systeme bei der Rehabilitation übernutzter bzw. erodierter Hanglagen herausgestellt. Diese Aussage wurde mittels eines Vergleichs mindestens 10-jähriger Rainforestationflächen an 7 verschiedenen Standorten zu jeweils direkt benachbarten Brachen bzw. Aufforstungen mit Gmelina arborea untersucht. Dazu wurden Bodenproben der Flächen auf chemische und biologische Parameter hin analysiert und per t-Test verglichen. Statistisch signifikante Unterschiede zwischen den Landnutzungen bestanden bei den pH-Werten und Gehalten an austauschbarem Mg2+, welche in Rainforestation-Böden unter denjenigen der jeweiligen Referenzflächen lagen. Für Ca2+ und P waren die Unterschiede weniger deutlich. Eine generelle Verarmung von Böden an basischen Kationen durch Festlegung in Pflanzen erscheint plausibel. So fand NYKVIST (1997) in malaysischen Waldökosystemen eine Einlagerung von 50% des im Ökosystem enthaltenen Ca2+ in der Pflanzenbiomasse. Zugleich würde eine mit der Sukzession einhergehende Versauerung des Bodens (wie von KELLMAN (1970) für Sekundärsukzessionen in Mindanao beschrieben) eine Entbasung, also den Verlust verfügbarer Kationen, des Bodens begünstigen. Dass Bäume als Nährstoffpumpen

• 198 • 9 Kurzfassung fungieren können, konnte u.a. am Standort Marcos gezeigt werden. Der P-Gehalt im Oberboden lag dort mit über 3mg kg-1 deutlich über den gemessenen P-Gehalten im darunter liegenden Horizont mit 0.47mg kg-1. Verbesserte Bodenfruchtbarkeit bzgl. chemischer oder biologischer Parameter durch Rainforestation konnte mit dem verwendeten Ansatz nicht festgestellt werden. Das System kann in diesem Sinne eher als Kapitalanlage von Nährstoffen angesehen werden, welche in unterschiedlichem Maße durch Rückschnitt oder Rodung wieder freigesetzt werden können.

Nahe einer der bestehenden Flächen wurde 2004 eine neue Rainforestation-Pflanzung, bestehend aus einheimischen Wertholzarten, Obstbäumen und Musa textilis (Abaca) angelegt. Hier wurde die Entwicklung der Pflanzen während der kritischen ersten Jahre hinsichtlich Überlebensraten und Biomassezuwachs während wiederholter Inventuren erfasst. Da die Parzelle von einem Hektar Größe hinsichtlich Geländeform, Bodeneigenschaften und Kronenschluss kleinräumig stark variierte, wurden 10 deutlich voneinander verschiedene Teilflächen zur intensiven Beprobung gewählt. Durch die Auswahl wurde ein möglichst weiter Bereich an Umweltbedingungen abgedeckt. Bodenphysikalische und -chemische Parameter sowie mikrobielle Biomasse und Aktivität im Boden und photosynthetisch aktive Einstrahlung wurden ebenso erfasst wie die Streuproduktion und Wurzellängendichte der vorhandenen Vegetation. Diese Parameter wurden zu Überlebens- und Zuwachsraten von Musa textilis in der Neupflanzung in Beziehung gesetzt: Überlebenswahrscheinlichkeiten wurden für jede Position unter den gegebenen Umweltfaktoren mit Hilfe von Logit-Funktionen errechnet. Dabei waren die wichtigsten Standortfaktoren bzgl. der Überlebensraten Humusgehalt, biologische Aktivität im Boden und Streuproduktion. Alle drei Größen hängen wesentlich mit der Bodenfeuchte zusammen. Dies war von Bedeutung, da die höchste Mortalität von Abacapflanzen stets während Trockenperioden auftrat. Zur Bewertung von Einflussgrößen auf Pflanzenbiomasse und -wachstum wurden multiple Regressionen erstellt. Organischer Kohlenstoff und N in der leicht abbaubaren organischen Bodensubstanz waren die bedeutendsten Steuergrößen für Biomasse und Zuwachs von M. textilis in einer anschließenden Sensitivitätsanalyse.

Zur Bewertung des Potenzials kleinbäuerlicher Agroforstpflanzungen für CDM-Projekte wurde ein Modellansatz gewählt. Mit Hilfe des Modells WaNuLCAS (Water, Nutrients, Light and Carbon in Agroforestry Systems) wurden Wachstumsverläufe von Pflanzen auf Grundlage eigener Messungen von Standorteigenschaften (s.o.) und Pflanzenparametern simuliert. Dabei wurden über Zeiträume von 10 bzw. 20 Jahren CO2-Bilanzen für verschiedene Management-Varianten verglichen. In allen Modellläufen war der C-Haushalt des Bodens von herausragender Bedeutung für die CO2-Bilanz der gesamten Parzelle. Damit erhalten Bodenschutzmaßnahmen wie Erosionsvermeidung und Humusmehrung besonderes Gewicht bei der Anlage und Pflege der Systeme. Bei Betrachtung der in der Pflanzenmasse gespeicherten Mengen an CO2 zeigte sich eine gute Übereinstimmung mit Literaturdaten von LASCO und Mitarbeitern für Leyte (diverse Publikationen, s. Kap. 6). Von KOLB (2003) erhobene Inventurdaten für zwei der beprobten Standorte wurden dagegen im Modell überschätzt. Ein nennenswerter Beitrag der schnellwachsenden M. textilis zur CO2-Sequestrierung ergab sich in den Modellsimulationen nur während der Anfangsphase. Spätestens zwei Jahre nach der Pflanzung wurde M. textilis jedoch ausschattiert und Bäume dominierten die CO2-Bilanz der Biomasse. Im Vergleich verschiedener Pflanzsysteme zeichnete sich Gmelina arborea

• 199 • 9 Kurzfassung als Pionierart durch schnellen Aufbau von Biomasse, verbunden mit hoher CO2-Fixierung, aus. Dieses System wird gewöhnlich im Aufforstungsansatz staatlicher Programme favorisiert. Eine dem Rainforestation-Ansatz entsprechende Kombination einheimischer Nutzholz- und Obstbaumarten benötigte dagegen längere Zeit zur Entwicklung, nahm aber insgesamt über knapp 20 Jahre größere Mengen an CO2 auf.

• 200 • 10 Acknowledgements

10 Acknowledgements

Conducting this study would not have been possible without the generous funding provided by an anonymous philanthropist and the Vater und Sohn Eiselen – Stiftung, Ulm. In this context, very special thanks go to Dr. Christoph v. Braun for his most valuable contribution, establishing the right contacts in the right moment. SEFAR, Sartorius and Metrohm donated working materials. Dr. Christian Hülsebusch gave useful hints on fundraising.

My sincere thanks, in random order, to everybody who helped me in the Philippines through advise, discussions, collecting data, facilitating logistic support and good company: Dr. Victor Asio, Cheryl Batistel, Joseph Allen Añura, Dr. Rodel D. Lasco, Claudia Kindermann, Daniel Diehl, Dr. Faustino Villamayor, Dr. Paciencia Milan, Dr. Jochen Weingart, Antonio Fernandez, Marcelo Fernandez, ICRAF Leyte, Janice Susaya, Prof. Manuel Posas, Luz Asio, Dr. Remberto Patindol, Dr. Alfred Conklin Jr., Dr. Marco Stark, Klemens Gann.

In Hohenheim, to my supervisor Prof. Dr. J. Sauerborn, Prof. Dr. Georg Cadisch, Dr. Thomas Gaiser, Prof. Dr. E. Kandeler, Dr. Thomas Hilger, Dr. Jan Grenz, Sabine Rudolph, Oliver Koch, Christian Poll, Dr. Dagmar Tscherko, PD Dr. Konrad Martin, Prof. Dr. K. Stahr, Prof. Dr. H.-P. Piepho, Karin Breuer, Dr. J. Breuer, Dr. Gerhard Langenberger, Dr. Gerd Dercon, Annika Badorreck, Dr. Mehdi Zahrei, Inge Koch, Annett Göltenboth and Nina Nikolic, who all gave very valuable input during discussions or carrying out lab analyses. Further to all colleagues at 380 who spread good humour.

In Halle, to Prof. Dr. Reinhold Jahn for his willingness to supervise this study and guidance during soil survey in the field.

Tributes go to Prof. Dr. Eckard George, Dr. Thomas Gaiser and Prof. Dr. W. Koch, who sparked my interest in science. They tought me that research cannot be separated from social responsibility and should be fueled by cooperation rather than competition.

Finally and most important, I am most grateful to my home base in Stuttgart, everybody who helped keeping my spirits up, when it was necessary (and it was):

My parents Jutta and Gustav Marohn to whom I dedicate this study with gratitude for their permanent unconditional support;

Isabel Marohn and all other family members;

KKKarin & Co. in Rohracker, Ste, Dunja, Barbara and Arbeitersportverein Filderstadt.

• 201 • 11 References

11 References rd • AG BODENKUNDE (1982): Bodenkundliche Kartieranleitung, 3 ed. Schweizerbart, Stuttgart, Germany. • AGRICULTURAL EXPERIMENT STATIONS OF CONNECTICUT, DELAWARE, MAINE, MARYLAND, MASSACHUSETTS, NEW HAMPSHIRE, NEW JERSEY, NEW YORK, PENNSYLVANIA, RHODE ISLAND, VERMONT AND WEST VIRGINIA (ED.)(1995): Bulletin # 493, 2nd edition. Agricultural Experiment Station, University of Delaware, Newark, USA. • AHRENS, O., S. HENDERS, M. LANGKAU, S. LINDEMANN, T. MÜLLER & M. PETRI (2004): Cost- benefit analysis – comparison of different land uses in Leyte, Philippines. Faculty of Forest Science and Wood Ecology, Georg-August University, Göttingen, Germany. Unpublished. • ALBRECHT, A. & S.T. KANDJI (2003): Carbon sequestration in tropical agroforestry systems. Agriculture, Ecosystems and Environment 99:15–27. • ALVAREZ, C.R., R. ALVAREZ, M.S. GRIGERA & R.S. LAVADO (1998): Associations between organic matter fractions and the active soil microbial biomass. Soil Biology & Biochemistry 30(6):767-733. • AKINNIFESI, F.K., B.T. KANG & D.O. LAPIDO (1999): Structural root form and fine root distribution of some woody species evaluated for agroforestry systems. Agroforestry Systems 42:121-138. • ALBRECHT, A. & S.T. KANDJI (2003): Carbon sequestration in tropical agroforestry systems. Agriculture, Ecosystems and Environment 99:15-27. • ANDERSON, A.B., A. GELY, J. STRUDWICK, G.L. SOBEL & M.G.C. PINTO (1985): Um sistema agroflorestal na várzea do estuário amazônico. Acta Amazonica 15 (1-2): 195-224. • ANDERSON, J.M.E. & J.S.I. INGRAM (EDS.)(1993): Tropical soil biology and fertility. A handbook of methods, 2nd edition. CAB International, Wallingford, UK. • ANDERSON, J.P.E. & K.H. DOMSCH (1978): A physiological method for the quantitative measurement of microbial biomass in soils. Soil Biology & Biochemistry 10:215-221. • ANDERSON, T.-H. (1994): Physiological analysis of microbial communities in soil: Applications and limitations. In: RITZ, K., DIGHTON, J. & K.E. GILLER (1994): Beyond the biomass. British Society of Soil Science, Wiley-Sayce, UK. • ANDERSON, T.-H. & K.H. DOMSCH (1985): Maintenance carbon requirements of actively- metabolizing microbial populations under in-situ conditions. Soil Biology & Biochemistry 17 (2):197-203. • ANDERSON, T.-H. & K.H. DOMSCH (1986): Carbon assimilation and microbial activity in soil. Journal of Plant Nutrition and Soil Science 149:457-468. • ANDERSON, T.-H. & K.H. DOMSCH (1989): Ratios of microbial biomass carbon to total organic carbon in arable soils. Soil Biology & Biochemistry 21 (4):471-479. • ANDERSON, T.-H. & K.H. DOMSCH (1990): Application of eco-physiological quotients (qCO2 and qD) on microbial biomasses from soils under different cropping histories. Soil Biology & Biochemistry 22(2):251-255. • ANDERSON, T.H. & R.G. JOERGENSEN (1997): Relationship between SIR and FE estimates of microbial biomass C in deciduous forest soils at different pH. Soil Biology & Biochemistry 29(7):1033-1042. • ANDERSSON, M., A. MICHELSEN, M. JENSEN, A. KJØLLER (2004): Tropical savannah woodland: Effects of experimental fire on soil microorganisms and soil emissions of carbon dioxide. Soil Biology & Biochemistry 36:849–858. • ANONYMOUS (1993): Greater Tongonan Geothermal Field integrated resource assessment. LGP Geoscientific, Manila, Philippines. • AQUINO, R. ET AL. (1983): Semi-detailed geology of the Bato-Lunas and vicinity Burauen

• 202 • 11 References

geothermal prospect. PNOC Leyte, Philippines. • ARAGON, J.S. (2004): Leaf litter decomposition of Dipterocarpus validus Brandis (Dipterocarpaceae) and Gmelina arborea Roxb. (Verbenaceae) in two forest sites in Mt. Pangasugan. B.Sc. thesis, Leyte State University, Philippines. • ARAKAWA, H. (ED.)(1969): World survey of climatology Vol.8. Climates of Northern and Eastern Asia. Elsevier, Amsterdam, The Netherlands. • ASHTON, M.S. & F. MONTAGNINI (EDS.)(2000): The silvicultural basis for agriforestry systems. CRC Press, Boca Raton, USA. • ASHTON, M.S. & M.J. DUCEY (2000): Agroforestry systems as successional analogs to native forests. In ASHTON, M.S. & F. MONTAGNINI (EDS.)(2000): The silvicultural basis for agriforestry systems. CRC Press, Boca Raton, USA. • ASIO, V.B., C.C. CABUNOS & Z.S. CHEN (2006): Morphology, physiochemical characteristics, and fertility of soils from quaternary limestone in Leyte, Philippines. Soil Science 171(8): 648-661. • ASIO V., R. JAHN & K. STAHR (1999): Changes in the properties of a volcanic soil (Andisol) in Leyte due to conversion of forest to other land uses. Philippine Journal of Science 128/1, 1-13. • ASIO, V.B. (1994): Upland agriculture, population pressure and environmental degradation in the Philippines: A review and analysis. Annals of Tropical Research Vol.X, ViSCA, Baybay, Leyte, Philippines. • ASIO, V.B. (1996): Characteristics, weathering, formation and degradation of soils from volcanic rocks in Leyte, Philippines. PhD Dissertation University of Hohenheim. Hohenheimer Bodenkundliche Hefte 33. • ASIO, V.B., R. JAHN, K. STAHR & J. MARGRAF (1998): Soils of the tropical forests of Leyte, Philippines. II: Impact of different land uses on status of organic matter and nutrient availability. In SCHULTE, A. & D. RUHIYAT (EDS.): Soils of tropical forest ecosystems. Springer Berlin, Germany. pp. 38-44. • ASSOCIACIÓN DE AGRICULTURA ECOLÓGICA (1998): Pasos hacía una agricultura más rentable y ecologicamente apropiada. Pto. Maldonado, Perú. • BACOLCOL, T., E. BARRIER & T. DUQUESNOY (2004): Study of the Philippine Fault in the Visayas through spatial geodesy. Western Pacific Geophysics Meeting, Honolulu, Hawaii, Aug 16-20, 2004. • BALÉE, W. (1989): The culture of Amazonian forests. In POSEY, D.A. & W. BALÉE (EDS.): Resource management in Amazonia: Indigenous and folk strategies. The New York Botanical Garden, USA. Advances in Economic Botany, Vol.7. • BALESDENT, J. (1996): The significance of organic separates to carbon dynamics and its modeling in some cultivated soils. European Journal of Soil Science 47:485-493. • BALZER, P. (1994): The climate in ViSCA, Baybay, Leyte, Philippines. Annals of Tropical Research Vol.X, ViSCA, Baybay, Leyte, Philippines. • BARETH, G. (2000): Emissionen klimarelevanter Gase aus der Landwirtschaft – Regionale Darstellung und Abschätzung unter Nutzung von GIS am Beispiel des württembergischen Allgäus. PhD Dissertation University of Hohenheim, Germany. Hohenheimer Bodenkundliche Hefte 56. • BARRERA, A., I. ARISTORENAS & J. TINGZON (1954): Soil survey of Leyte province, Philippines. Bureau of Print, Manila, Philippines. • BARRIOS, E., F KWESIGA, R.J. BURESH & J.I. SPRENT (1997): Light fraction soil organic matter and available nitrogen following trees and maize. Soil Sciences of America Journal 61:826-831. • BASS, S., O. DUBOIS, P. MOURA COSTA, M. PINARD, R. TIPPER & C. WILSON (2000): Rural livelihood and carbon management. IIED Natural Resources Issues Paper No.1. IIED,

• 203 • 11 References

London, UK. • BATISTEL, C.C. (2004): Effects of leaf decomposition of selected exotic and indigenous tree species on forest soil quality. B.Sc. Thesis, Leyte State University, Philippines. • BATTAGLIA, M. (2004): What is CABALA? http://www.ffp.csiro.au/downloads/CABALA.pdf, download 04/03/07. • BAUHUS, J. & C. MESSIER (1999): Image analysis – evaluation of fine root length and diameter measurements obtained using RHIZO image analysis. Agronomy Journal 91:142-147. • BAYALA, J., J.S.J. OUEDRAOGO, M.V. NOORDWIJK, B. LUSIANA & Z. TEKLEHAIMANOT (2004): Separating tree-soil-crop interactions in agroforestry parkland systems in Saponé (Burkina Faso) using WaNuLCAS. Presentation 1st Agroforestry World Congress, Florida, USA. • BAYRANTE, L.F. (1982): The geology of the Burauen Geothermal Prospect Leyte, Central Philippines. Technical Bulletin Vol. I, No.1:35-47. PNOC Geothermal Division, Manila, Philippines. • BECK, L. & L. GASPAROTTO (2000): Fauna de solo e decomposição da serapilheira em florestas primárias e secundárias e em sistema de cultivo misto na Amazônia. Relatório Final 1996-1999. Internet Source http://www.cpaa.embrapa.br/env52 download Feb 11th, 2007. • BECK, T., R.G. JOERGENSEN, E. KANDELER, F. MAKESCHIN, E. NUSS, H.R. OBERHOLZER & S. SCHEU (1997): An inter-laboratory comparison of ten different ways of measuring soil microbial biomass C. Soil Biology & Biochemistry 29 (7):1023-1032. • BEERBAUM, S. (2001): Kosteneffiziente CO2- Minderungsmaßnahmen im Rahmen des Clean Development Mechanism – dargestellt am Beispiel Deutschland und Indien. PhD Dissertation University of Hohenheim. Logos, Berlin, Germany. rd • BIRKELAND, P.W. (1999): Soils and geomorphology. 3 ed. Oxford University Press, New York, USA. • BOOGAARD, H.L., C.A. VAN DIEPEN, R.P. ROTTER, J.M.C.A. CABRERA & H.H. VAN LAAR (1998): WOFOST 7.1; user's guide for the WOFOST 7.1 crop growth simulation model and WOFOST control center 1.5. SC-DLO.Techn. Doc. 52, Wageningen, The Netherlands. • BORINES, L.M. & V.G. PALERMO (2006): Major diseases of abaca and the role of plant disease diagnostic laboratory (PDDL). Flyer, Leyte State University, Philippines. • Brack Egg, A. (1999): Diccionario enciclopédico de plantas útiles del Perú. CBC, Cusco, Perú. • BRAKE, M., H. HÖPER & R.G. JÖRGENSEN (1999): Land-use induced changes in activity and biomass of microorganisms in raised bog peats at different depths. Soil Biology & Biochemistry 31:1489-1497. • BRECKLING, B. & P. BIRKENMEIER (2000): Landnutzungsalternativen im Regenwald. P. Lang, Frankfurt/Main, Germany. • BROWN S. (1997): Estimating biomass and biomass change of tropical forests, a primer. FAO Forestry paper 134, FAO, Rome, Italy. • BROWN, S., L.R. IVERSON, A. PRASAD & D. LIU (1993): Geographical distribution of carbon in biomass and soils of tropical Asian forests. Geocarto International 4. • BROWN, W.H. (1919): Philippine fiber plants. Dept. Of Agriculture and Natural Resources, Bureau of Forestry. Bulletin No.19. Bureau of Printing, Manila, Philippines. • BROWN, S. & A.E. LUGO (1990): Effects of forest clearing and succession on the carbon and nitrogen contentof soils in Puerto Rico and US Virgin Islands. Plant and Soil 124:53-64. • BUGAYONG, L.A. (2004): Government policy as determinant of local agroforestry practices in the Philippines. In: First World Congress of Agroforestry, Book of Abstracts. Orlando,

• 204 • 11 References

FL, USA. • BULLECER, R.C. & M. STARK (2004): Coconut-based agroforestry farming systems in Central Philippines. Presentation 1st World Congress of Agroforestry, Book of Abstracts. Orlando, FL, USA. • BURESH, R.J. (1999): The contribution of trees to nutrient cycling and availability. In: Report of the first research co-ordination meeting of the FAO/IAEA co-ordinated research project on the use of nuclear techniques for developing integrated nutrient and water management practices for agroforestry systems. FAO/IAEA. • BUREAU OF COAST AND GEODETIC SERVICE (1982): Map of Baybay, Scale 1:50000. Sheet 3952 II. DENR, Cebu, Philippines. • CAB INTERNATIONAL (2002): Tree compendium. Online - data base. www.cabicompendium.org download Feb 11th, 2007. • CADISCH, G. & K.E. GILLER (1997): Driven by nature. Plant litter quality and decomposition. CAB International, Wallingford, UK. • CADISCH, G., H. IMHOF, S. URQUIAGA, R.M. BODDEY & K.E. GILLER (1996): Carbon turnover (13C) and nitrogen mineralization potential of particulate light soil organic matter after rainforest clearing. Soil Biology & Biochemistry 28(12):1555-1567. • CANNELL, M.G.R., M. V.NOORDWIJK & C.K. ONG (1996): The central agroforestry hypothesis: The trees must acquire resources that the crop would not otherwise acquire. Agroforestry Systems 34: 27-31. • CAREY, E.V., A. SALA, R. KEANE & C.R. MU (2001) in Global Change Biology: Are old forests underestimated as global carbon sinks? pp. 339-344. Blackwell. • CENIZA, M.J.C., P.P. MILAN, V.B. ASIO, S.B. BULAYOG, M.D. NAPIZA, M.J.M. BANDE, H.L. MONDAL, D.B. POSAS & J. POGOSA (2004): Status, ecological changes and market study of rainforestation farms. Unpublished report. • CHAMBERS, J.Q., N. HIGUCHI, E.S. TRIBUZY & S.E. TRUMBORE (2001): Carbon sink for a century. Nature 410:429. • CHERTOV, O.G. & A.S. KOMAROV (1996): SOMM – a model of soil organic matter and nitrogen dynamics in terrestrial ecosystems. In: POWLSON, D.S., P. SMITH & J.U. SMITH (EDS.): Evaluation of soil organic matter models. Springer, Berlin, Germany, in cooperation with NATO Scientific Affairs Division. • CHOKKALINGAM, U., A.P. CARANDANG, J.M. PULHIN, R.D. LASCO, R.J.J. PERAS & T. TOMA (EDS.) (2006): One century of forest rehabilitation in the Philippines. Approaches, outcomes and lessons. SMK Grafika Desa Putera Jakarta, Indonesia. • CLARK, D.A., S. BROWN, D.W. KICKLIGHTER, J.Q. CHAMBERS, J.R. THOMLINSON & J. NI (2001): Measuring net primary productivity in forests: Concepts and field methods. Ecological Applications 11(2):356-370. • CLARK, D.A., S. BROWN, D.W. KICKLIGHTER, J.Q. CHAMBERS, J.R. THOMLINSON, J. NI & E.A. HOLLAND (2000): Net primary production in tropical forests: An evaluation and synthesis of existing field data. Ecological Applications 11(2):371–384. • CODY, R.P. & J.K. SMITH (1997): Applied statistics and the SAS programming language, 4th ed. Prentice-Hall, NJ, USA. • COICAP (COORDINADORA INDÍGENA CAMPESINA DEL PERÚ (1999): Manual práctico de forestería análoga. Pto. Maldonado, Perú. • COLEMAN, K. & D.S. JENKINSON (1996): RothC-26.3 – A model for the turnover of carbon in soil. In: POWLSON, D.S., P. SMITH & J.U. SMITH (EDS.): Evaluation of soil organic matter models. Springer, Berlin, Germany, in Cooperation with NATO Scientific Affairs Division. • COLEMAN, K., D.S. JENKINSON, G.J. CROCKER, P.R. GRACE, J. KLÍR, M. KÖRSCHENS, P.R. POULTON & D.D. RICHTER (1997): Simulating trends in soil organic carbon in long-term

• 205 • 11 References

experiments using RothC-26.3. Geoderma 81:29-44. • CONKLIN, H.C. (1957): Hanunóo agriculture. A report on an integral system of shifting cultivation in the Philippines. FAO Forestry Development Paper No. 12. Food and Agriculture Organization, Rome, Italy. • CONNELL, J.H. (1978): Diversity in tropical forests and coral reefs. Science 199:1302-1310. • CRASWELL, E.T., A. SAJJAPONGSE, D.J.B. HOWLETT & A.J. DOWLING (1997): Agroforestry in the management of sloping lands in Asia and the Pacific. Agroforestry Systems 38(1-3): 121-137. • CUENCA, G., J. ARANGUREN & R. HERRERA (1983): Root growth and litter decomposition in a coffee plantation under shade trees. Plant and Soil 71:477-486. • CUEVAS, V. & P.E. SAJISE (1978): Litterfall and leaf litter decomposition in a Philippine secondary forest. Kalikasan, Philippine Journal of Biology 7(II):99-109. • DAGOY, F.J., F.D. POLIQUIT, L.G. HINAY & L.S.E. RACHO (1994): Sociodemographic profile of the upland farmers in the MPSAP pilot barangays. Annals of Tropical Research. Special Issue on Tropical Ecology No.1. ViSCA, Baybay, Leyte, Philippines. • DARGANTES, B.D. & W. KOCH (1994): Case studies on the occupation and cultivation of the forest lands of Leyte, Philippines. Annals of Tropical Research, Special Issue on Tropical Ecology No.1, ViSCA, Baybay, Leyte, Philippines. • DAUB, M. (2002): Litter arthropod community and its impact on decomposition – a contribution to restoration biodiversity in rainforestation systems, Leyte; Philippines. Master thesis University of Hohenheim. • DAVIS, S.D., V.H. HEYWOOD & A.C. HAMILTON (EDS.)(1995): Centres of plant diversity: A guide and strategy for their conservation. Vol. II: Asia, Australasia and the Pacific. IUCN, Cambridge, UK. • DE CONNINCK, H.C. & N.H. VAN DER LINDEN (2003): An overview of carbon transactions. Report to ECN Project 7.7513.03.01. Energy Research Centre of the Netherlands. http://www.ecn.nl/library/reports/2003/c03022.html download Feb 11th, 2007. • DEL RIO, J. & A. GUTIERREZ (2006): Chemical composition of abaca (Musa textilis) leaf fibers used for manufacturing of high quality paper pulps. J. Agric. Food Chem. (54): 4600-4610. • DENICH, M. &M. KANASHIRO (EDS.)(1998): Potential land-use and management of altered and abandoned areas in the Brazilian Amazon region. MCT/CNPq, Brasília, Brasil. • DIEHL, D. (2005): Growth of tree seedlings on Leyte/Philippines correlated with litter production and decomposition. BSc Thesis, Hohenheim University, unpublished. • DILLY, O. (2001): Microbial respiratory quotient during basal metabolism and after glucose amendment in soils and litter. Soil Biology & Biochemistry 33:117-127. • DILLY, O. & J.-C. MUNCH (1998): Ratios between estimates of microbial biomass content and microbial activity in soils. Biology & Fertility of Soils 27:374-379. • DINESH, R., S.G. CHAUDHURI & T.E. SHEEJA (2004): Soil biochemical and microbial indices in wet tropical forests: Effects of deforestation and cultivation. Journal of Plant Nutrition and Soil Sciences 167:24-32. • DIRKSMEYER, W. (1998): Rainforestation farming: A farmer's guide to biodiversity management in the Philippines. • DIXON, R.K. (1995): Agroforestry Systems: Sources or sinks of greenhouse gases? Agroforestry Systems 31:99-116. • DODOR, D.E. & M.A. TABATABAI (2003): Effect of cropping systems on phosphatases in soils. Journal of Plant Nutrition and Soil Sciences 1/2003:7-13. • DONALD, R.G., D.W. ANDERSON & J.W.B. STEWART (1993): Potential role of dissolved organic carbon in phosphorus transport in forested soils. Soil Science Society of

• 206 • 11 References

America Journal 57:1611-1618. nd • DUNGER, W. & H.J. FIEDLER (EDS.)(1997): Methoden der Bodenbiologie. 2 ed. Fischer, Jena, Germany. • DURST, P.B., T.R. WAGGENER, T. ENTERS & T.L. CHENG (2001): Forests out of bounds: Impacts and effectiveness of logging bans in natural forests in Asia-Pacific. FAO Regional Office for Asia and the Pacific. Bangkok, Thailand. • ECKSTEIN, K. & J.C. ROBINSON (1996): Physiological responses of banana (Musa AAA, Cavendish subgroup) in the Subtropics. VI. Seasonal responses of leaf-gas exchange to short-term water stress. Journal of Horticultural Science 71(5):679-692. • ECKSTEIN, K., J.C. ROBINSON & C. FRASER (1997): Physiological responses of banana (Musa AAA, Cavendish subgroup) in the Subtropics. VII. Effects of windbreak-shading on phenology, physiology and yield. Journal of Horticultural Science 72(3):389-396. • EISENBEIS, G. (2004): Soil Ecology Methods. Internet source: http://www.uni- mainz.de/FB/Biologie/Zoologie/abt1/eisenbeis/Homepage_Minicontainer.htm download Feb 11th, 2007. • EWEL, J.J. (1999): Natural systems as models for the design of sustainable systems of land-use. In: LEFROY, T., R. HOBBS, M. O´CONNOR & J. PATE (EDS): Agriculture as mimic of natural ecosystems. Kluwer, Dordrecht, The Netherlands. • FALLOON, P.D., P. SMITH, J.U. SMITH, J. SZABÓ, K. COLEMAN, S. MARSHALL (1998): Regional estimates of carbon sequestration potential: Linking the Rothamsted carbon model to GIS databases. In Biology and Fertility of Soils 27:236-241. • FANG, C., P. SMITH, J.B. MONCRIEFF & J.U. SMITH (2005): Similar response of labile and resistant soil organic matter pools to changes in temperature. Nature 433:57-59. • FAO Gateway for Land and Water Information. www.fao.org download Feb 11th, 2007. • FAO, ISSS, ISRIC (1999): World reference base for Soil Resources. FAO, Rome, Italy. • FAO (1995): Mangosteen cultivation. FAO plant production and protection paper 129. FAO, Rome, Italy. • FAO (1982): Fruit-bearing forest trees. Forestry Paper 34. FAO, Rome, Italy. • FEIGL, B.J., G.P. SPARLING, D.J. ROSS & C.C. CERRI (1995): Soil microbial biomass in Amazonian soils: Evaluation of methods and estimates of pool sizes. Soil Biology & Biochemistry 27(11):1467-1472. • FELLER, C. & M.H. BEARE (1997): Physical control of soil organic matter dynamics in the Tropics. Geoderma 79:69-116. • FIBER INDUSTRY DEVELOPMENT AUTHORITY (FIDA): Technoguide on abaca. Extension Flyer. • FIELD, C.B. (2001): Plant physiology of the „missing“ carbon sink. In Plant Physiology 125:25-28. • FILIP, Z. (2002): International approach to assessing soil quality by ecologically-related biological parameters. Agriculture, Ecosystems and Environment 88:169-174. • FOXWORTHY, F.W. (1932): Dipterocarpaceae of the Malay Peninsula. Malayan Forest Records 10. Cited after NG (1966). • FRANKO, U. (1996): Modeling approaches of soil organic matter turnover within the CANDY system. In: POWLSON, D.S., P. SMITH & J.U. SMITH (EDS.): Evaluation of soil organic matter models. Springer, Berlin, Germany, in cooperation with NATO Scientific Affairs Division. • FRANZEL, S., G.L. DENNIN, J.P.B. LILLESØ & A.R. MERCADO JR. (2004): Scaling up the impact of agroforestry: Lessons from three sites in Africa and Asia. Agroforestry Systems 61-62(1-3): 329-344. • GAISER, T. (1993): Bedeutung der organischen Bodensubstanz für Eigenschaften und Ertragfähigkeit von Vertisolen und Acrisolen in Süd-Benin. PhD Dissertation University of Hohenheim. Hohenheimer Bodenkundliche Hefte Vol. 12, Hohenheim, Germany.

• 207 • 11 References

• GAUNT, J.L., S.P. SOHI, J.R.M. ARAH, N. MAHIEU, H. YANG & D.S. POWLSON in KIRK, G.J.D. & D.C. OLK (EDS.)(2000): Carbon and nitrogen dynamics in flooded soils, pp. 89-99: A physical soil fractionation scheme to obtain organic matter fractions suitable for modelling. IRRI, Manila, Philippines. • GEORGE, T.S., P.J. GREGORYA, M. WOODA, D. READA & R.J. BURESH (2002): Phosphatase activity and organic acids in the rhizosphere of potential agroforestry species and maize. Soil Biology & Biochemistry 34(10):1487-1494. • GEROLD, G., M. FREMEREY & E. GUHARDJA (EDS.) (2004): Land use, nature conservation and the stability of rainforest margins in Southeast Asia. Springer, Berlin, Heidelberg, Germany. • GIFFORD, R.M. (2000): The CO2 fertilizing effect: Relevance to the global carbon Cycle (Ch. 5). In WIGLEY, T.M.L. & D.S. SCHIMEL (EDS.): The carbon cycle. Cambridge University Press, UK. • GINOGA, K., Y.C. WULAN, & M. LUGINA (2005): Potential of agroforestry and plantation systems in Indonesia for carbon stocks: An economic perspective. ACIAR Working paper CC14, ACIAR Project ASEM 2002/066. http://www.une.edu.au/febl/Economics/carbon download Feb 12th, 2007. • GÖLTENBOTH, A. (1998): Influence of chemical soil characteristics on the growth of selected tree species in rainforestation systems on Leyte, Philippines. Diploma Thesis, unpublished. • GOLLER, R., W. WILCKE, K. FLEISCHBEIN, C. VALAREZO & W. ZECH (2006): Dissolved nitrogen, phosphorus, and sulfur forms in the ecosystem fluxes of a montane forest in Ecuador. Biogeochemistry 77(1):57-89. • GREGORICH, E.G., M.R. CARTER, D.A. ANGERS, C.M. MONREAL & B.H. ELLERT (1994): Towards a minimum data set to assess soil organic matter quality in agricultural soils. Canadian Journal of Soil Science 74:367-385. • GROETSCHEL, A., R.R. AQUINO, I. BUCHHOLZ, A. IBKENDANZ, T.G. MAZO, N.A. SALES, J. SEVEN & K.C. VICENTUAN (2001): Natural resource management strategies on Leyte Island, Philippines. SLE, Berlin, Germany. • GUARTE, R.C. & F.G. SINON (2005): Processing of high quality abaca (Musa textilis) fibers for industrial application. Presentation 2nd S-S-N Natural Fibre Dialogue, Leyte State University, Leyte, 4th-6th March 2005. • GUO, L.G. & R.M. GIFFORD (2002) in Global Change Biology 8/4: 345-360. Soil carbon stocks and land use change: A meta analysis. Blackwell Science, UK. • HAIRIAH, K., S.M. SITOMPUL, M. V. NOORDWIJK & C. PALM (2001): Methods for sampling carbon stocks above and below ground. ASB Lecture Notes. International Centre for Research in Agroforestry. • HARCOMBE, P.A. (1980): Soil nutrient loss as a factor in early tropical secondary succession. In: Biotropica 12(2):8-14; Special Issue on Tropical Succession. • HART, R.D. (1980): A natural ecosystem analog approach to the design of a successional crop system for tropical forest ecosystems. Biotropica 12(2):73-82; Special Issue on Tropical Succession. rd • HARTGE, K.H. & R. HORN (1999): Einführung in die Bodenphysik. 3 ed. Enke, Stuttgart. • HARTGE, K.H. & R. HORN (1989): Die physikalische Untersuchung von Böden. Enke, Stuttgart, Germany. • HASHIMOTO, S., N. TANAKA, M. SUZUKI, A. INOUE, H. TAKIZAWA, I. KOSAKA, K. TANAKA, C. TANTASIRIN & N. TANGTHAM (2004): Soil respiration and soil CO2 concentration in a tropical forest, Thailand. Journal of Forest Research 9(1):75-79. • HE, J.-S., K.S. WOLFE-BELLIN, B. SCHMID & F.A. BAZZAZ (2005): Density may alter diversity- productivity relationships in experimental plant communities. Basic and Applied Ecology

• 208 • 11 References

6:505-517. • HECHT, S.B. & D.A. POSEY (1989): Preliminary results on soil management techniques of the Kayapó indians. In POSEY, D.A. & W. BALÉE (EDS.): Resource management in Amazonia: Indigenous and folk strategies. The New York Botanical Garden, Advances in Economic Botany, Vol.7. • HOUGHTON, J.T., Y. DING, D.J. GRIGGS, M. NOGUER, P.J. VAN DER LINDEN, X. DAI, K. MASKELL & C.A. JOHNSON (2001): Climate change 2001: The scientific basis. Contribution of working group I to the third assessment report of the Intergovernmental Panel on Climate Change. Published for the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK. • INSAM, H. & K.H. DOMSCH (1988): Relationship between soil organic carbon and microbial biomass on chronosequences of reclamation sites. Microb. Ecol. 15:177-188. • INSAM & HASELWANDTER (1989): Metabolic quotient of the soil microflora in relation to plant succession. Oecologia 79:174-178. • IPCC (2003): Meeting on current scientific understanding of the processes affecting terrestrial carbon stocks and human influences upon them. Geneva, Switzerland, 21–23 July 2003. www.ipcc.ch/pub/carbon.pdf download Feb 11th, 2007. • IPCC (2001): Third Assessment Report. WGII: Climate Change 2001 – Impacts, adaptation and vulnerability. IPCC, Geneva, Switzerland. • IPCC (2000): Summary for policy makers: Land use, land use change and forestry. Online source: http://www.ipcc.ch/pub/srlulucf-e.pdf download Feb 11th, 2007. • IPCC (1996): Greenhouse gas inventories. Vol. II: Workbook. Vol. III: Reference manual: Guidelines for national greenhouse gas inventories. Bracknell, UK. • ISAAC, S.R. & M.A. NAIR (2005): Biodegradation of leaf litter in the warm humid Tropics of Kerala, India. Biology & Fertility of Soils 37:1656-1664. • IZQUIERDO, I., F. CARAVACA, M.M. ALGUACIL & A. ROLDÁN (2003): Changes in physical and biological soil quality indicators in a tropical crop system (La Havana, Cuba) in response to different agroecological management practices. Environmental Management 32(5) 639-645. • JAHN, R. (1998): The soil resources of the Tropics — the necessity of intensified nutrient cycling within ecological land use systems. Proceedings of the International Conference on Applied Tropical Ecology in Visayas State College of Agriculture, Baybay, Leyte, Philippines. • JAHN, R. & V. ASIO (1998): Soils of the tropical forests of Leyte, Philippines: I. Weathering, soil characteristics, classification and site qualities. In SCHULTE, A. & D. RUHIYAT (EDS.): Soils of tropical forest ecosystems: Characteristics, ecology and management. Springer, Berlin, Germany. pp.29-38. • JAHN, R., H.-P. BLUME & V. ASIO (EDS.)(2003): Students guide for soil description, soil classification and site evaluation. Institute of Soil Science and Plant Nutrition, Martin Luther University, Halle/Saale, Germany. • JAMALUDHEEN, V. & B. MOHAN KUMAR (1999): Litter of multipurpose trees in Kerala, India: Variations in the amount, quality, decay rates and release of nutrients. Forest Ecology and Management 115(1):1-11. • JOERGENSEN, R.G. & X. CASTILLO (2001): Interrelationships between microbial and soil properties in young volcanic ash soils of Nicaragua. Soil Biology & Biochemistry 33:1581-1589. • JOERGENSEN RG & S. SCHEU (1999): Depth gradients of microbial and chemical properties in moder soils under beech and spruce. Pedobiologia 43:134–199 • KAISER, E.A., T. MUELLER, R.G. JOERGENSEN, H. INSAM & O. HEINEMEYER (1992): Evaluation of methods to estimate the soil microbial biomass and the relationship with soil texture

• 209 • 11 References

and organic matter. Soil Biology and Biochemistry 24(7):675-683. • KANNINEN, M. (2004): Agroforestry Systems and Carbon Sequestration: Potential and Perspectives. Presentation 1st Agroforestry World Congress 2004, Florida, USA. • KELLMAN, M.C. (1970): Secondary plant succession in tropical montane Mindanao. Publication BG/2. Australian National University, Canberra, Australia. • KETTERINGS, Q.M., R. COE, M. V.NOORDWIJK, Y. AMBAGAU & C.A. PALM (2001): Reducing uncertainty in the use of allometric biomass equations for predicting above-ground tree biomass in mixed secondary forests. Forest Ecology and Management 146:199-209. • KHANNA, P.K. (1998): Nutrient cycling under mixed-species tree systems in Southeast Asia. Agroforestry Systems 38:99-120. • KÖPPEN, W. (1931): Grundriss der Klimakunde. 2. Aufl. W. de Gruyter & Co., Berlin, Germany. • KOLB, M. (2003): Silvicultural analysis of rainforestation farming areas on Leyte Island, Philippines. Magister Thesis, University of Göttingen, Germany. • KUMMER, D.M. & B.L. TURNER II (1994): The human causes of deforestation in Southeast Asia. BioScience 44(5):323-328. • KUMMER, D.M. (1992): Deforestation in the Postwar Philippines. Ateneo de Manila University Press, Manila, Philippines. • LANGENBERGER, G. (2005): Age of first fruiting in two planted Philippine Dipterocarp species. Flora Malesiana Buletin 13(4):369-370. • LANGENBERGER, G. (2003): Diversität, Struktur und Reliefabhängigkeit der Vegetation in einem Tieflandregenwald der Insel Leyte, Philippinen. Apia Verlag, Laubach, Germany. • LASCO, R.D., F.B. PULHIN, R.V.O. CRUZ, J.M. PULHIN & S.S.N. ROY (2005A): Carbon budgets of terrestrial ecosystems in the Pantabangan-Carranglan Watershed. AIACC Working Paper No.10. • LASCO, R.D., F.B. PULHIN & M.R.N. BANATICLA (2005B): Potential carbon sequestration projects in the Philippines. In: J.E. PADILLA, E.E. TONGSON & R.D. LASCO (EDS.): PES: Sustainable financing for conservation and development. WWF, ICRAF, Manila, Philippines. • LASCO, R.D. & F.B. PULHIN (2004): Carbon Sequestration of Agroforestry Systems in Southeast Asia: Potentials for the Clean Development Mechanism (CDM). Presentation 1st Agroforestry World Congress, Florida, USA. • LASCO R.D., F.B. PULHIN, J.M. ROSHETKO & R. BANATICLA (2004): LULUCF climate change mitigation projects in the Philippines: A Primer. Laguna, Philippines. • LASCO R.D., J.S. LALES, M.T. ARNUEVO, I.Q. GUILLERMO, A.C. DE JESUS, R. MEDRANO, O.F. BAJAR & C.V. MENDOZA (2002): Carbon dioxide (CO2) storage and sequestration of land cover in the Leyte Geothermal Reservation. Renewable Energy 25(2):307-315. • LASCO, R.D., R.G. VISCO & J.M. PULHIN (2001): Secondary forests in the Philippines: Formation and transformation in the 20th century. Journal of Tropical Forest Science 13 (4): 652-670. • LASCO, R.D. & P.D. SUSON (1999): A Leucaena leucocephala -based indigenous fallow system in central Philippines: The Naalad. Int. Tree Crops J. 10:161-174. • LAVELLE, P. & A.V. SPAIN (2005): Soil ecology. Springer, Dordrecht, The Netherlands. • LAVELLE, P., E. BLANCHART, A. MARTIN, S. MARTIN & A. SPAIN (1993): A hierarchical model for decomposition in terrestrial ecosystems. Application to soils in the Humid Tropics. Biotropica 25(2):130-150. • LEHMANN, J. & J. GAUNT (2004): Carbon Sequestration in Soils under Agroforestry. Presentation 1st Agroforestry World Congress, Florida, USA. • LEHMANN, J., D. GÜNTHER, M. SOCORRO DA MOTA, M. PEREIRADE ALMEIDA, W. ZECH & K. KAISER (2001): Inorganic and organic soil phosphorus and sulfur pools in an Amazonian

• 210 • 11 References

multistrata agroforestry system. Agroforestry Systems 53:113-124. • LEITE, L.F.C., E.S. MENDONÇA & P.L.O.A. MACHADO (2004): Simulação pelo modelo CENTURY da dinâmica da matéria orgânica de um Argissolo sob adubação mineral e orgânica. R. Bras. Ci. Solo 28:347-358. • LIN, Q. & P.C. BROOKES (1999): An evaluation of the substrate-induced respiration method. Soil Biology & Biochemistry 31:1969-1983. • LLOYD, J. (1999): The CO2 dependence of photosynthesis, plant growth responses to elevated CO2 concentrations and their interactions with soil nutrient status. Temperate and boreal forest productivity and the combined effects of increasing CO2 concentrations and increased nitrogen deposition at a global scale. Functional Ecology 13/4:439-459. • LOAGUE, K. & R.E. GREEN (1991): Statistical and graphical methods for evaluating solute transport models: Overview and application. Journal of Contaminant Hydrology, 7:51-73. • LUIZÃO, R.C.C., T.A. BONDE & T. ROSSWALL (1992): Seasonal variation of soil microbial biomass – the effects of clearfelling a tropical rainforest and establishment of pasture in the Central Amazon. Soil Biology & Biochemistry 24(8):805-813. • MACKENSEN, J., R. KLINGE, D. RUHIYAT, & H. FOLSTER (2003): Assessment of management- dependent nutrient losses in tropical industrial tree plantations. Ambio 32(2):106-112. • MAGID, J., G. CADISCH & K.E. GILLER (2002): Short and medium term plant litter decomposition in a tropical Ultisol elucidated by physical fractionation in a dual 13C and 14C isotope study. Soil Biology & Biochemistry 34(9):1273-1281. • MAMILOV, A.S. & O.M. DILLY (2002): Soil microbial eco-physiology as affected by short- term variations in environmental conditions. Soil Biology & Biochemistry 34:1283-1290. • MAO, D.M., Y.W. MIN, L.L. YU, R. MARTENS & H. INSAM (1992): Effect of afforestation on microbial biomass and activity in soils of tropical China. Soil Biology & Biochemistry 24(9):865-872. • MARGRAF, J. & P.P. MILAN (1996): Rainforestation farming. A farmer's guide to biodiversity management for the Philippines. GtZ Applied Tropical Ecology Program, Manila, Philippines. • MARSCHNER, H. (1993): Mineral Nutrition of Higher Plants. Academic Press, London. • MARTIUS, C., H. HÖFER, M.V.B. GARCIA, J. RÖMBKE, B. FÖRSTER & W. HANAGARTH (2004): Microclimate in agroforestry systems in central Amazonia: Does canopy closure matter to soil organisms? Agroforestry Systems 60:291-304. • MASRI, M. (1991): Root distribution of durian (Durio zibethinus Murr.) cv. D 24. MARDI Res. J 19(2):183-189. Malaysian Agricultural Research and Development Institute, KL, Malaysia. • MENYAILO, O.V., J. LEHMANN, M. DA SILVA CRAVO & W. ZECH (2003): Soil microbial activities in tree-based cropping systems and natural forests of the central Amazon, Brazil. Biology & Fertility of Soils 38:1-9. • MERCADO JR., A., M.V. NOORDWIJK, T. HILGER & G. CADISCH (2006): Acacia mangium as an intercrop in mixed tree-based contour hedgerow systems on sloping acid upland soils. Submitted journal article (Agroforestry systems). • MERCADO JR., A., G. CADISCH, M. V.NOORDWIJK & T. HILGER (2005): Nitrogen facilitation and competition in timber tree – maize agroforestry systems. Presentation. Smallholder agroforestry options for degraded tropical soils. International Consultation Workshop 18-21 August 2005, Malang, Indonesia. • MILZ, J. (2001): Guía para el establecimiento de sistemas agroforestales en Alto Beni, Yucumo y Rurrenabaque. Edited by DED & CARE - MIRNA. Grupo Design, La Paz, Bolivia.

• 211 • 11 References

• MOLINA, J.A.E. (1996): Description of the model NCSOIL. In: POWLSON, D.S., P. SMITH & J.U. SMITH (EDS.): Evaluation of soil organic matter models. Springer, Berlin, Germany in cooperation with NATO Scientific Affairs Division. • MOLLISON, B. (1988): Permaculture. A designer's manual. Tagari Publications, Tasmania, Australia. • MONGELI PENEIREIRO, F. (1999): Sistemas agroflorestais dirigidos pela suçessão natural: Um estudo de caso. MSc Thesis Universidade de Sao Paulo. Unpublished. • MONTAGNINI, F. (2002) IN REDDY, V. (ED.)(2002): Management of tropical plantation forests and their soil-litter system. Science Publishers, Enfield, UK. • MONTAGNINI, F. & P.K.R. NAIR (2004): Carbon sequestration: An underexploited environmental benefit of agroforestry systems. Agroforestry Systems 61:281-295. • MORENO, L.O., A.A. PARAC & F.L. OCON (2005): Fiber characteristics of promising abaca accessions in NARC germplasm suited for specific iIndustry end uses. Presentation 2nd S-S-N Natural Fibre Dialogue, Leyte State University, Leyte, Philippines, 4th-6th March 2005. • MOREIRA, A. & E. MALAVOLTA (2004): Dinâmica da matéria orgânica e da biomassa microbiana em solo submetido a diferentes sistemas de manejo na Amazônia Ocidental. Pesqisa agropecuaria brasileira 39(11):1103-1110. • MULIA, R., P. PURNOMOSIDDHI & B. LUSIANA (2001): Tree parameterization in WaNuLCAS model. International Centre for Research in Agroforestry, Bogor, Indonesia. • MULIA, R. (2001): Brief explanation for WanFBA. A program for estimating shoot and root biomass. International Centre for Research in Agroforestry. • MURTHY, S.V.K. & B.R.V IYENGAR (1997): Root distribution and morphology in some banana (Musa paradisiaca) varieties. Indian Journal of Agricultural Sciences (IND) 67, (11):495-499. • MURTY, D., M.F. KIRSCHBAUM, R.E. MCMURTRIE & H. MCGILVRAY (2002): Does conversion of forest to agricultural land change soil carbon and nitrogen? A review of the literature. In Global Change Biology 8(2):105-123. Blackwell. • NAMRIA (National Mapping and Resource Information Authority): Land cover statistics 2003. Internet Source www.namria.gov.ph, download Feb 11th, 2007. • NATIONAL STATISTICAL COORDINATION BOARD: Gross regional domestic product 2004. Internet source, download 20/05/07 from www.nscb.gov.ph • NATIONAL STATISTICS OFFICE (2000): 2000 census of population and housing – reports by region. Internet source, download 20/05/07 from www.census.gov.ph • NG, F.S.G. (1966): Age of first flowering in Dipterocarps. Malay. Forester 29(4):290-295. • NIEUWOLT, S. (1977): Tropical climatology. J. Wiley & Sons, Chichester, UK. • NYKVIST, N. (1997): Total distribution of plant nutrients in a tropical rainforest ecosystem, Sabah, Malaysia. Ambio 26(3):152-157. • NYKVIST, N., H. GRIP, B.L. SIM, A. MALMER & F.K. WONG (1994): Nutrient losses in forest plantations in Sabah, Malaysia. Ambio 23(3):210-215. • ODUM, E.P. (1969): The strategy of ecosystem development. Science 164:262-270.

• ONG, C.K., M. V.NOORDWIJK, & G. CADISCH (EDS.)(2003): Below-ground interactions in tropical agro-ecosystems. Concepts and models with multiple plant components. CAB International, Wallingford, UK. • PADOCH, C., J.C. INUMA, W. DE JONG & J. UNRUH (1985): Amazonian agroforestry: A market-oriented system in Peru. In: Agroforestry Systems 3:47-58. • PAGEL, H., J. ENZMANN & H. MUTSCHER (1982): Pflanzennährstoffe in tropischen Böden - ihre Bestimmung und Bewertung. VEB Deutscher Landwirtschaftsverlag, Berlin, German Democratic Republic. • PALM, C.A., C.N. GACHENGO, R.J. DELVE, G. CADISCH & K.E. GILLER (2000): Organic inputs

• 212 • 11 References

for soil fertility management in tropical agroecosystems: Application of an organic resource database. Agriculture, Ecosystems and Environment 1692:1–16. • PANEM, C.C. (1992): Structures of the Greater Tongonan Area, Central Leyte, Philippines. PNOC Geothermal Division, Manila, Philippines. • PASSIOURA, J.B. (1996): Simulation models: Science, snake oil, education, or engineering? Agronomy Journal 88:690-694. • PAUSTIAN, K., E. LEVINE, W.M. POST & I.M. RYZHOVA (1997A): The use of models to integrate information and understanding of soil C at the regional scale. Geoderma 79:227-260. • PAUSTIAN, K., G.I. ÅGREN & E. BOSATTA (1997B): Modelling litter quality effects on decomposition and soil organic matter dynamics. In CADISCH, G. & K. GILLER (EDS.): Driven by nature. CAB International, Wallingford, UK. • PENA, N. (2004): LULUCF: Strategies for the future. 1Presentation 1st Agroforestry World Congress 2004, Florida, USA. • POLINAR, A.N. (2004):Constraints to on-farm tree domestication among upland farmers in Southern Leyte, Philippines. Presentation 1st Agroforestry World Congress 2004, Florida, USA. • POSEY, D.A. & W. BALÉE (EDS.)(1989): Resource management in Amazonia: Indigenous and folk Strategies. Advances in Economic Botany, Vol.7. The New York Botanical Garden, NY, USA. • POWLSON, D. (2005): Will soil amplify climate change? Nature 433:204f. • POWLSON, D.S., P. SMITH & J.U. SMITH (eds.)(1996): Evaluation of soil organic matter models. Springer, Berlin, Germany, in cooperation with NATO Scientific Affairs Division. • QUIMIO, J.M., E.A. HERMONES & L. TAN (1998): Vegetation structure of five-year rainforestation farm in ViSCA. Poster, International Conference on Applied Tropical Ecology, 8-10 September 1998, ViSCA, Baybay, Leyte, Philippines. • RAHAYU, N., M. V.NOORDWIJK & B. LUSIANA: Site-specific potential crop growth input parameters for WaNuLCAS, using WOFOST_Excel. Available online at www.worldagroforestrycentre.org/sea download Feb 11th, 2007. • RAICH, J.W., C.S. POTTER, & D. BHAGAWATI (2003): Interannual variability in global soil respiration on a 0.5 degree grid cell basis (1980-1994). NDP-081. Carbon Dioxide Information Analysis Center, U.S. Department of Energy, Oak Ridge National Laboratory, Oak Ridge, Tennessee, USA. • RAICH, J.W., A.E. RUSSELL & P.M. VITOUSEK (1997): Primary productivity and ecosystem development along an elevational gradient on Mauna Loa, Hawai´i. Ecology 78 (3): 707-721. • RANGIN, C., C. MÜLLER & H.PORTH (1989): Neogene geodynamic evolution of the Visayan Region. Geologisches Jahrbuch Vol. 70:7-27. Bundesanstalt für Bodenforschung, Hannover, Germany. • RAO, M.R., P.K.R. NAIR & C.K. ONG (1998): Biophysical interactions in tropical agroforestry systems. Agroforestry Systems 38:3-50. • REDDY, M.V. (ED.)(2002): Management of tropical plantation forests and their soil-litter system. Science Publishers, Enfield, UK. • REES, R.M., B.C. BALL, C.D. CAMPBELL & C.A. WATSON (2001): Sustainable management of soil organic matter. CAB International Wallingford, UK. • REHM, S. & G. ESPIG (1996): Die Kulturpflanzen der Tropen und Subtropen. Ulmer, Stuttgart, Germany. • RODRIGO, V.H.L., C.M. STIRLING, Z. TEKLEHAIMANOT & A. NUGAWELA (1997): The effect of planting density on growth and development of component crops in rubber / banana intercropping systems. Field Crops Research 52:95-108. • RUHIGWA, B.A., M.P. GICHURU, B. MAMBANF & N.M. TARIAH (1992): Root distribution of Acioa

• 213 • 11 References

barteri, Alchornea cordifolia, Cassia siamea and Gmelina arborea in an acid Ultisol. Agroforestry Systems 19: 67-78. • RUSSELL, A.E. (2002): Relationships between crop-species diversity and soil characteristics in Southwest Indian agroecosystems. Agriculture, Ecosystems and Environment 92:235-249. • SALES, R.F., R.D. LASCO & M.R.N. BANATICLA (2005): Carbon storage and sequestration potential of smallholder tree farms on Leyte Island, The Philippines. In: ACIAR smallholder forestry project ASEM 200/008 Redevelopment of a timber industry following extensive land clearing: Proceedings from the end-of project workshop, Ormoc City, Philippines 19-21 August 2004. ACIAR, Ormoc, Philippines. • SALICK, J. (1989): Ecological base of Amuesha agriculture, Peruvian Upper Amazon. In POSEY, D.A. & W.BALÉE (EDS.)(1989): Resource management in Amazonia: Indigenous and folk Strategies. Advances in Economic Botany, Vol.7. The New York Botanical Garden, NY, USA. • SANTOS, F. & S.G. RAMOS (1995): The geology of the Bato-Lunas Geothermal Prospect, Province of Leyte. Geoscientific Dept., Geothermal Division, PNOC, Makati, Philippines. • SCHACHTSCHABEL, P., H.-P. BLUME, G. BRÜMMER, K.-H. HARTGE & U. SCHWERTMANN (1992): Lehrbuch der Bodenkunde, 13. Auflage. Enke, Stuttgart, Germany. • SCHIMEL, D.S., J.I. HOUSE, K.A. HIBBARD, P. BOUSQUET, P. CIAIS, P. PEYLIN, B.H. BRASWELL, M.J. APPS, D. BAKER, A. BONDEAU, J. CANADELL, G. CHURKINA, W. CRAMER, A.S. DENNING, C.B. FIELD, P. FRIEDLINGSTEIN, C. GOODALE, M. HEIMANN, R.A. HOUGHTON, J.M. MELILLO, B. MOORE 3RD, D. MURDIYARSO, I. NOBLE, S.W. PACALA, I.C. PRENTICE, M.R. RAUPACH, P.J. RAYNER , R.J. SCHOLES, W.L. STEFFEN & C. WIRTH (2001): Recent patterns and mechanisms of carbon exchange by terrestrial ecosystems. Nature 414:169-172. • SCHINNER, F., E. KANDELER, R. ÖHLINGER & R. MARGESIN (1993): Bodenbiologische Arbeitsmethoden. 2nd edition. Springer, Berlin, Heidelberg, Germany. nd • SCHLICHTING, E., H.-P. BLUME & K. STAHR (1995): Bodenkundliches Praktikum. 2 ed. Blackwell, Berlin, Germany.

• SCHNEIDER V. DEIMLING, T., H. HELD, A. GANOPOLSKI & S. RAHMSTORF (2006): Climate sensitivity estimated from ensemble simulations of glacial climate. Climate Dynamics 27(2-3):149-163. • SCHOENAU, J.J. & J.R. BETTANY (1987): Organic matter leaching as a component of carbon, nitrogen, phosphorus and sulphur cycles in a forest, grassland, and gleyed soil. Soil Science Society of America Journal 51:646-651. • SCHROEDER, P. (1994): Carbon storage benefits of agroforestry systems. Agroforestry Systems 27:89-97. • SCHROTH, G. & F.L. SINCLAIR (EDS.)(2003): Trees, crops and soil fertility - Concepts and research methods. CABI Publishing, Wallingford, UK. • SCHROTH, G., J. LEHMANN, M.R.L. RODRIGUES, E. BARROS & J.L.V. MACÊDO (2001): Plant-soil interactions in multi-strata agroforestry in the Humid Tropics. Agroforestry Systems 53:85-102. • SCHROTH, G., U. KRAUSS, L. GASPAROTTO, J.A. DUARTE A., & K. VOHLAND (2000): Pests and diseases in agroforestry systems of the Humid Tropics. Agroforestry Systems 50:199-241. • SCHULTE, E.E. (1995): Recommended soil testing procedures for the Northeastern United States. In: AGRICULTURAL EXPERIMENT STATIONS OF CONNECTICUT, DELAWARE, MAINE, MARYLAND, MASSACHUSETTS, NEW HAMPSHIRE, NEW JERSEY, NEW YORK, PENNSYLVANIA, RHODE ISLAND, VERMONT AND WEST VIRGINIA (ED.), 2nd edition. Bulletin # 493. Agricultural Experiment Station, University of Delaware, Newark, USA. • SCHÜTT, P. (ED.)(2004): Bäume der Tropen. Nikol, Hamburg, Germany.

• 214 • 11 References

• SCHWENDENMANN, L.C. (2002): Below-ground carbon dynamics as a function of climate variability in undisturbed soils of a neotropical rain forest. PhD Dissertation University of Göttingen, Germany. • SHIVELY, G.E., C.A. ZELEK, D.J. MIDMORE & T.M. NISSEN (2004): Carbon sequestration in a tropical landscape: An economic model to measure its incremental cost. Agroforestry Systems 60:189-197. • SMILEY, G.L. (2006): Biophysical resource use and productivity in cocoa-gliricidia agroforests of central Sulawesi. PhD Dissertation University of Hohenheim. Margraf, Weikersheim, Germany. • SMITH, P., J.U. SMITH, D.S. POWLSON, W.B. MCGILL, J.R.M. ARAH, O.G. CHERTOV, K. COLEMAN, U. FRANKO, S. FROLKING, D.S. JENKINSON, L.S. JENSEN, R.H. KELLY, H. KLEIN- GUNNEWIEK, A.S. KOMAROV, C. LI, J.A.E. MOLINA, T. MUELLER, W.J. PARTON, J.H.M. THORNLEY & A.P. WHITMORE (1997): A comparison of the performance of nine soil organic matter models using datasets from seven long-term experiments. Geoderma 81:153-225. • SMITH, W.B. & G.J. BRAND (1983): Allometric biomass equations for 98 species of herbs, shrubs and small trees. USDA Forest Service, North Central Forest Experiment Station, Research Note NC-299. • SOERIANEGARA, I. & R.H.M.J. LEMMENS (EDS.)(1993): Plant resources of South East Asia, Vol.5,1: Major commercial timbers. PROSEA. Backhuys Publ., The Netherlands. • SOILS RESEARCH, TESTING AND PLANT ANALYSIS LABORATORY, DASS, LSU (not dated): Standard methods of analysis for soil and plant tissue samples. Unpublished. • SPEK, L.Y. & M. V.NOORDWIJK (1994): Proximal root diameters as predictors of total root system size for fractal branching models. II. Numerical Model. Plant and Soil 164:119-128. • STOVER, R.H. (1982): Valery and Grand Nain: Plant and foliage characteristics and a proposed banana ideotype. Tropical Agriculture 59:303-305. • SVENSSON, K. & M. PELL (2000): Soil microbial tests for discriminating between different cropping systems and fertiliser regimes. Biology & Fertility of Soils 33:91-99. • TREXLER, M.C. & C. HAUGEN (1991): Using tropical forestry to mitigate global warming: A preliminary carbon assessment. Draft final report to the US Environmental Protection Agency, Washington DC, USA. • TSEGAYE, A. & P.C. STRUIK (2003): Growth, radiation use efficiency and yield potential of enset (Ensete ventricosum) at different sites in southern Ethiopia. Annuals of Applied Biology 142:71-81. • TURNER, D.W. (1990): Modelling demand for nitrogen in the banana. Acta Horticulturae 275:497-503. • TURNER, D.W. (1972A): Banana plant growth 1. Gross morphology. Australian Journal of Experimental Agriculture and Animal Husbandry 12:209-215. • TURNER, D.W. (1972B): Banana plant growth: Dry matter production, leaf area growth and analysis. Australian Journal of Experimental Agriculture and Animal Husbandry 2:217-224. • UHL, C. & C.F. JORDAN (1984): Succession and nutrient dynamics following forest cutting and burning in Amazonia. Ecology 65(5):1476–1490. • UHL, C.(1987): Factors controlling succession following slash and burn agriculture in Amazonia. Journal of Ecology 75:377-407. • V. NOORDWIJK, M., G. CADISCH & C.K. ONG (2004): Below-ground interactions in tropical agroecosystems: Concepts and models with multiple plant components. CAB International, Wallingford, UK. • V. NOORDWIJK, M., B. LUSIANA & N. KHASANAH (2004): WaNuLCAS 3.0. Background on a model of water, nutrient and light capture in agroforestry systems. ICRAF, Bogor,

• 215 • 11 References

Indonesia. • V. NOORDWIJK, M. & R. MULIA (2002): Functional branch analysis as tool for fractal scaling above- and belowground trees for their additive and non-additive properties. Ecological Modelling 149(1-2):41-51. • V. NOORDWIJK, M., S. RAHAYU, K. HAIRIAH, Y.C. WULAN, A. FARIDA & B.VERBIST (2002): Carbon stock assessment for a forest-to-coffee conversion landscape in Sumber-Jaya (Lampung, Indonesia): From allometric equations to land use change analysis. Science in China, Series C, 45, Suppl. • V. NOORDWIJK, M. (2002): Scaling trade-offs between crop productivity, carbon stocks and biodiversity in shifting cultivation landscape mosaics: The FALLOW model. Ecological Modelling 149:113-126. • V. NOORDWIJK, M. (1999): Scale effects in crop-fallow rotations. Agroforestry Systems 47:239-251. • V. NOORDWIJK, M., C. CERRI, P.L. WOOMER, K. NUGROHO & M. BERNOUX (1997): Soil carbon dynamics in the humid tropical forest zone. Geoderma 79:187-225. • V. NOORDWIJK, M., L.Y. SPEK & P. DE WILLIGEN (1994): Proximal root diameters as predictors of total root system size for fractal branching models. I. Theory. Plant and Soil 164:107-118. • V.CARLOWITZ, H.C. (1713): Sylvicultura oeconomica, oder haußwirthliche Nachricht und naturmäßige Anweisung zur Wilden Baum Zucht. J.Fr. Braun, Leipzig, Germany. • VELDKAMP, E. (1994): Organic carbon turnover in three tropical soils under pasture after deforestation. Soil Science Society of America Journal 58:175-180. • VELICOGNA, I. & J. WAHR (2006): Acceleration of greenland ice mass loss in spring 2004. Nature 443:329-331. • VERCHOT, L.V., M. V.NOORDWIJK, T. TOMICH & S. KANDJI (2004): Climate Change: Linking adaptation and mitigation through agroforestry. Presentation, 1st Agroforestry World Congress, Florida, USA. • VERHEIJ, E.W.M. & R.E. CORONEL (ed.)(1991): Plant resources of South East Asia, Vol.2. Pudoc, Wageningen, The Netherlands. • WALDROP, M.P., T.C. BALSER & M.K. FIRESTONE (2000): Linking microbial community composition to function in a tropical soil. Soil Biology & Biochemistry 32(13):1837-1846. • WARDLE, D.A., G.W. YEATES, K.S. NICHOLSON, K.I. BONNER & R.N. WATSON (1999): Response of soil microbial biomass dynamics, activity and plant litter decomposition to agricultural intensification over a seven-year period. Soil Biology & Biochemistry 31:1707-1720. • WARDLE, D.A. & D. PARKINSON (1990): Interactions between microclimatic variables and the soil microbial biomass. Biology and Fertility of Soils 9:273-280. • WEIDELT, H.J. & V.S. BANAAG (1982): Aspects of management and silviculture of Philippine Dipterocarp forests. GtZ Eschborn, Germany. • WHITFORD, H.N. (1911): The forests of the Philippines. Part I: Forest types and products. Department of the Interior, Bureau of Forestry, Bulletin No. 10. Manila, Philippines. • WISCHMEIER, W.H. & D.D. SMITH (1978): Predicting rainfall erosion losses – a guide to conservatrion planning. Agriculture handbook 537. USDA, Washington, USA. • WORLD AGROFORESTRY CENTRE: Online Agroforestry Tree database. Online resource. Download from www.worldagroforestrycentre.org on 20/05/2007. • WORLD AGROFORESTRY CENTRE: Online Wood Density database. Online resource. Download from www.worldagroforestrycentre.org on 20/05/2007. • YADAV, R.S. & J.C. TARAFDAR (2001): Influence of organic and inorganic phosphorus supply on the maximum secretion of acid phosphatase by plants. Biology & Fertility of Soils 34:140-143.

• 216 • 11 References

• YAN, T., L. YANG & C.D. CAMPBELL (2003): Microbial biomass and metabolic quotient of soils under different land use in the Three Gorges Reservoir. Geoderma 115:129-138. • YANG, J., D.J. GREENWOOD, D.L. ROWELL, G.A. WADSWORTH & I.G. BURNS (2000): Statistical methods for evaluating a crop nitrogen simulation model, N_ABLE. Agricultural Systems 64(1):37-53. • YOUNG, A. (1997): Agroforestry for soil management. CAB International, Wallingford, UK. • YUMUL, G.P., C.B. DIMALANTA, R.A. TAMAYO & R.C. MAURY (2003): Collision, subduction and accretion events in the Philippines: A synthesis. Island Arc 12(2):77-91. • ZANG, S.X. & J.Y. NING (2002): Interaction between Philippine Sea Plate (PH) and Eurasia (EU) Plate and its influence on the movement of Eastern Asia. Chinese Journal of Geophysics 45(2):188-197. • ZECH, W., N. SENESI, G. GUGGENBERGER, K. KAISER, J. LEHMANN, T.M. MILANO, A. MILTNER & G. SCHROTH (1997): Factors controlling humification and mineralization of soil organic matter in the Tropics. Geoderma 79:117-161. • ZIKELI, S. (1998): Nutrient status and nutrient cycle of the tropical rain forest. Master Thesis University of Halle, Germany. • ZÖFEL, K. (2004): Regeneration of soil fertility through reforestation? Studies on P- availability on deforested and forested sites on Leyte, Philippines. M.Sc. Thesis (Diplomarbeit) University of Hohenheim. Unpublished.

• 217 12 Appendices Table 30: Overview soil chemical data paired plots chapter 4

PH C N Phosphatase P Ca2+ Mg2+ K+ Na+ BR 24h BR 72h C org T Bray mic -1 -1 µgCO g-1 h-1 -1 [%] µgPhenol g 3h mg/kg 2 µg g Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ci-Rainfo 4.67 0.22 3.50 0.33 0.29 0.04 37.05 17.10 0.52 0.33 415.8 58.1 412.0 103.6 101.82 8.06 62.15 31.41 6.46 1.31 4.39 1.15 977.2 246.6 Ci-Grass 4.97 0.27 3.51 0.67 0.30 0.04 113.41 20.78 0.43 0.12 529.2 144.0 532.7 218.9 115.96 33.77 172.12 158.10 7.79 1.66 3.84 1.24 988.3 309.0 LSU-Rainfo 4.96 0.15 1.98 0.25 0.22 0.03 23.60 6.13 0.46 0.03 3009.0 50.5 701.9 36.0 177.84 46.85 50.54 11.99 2.40 0.56 2.95 0.40 295.2 109.0 LSU-Annuals 5.05 0.27 2.23 0.25 0.25 0.03 24.59 7.58 0.51 0.10 3307.4 203.4 929.0 97.8 223.61 145.05 69.91 13.89 2.69 0.52 3.14 0.42 491.6 197.1 LSU-Grass nd nd 2.24 0.19 0.23 0.02 35.34 6.32 0.52 0.03 2156.2 45.7 1446.4 83.5 107.74 21.67 73.65 12.69 3.44 0.43 3.62 0.38 365.4 95.3 Mar-Rainfo 4.98 0.11 2.67 0.44 0.26 0.04 58.41 15.31 0.47 0.05 3046.0 491.4 782.1 36.3 110.65 48.95 153.62 76.82 3.23 1.14 3.25 0.36 393.8 154.9 Mar-Grass 5.36 0.29 2.76 0.46 0.30 0.05 32.15 nd 0.63 0.08 4314.7 195.9 1113.4 78.1 247.02 79.73 101.88 25.04 2.98 0.83 3.17 0.49 347.6 103.2 Mar-Gme 5.04 0.19 3.32 0.52 0.32 0.04 47.76 15.27 0.54 0.07 4352.2 9.2 1020.3 2.7 188.05 76.07 118.98 1.38 4.12 1.25 4.01 0.87 464.9 148.8 Pang-Rainfo 4.96 0.08 2.77 0.20 0.26 0.02 28.40 nd 0.38 0.06 920.6 141.2 469.3 34.9 95.95 20.66 40.03 7.54 2.74 1.34 2.45 1.08 372.0 156.4 Pang-Grass 4.75 0.35 2.28 0.25 0.21 0.02 31.88 nd 0.39 0.07 998.0 106.6 311.5 31.9 79.22 20.17 70.22 29.35 1.76 0.89 1.35 0.41 287.7 114.5 Pat-Rainfo 4.79 0.22 1.76 0.10 0.17 0.01 37.69 9.19 0.41 0.07 2001.0 9.1 1116.6 21.2 130.27 79.02 107.94 32.72 1.37 0.13 1.33 0.12 294.2 78.2 Pat-Grass 4.98 0.20 2.08 0.24 0.18 0.02 36.27 22.14 0.53 0.17 976.8 8.2 544.5 5.1 303.20 9.89 90.42 0.07 1.62 0.33 1.71 0.13 371.0 85.5 Pat-Gme 5.36 0.12 2.29 0.46 0.19 0.04 24.12 23.70 0.39 0.02 1533.4 13.8 877.3 3.8 304.72 50.19 90.58 4.72 4.13 1.08 3.78 1.01 381.0 200.7 Mai-Rainfo 6.61 0.13 2.68 0.26 0.23 0.02 22.00 14.99 34.72 4.93 5640.7 876.8 1021.4 89.2 136.14 21.99 70.30 9.40 4.32 0.60 2.82 0.42 564.4 162.3 Mai-Grass 6.66 0.12 2.21 0.34 0.20 0.04 27.39 26.86 17.94 0.11 6029.4 200.4 1262.2 119.6 108.78 24.83 75.76 7.49 5.05 1.54 3.03 0.68 601.1 83.8 Pun-Rainfo 8.02 0.18 4.17 0.30 0.36 0.03 1.69 0.53 2.73 0.23 11717.3 179.3 246.4 0.6 189.37 81.92 80.74 29.45 22.37 0.97 8.68 0.36 526.8 90.8 Pun-Grass 7.83 0.33 4.14 0.08 0.33 0.02 0.95 0.50 2.87 0.55 12492.4 801.3 197.4 8.6 88.37 34.21 60.06 20.15 22.65 0.91 8.53 0.27 441.2 55.6 Laboratory data chapter 5:

Table 31: Soil organic carbon at Cienda subplots, 0-5cm and 7-12cm depth

Carbon contents Loss on Ignition method, 0-5cm depth Carbon contents Loss on Ignition method, 7-12cm depth Subplot 11 [%] [%] 0-5cm 7-12cm Subplot 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 6.71 6.14 5.37 6.07 6.55 5.86 5.41 4.51 4.99 5.02 5.65 4.29 4.03 4.91 4.18 4.14 4.19 3.66 3.68 4.19 4.67 3.65 6.14 5.93 5.35 5.53 6.31 5.64 5.91 4.46 5.47 4.84 6.63 5.34 4.00 4.77 4.19 4.09 3.59 4.00 4.01 4.24 5.48 4.46 5.75 4.97 4.87 5.96 5.40 5.40 5.43 4.66 4.88 5.25 5.46 7.47 3.98 4.27 3.65 4.70 4.03 4.24 3.90 4.27 4.16 4.70 5.71 5.87 5.53 5.57 5.58 5.65 5.89 5.40 4.01 4.90 5.76 6.36 4.94 4.19 3.81 4.20 4.37 3.86 3.56 4.15 4.33 5.10 5.63 5.27 4.21 6.06 5.20 5.23 5.44 4.55 4.48 4.58 6.52 4.63 4.04 4.56 3.91 3.88 4.23 3.70 3.86 4.21 5.25 4.83 6.10 4.88 5.30 4.80 5.36 4.75 4.20 4.61 5.30 4.02 5.02 5.52 5.13 3.85 5.37 3.96 4.35 4.23 3.78 3.97 4.23 3.68 5.70 5.18 5.47 4.76 6.53 5.47 4.95 5.52 5.28 3.95 5.60 6.32 4.29 4.03 4.13 2.26 3.45 3.76 4.05 3.82 5.53 4.02 5.27 2.66 6.82 5.57 5.63 4.90 5.49 5.22 5.64 5.19 4.63 5.50 4.09 3.83 3.42 4.48 3.76 4.37 4.43 3.83 4.14 4.09 5.43 4.47 5.98 6.86 5.02 5.57 7.34 4.63 5.39 4.40 5.01 5.21 4.45 3.94 4.80 4.03 4.62 3.93 4.10 3.77 3.49 5.14 4.97 4.75 5.46 5.31 4.19 4.72 6.90 5.10 4.83 4.96 4.79 4.98 3.89 3.69 5.57 3.66 4.32 3.65 4.07 4.27 4.57 4.28 5.21 4.82 5.73 5.05 5.76 5.52 4.85 4.83 5.44 5.35 5.68 4.94 3.79 4.05 4.58 4.35 4.75 3.57 4.11 4.18 4.11 3.70 5.44 5.40 5.08 4.91 5.72 5.46 4.63 4.40 5.84 5.01 5.75 7.78 3.94 4.34 4.04 4.22 4.78 3.52 4.21 4.14 4.39 3.70 5.46 4.86 4.86 5.16 5.65 5.53 5.22 5.15 6.39 4.13 5.70 5.94 4.03 3.89 4.23 4.54 3.64 4.49 4.76 4.08 4.81 4.54 6.42 4.69 5.33 5.78 6.04 4.81 5.54 5.85 5.56 4.22 5.18 6.12 4.07 4.20 4.75 5.06 4.09 3.85 4.04 3.94 4.17 4.64 6.42 4.80 5.29 5.80 6.02 4.61 5.00 6.14 4.79 4.93 5.19 5.45 5.19 4.16 3.93 4.04 3.57 3.90 3.76 3.87 4.85 4.35 5.89 4.65 4.34 6.21 4.81 5.38 5.33 6.25 4.89 4.21 5.12 5.79 4.30 4.15 4.21 4.06 4.02 4.75 3.92 3.72 3.83 4.12 6.58 5.26 5.79 5.46 5.54 4.35 5.35 6.00 5.18 4.12 5.06 5.45 2.31 4.09 4.63 4.29 4.06 4.59 3.70 3.62 3.68 3.85 6.09 5.01 6.02 5.53 6.05 5.09 4.31 5.76 4.64 4.51 4.59 5.48 4.00 4.06 4.39 3.98 3.60 3.88 3.81 4.16 3.70 4.01 6.37 5.07 5.83 7.68 6.36 5.37 5.23 5.66 4.80 4.90 4.47 5.72 3.33 3.82 3.94 3.71 3.64 3.68 3.95 3.73 3.87 4.13 5.74 6.00 5.14 5.90 6.44 5.42 6.13 5.85 5.61 4.57 5.26 4.95 4.20 5.37 4.34 4.16 4.12 3.67 3.94 3.99 3.83 5.97 5.17 4.76 6.38 5.30 3.96 5.04 5.48 4.73 4.81 5.37 4.00 4.16 4.88 4.34 4.05 3.77 3.98 3.94 3.73 5.80 5.21 4.05 5.01 4.79 5.07 5.08 5.34 4.32 4.02 3.59 3.54 3.86 3.79 5.31 4.83 5.00 5.73 4.65 Average 5.36 5.70 5.74 5.24 5.30 5.21 5.17 4.68 5.30 5.63 4.09 4.21 4.33 4.10 4.04 3.94 3.98 4.01 4.34 4.20 6.82 4.94 StDev 0.59 0.70 0.61 0.40 0.84 0.59 0.52 0.43 0.56 0.82 0.60 0.40 0.53 0.52 0.40 0.36 0.26 0.21 0.58 0.47 5.62 5.36 4.93 4.53 4.80 5.17 5.63 4.57 5.38 4.61 5.76 4.38 5.78 4.99 6.31 4.95 4.78 5.53 4.66 4.26 4.79 4.46 Average 5.78 4.94 StDev 0.51 0.58 • 12 Appendices

Table 32: Organic matter fractions at Cienda subplots

Organic Matter Fractions obtained by sieving and flotation in water, Cienda subplots

Fraction Element / Aboveground Roots Entire LOM HOM Lost Labile Subplot biomass >2mm <2mm <2mm <2mm >2mm C g kg-1 soil [%] 1 0.73 2.08 22.07 0.8 16.37 22 18.1 2 1.73 2.13 21.78 0.9 17.32 16 21.5 3 0.95 2.28 19.84 0.75 17.06 10 18.9 4 0.89 2.22 20.99 1.3 16.3 16 21.3 5 0.93 2.47 27.46 1.25 21.54 17 17.8 6 0.7 1.34 21.01 0.58 14.84 27 15.0 7 0.53 2.11 20.7 0.67 16.06 19 17.1 8 0.59 0.47 18.63 0.72 15.83 11 10.1 9 0.47 2.45 19.95 0.84 16.17 15 18.9 10 1.12 2.25 22.05 0.96 18.13 13 19.3 11 0.72 2.67 22.92 0.96 18.04 17 19.5 12 0.56 2.4 19.84 0.68 18.44 4 16.5 13 0.67 2.76 26.12 0.9 22.02 12 16.4

N g kg-1 soil [%] 1 0.01 0.02 2.21 0.03 1.64 25 3.7 2 0.04 0.03 2.28 0.03 1.73 23 5.0 3 0.02 0.03 1.98 0.02 1.71 13 3.9 4 0.01 0.02 1.89 0.04 1.63 12 4.4 5 0.02 0.03 2.43 0.04 1.91 20 4.8 6 0.01 0.02 1.99 0.02 1.39 29 3.4 7 0.01 0.03 1.97 0.02 1.52 22 3.6 8 0.02 0.01 1.97 0.03 1.58 18 3.2 9 0.01 0.03 1.79 0.03 1.54 12 4.5 10 0.02 0.03 2.21 0.03 1.73 20 4.6 11 0.02 0.03 1.88 0.03 1.56 15 4.9 12 0.01 0.03 1.88 0.02 1.68 9 3.5 13 0.01 0.03 4.91 0.03 1.71 64 4.0

P mg kg-1 soil [%] 1 0.88 1.53 409 1.97 327.31 19 1.3 2 1.81 1.55 422 1.82 330.78 21 1.5 3 1.03 2.26 412 1.76 364.69 11 1.4 4 0.87 1.41 391 2.98 322.69 17 1.6 5 1.14 1.98 406 2.52 376.91 7 1.5 6 0.95 1.37 399 0.95 350.51 12 0.9 7 0.48 1.7 392 1.32 338.05 13 1.0 8 0.94 0.51 380 1.77 303.33 20 1.0 9 0.4 1.76 371 1.41 301.56 18 1.2 10 0.92 1.73 416 2.28 484.56 -17 1.0 11 0.79 2.6 559 1.71 270.55 51 1.9 12 0.61 2.17 490 1.36 417.68 14 1.0 13 1.43 2 507 2.17 446.93 11 1.2

LOM = light organic matter (<2mm, floating in water) HOM = heavy organic matter (<2mm, not floating in water) Labile = share of C / N / P in LOM to such in the entire fraction Lost = difference of entire fraction minus LOM and HOM as share of the entire fraction: (Entire<2 – LOM<2 - HOM<2) / Entire<2 x 100

• 220 • 12 Appendices

Table 33: Substrate-induced respiration and Cmic at Cienda subplots Substrate-Induced Respiration at 32° C and microbial carbon Subplot 1 2 3 4 5 6 7 8 9 10 11 Soil dry matter [%] 55.19 53.69 56.63 57.57 58.65 78.12 61.95 64.62 51.88 58.38 64.09 4.65 5.46 4.86 5.10 5.63 3.52 3.26 2.55 6.36 3.77 3.72 4.98 5.81 5.18 5.10 5.78 3.52 3.55 2.55 6.54 3.77 3.86 CO [mg 100g-1 h-1] 4.98 5.46 5.18 5.10 5.31 3.29 3.55 2.55 6.36 3.77 3.72 2 4.98 5.46 5.02 6.05 5.31 3.52 3.85 2.55 6.36 3.61 4.00 4.82 5.81 5.18 5.41 5.78 3.52 3.55 2.55 6.71 3.45 4.15 Average 4.88 5.60 5.08 5.35 5.56 3.47 3.55 2.55 6.47 3.67 3.89 StDev 0.15 0.19 0.14 0.42 0.24 0.10 0.21 0.00 0.16 0.14 0.19 Coeff. Var. [%] 3.0 3.3 2.8 7.8 4.3 3.0 5.9 0.0 2.4 3.8 4.8 CO [µg g-1 h-1] 2 48.84 56.01 50.82 53.50 55.64 34.73 35.51 25.53 64.67 36.74 38.90 C [µg g-1] mic 1.61 1.85 1.68 1.76 1.83 1.14 1.17 0.84 2.13 1.21 1.28

Table 34: Basal respiration at Cienda subplots in a 30 day incubation experiment (non-cumulative depiction) Successive1 basal respiration over 30 days [µg CO g-1 oven-dry soil h-1] 2 1d 3d 6d 12d 18d 24d 30d Subplot Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD Ø SD 1 3.51 0.24 2.55 0.36 1.78 0.17 1.69 0.53 1.12 0.37 1.12 0.25 1.02 0.27 2 4.16 0.13 3.26 0.24 2.18 0.12 1.69 0.24 1.48 0.26 1.12 0.27 1.02 0.08 3 4.27 0.51 3.91 0.37 2.13 0.25 2.10 0.19 1.58 0.09 1.50 0.17 1.54 0.24 4 3.72 0.38 3.71 0.49 1.53 0.07 1.61 0.14 1.39 0.21 1.16 0.19 1.19 0.17 5 3.01 0.29 3.86 0.19 2.14 0.24 1.79 0.15 1.69 0.38 1.35 0.11 1.30 0.08 6 1.25 0.81 2.20 0.23 1.32 0.14 0.96 0.12 0.94 0.09 0.86 0.15 0.79 0.09 7 1.13 0.27 2.12 0.52 0.77 0.06 0.96 0.11 0.92 0.21 0.83 0.29 0.79 0.11 8 1.61 0.30 2.43 0.17 1.20 0.08 1.14 0.13 1.33 0.37 0.91 0.14 0.88 0.07 9 2.63 0.34 2.82 0.17 1.50 0.14 1.33 0.06 1.45 0.31 1.31 0.26 1.28 0.13 10 2.26 0.41 2.83 0.20 1.18 0.13 1.22 0.13 1.50 0.46 1.09 0.10 1.27 0.31 11 4.48 0.51 3.04 0.19 1.45 0.13 1.53 0.04 2.01 0.60 1.57 0.38 1.39 0.19 12 3.89 0.72 2.63 0.35 1.43 0.16 1.22 0.07 1.23 0.42 1.03 0.09 1.04 0.06 13 5.17 0.50 3.65 0.21 2.00 0.37 1.74 0.29 1.58 0.19 1.44 0.18 1.36 0.13 1 = not cumulative Ø = average, SD = st.dev. Table 35: Root weight and root length density at Cienda

Root weight density [mg cm-3] and root length density [cm cm-3] Cienda subplots

Weight Total RLD 0-15cm depth 15-30cm depth Ø Ø Total Ø Ø Total 0-15cm 15-30cm Subplot <1.5mm >1.5mm <1.5mm >1.5mm depth depth 1 1.99 2.79 4.77 0.90 1.81 2.71 1.31 1.57 2 2.25 0.80 3.05 0.99 0.20 1.19 0.94 0.59 3 1.42 2.42 3.84 1.16 2.65 3.81 2.82 0.74 4 7.16 9.02 16.18 3.58 11.94 15.52 3.42 0.48 5 5.44 6.76 12.20 1.63 1.99 3.62 1.68 0.54 6 2.92 1.06 3.98 0.99 0.99 1.99 1.78 0.66 7 nd nd nd 0.57 1.42 1.99 nd 0.27 8 1.99 0.27 2.25 0.72 0.00 0.72 0.65 0.47 9 2.92 2.92 5.84 0.83 3.32 4.14 1.47 0.51 10 4.74 7.65 12.40 0.80 0.27 1.06 1.33 0.35 11 6.90 5.31 12.20 2.75 1.07 3.83 2.26 0.63 12 2.82 4.48 7.29 0.61 0.15 0.77 2.38 0.46

• 221 • 12 Appendices

Table 36: Photosynthetically Active Radiation (PAR) at Cienda

Photosynthetically Active Radiation Cienda subplots [% of open area]

Row / X W V U T S R Q P O N M L K J I H G F E D C B A Line* 0 100 85 100 100 8 100 100 15 100 100 1 100 100 100 87 100 25 100 8 100 100 99 100 100 42 11 19 2 16 3 75 3 31 4 88 2 100 100 100 100 24 100 100 100 100 8 20 52 54 57 6 58 10 2 3 6 46 100 3 2 3 100 100 100 9 62 100 20 100 5 100 57 100 33 18 96 9 3 52 97 2 97 3 3 1 4 100 100 100 34 100 100 100 100 100 11 57 96 100 100 8 100 19 7 4 2 15 44 8 2 5 100 100 100 5 89 9 11 6 100 100 38 100 4 3 100 2 3 96 100 96 3 57 8 6 24 97 85 6 40 100 100 100 100 7 7 100 91 8 19 61 26 98 2 94 58 40 3 7 72 83 48 99 83 9 100 8 96 100 100 100 49 9 3 28 12 3 30 4 2 92 3 2 8 21 67 27 96 100 20 100 4 36 31 100 51 72 4 6 100 70 15 15 11 17 6 24 2 9 19 28 39 9 100 14 8 100 100 38 87 100 7 2 2 100 6 60 69 2 22 5 4 2 10 25 32 38 98 14 6 100 100 100 9 100 39 13 55 94 18 95 4 100 2 3 2 3 1 11 61 56 55 100 7 6 5 100 4 2 100 4 3 8 4 96 19 3 2 4 92 3 19 3 12 47 42 42 13 10 8 17 4 8 4 60 100 12 100 19 46 13 3 53 68 6 5 2 13 32 44 57 95 99 100 100 6 4 20 4 3 2 92 98 8 8 43 5 25 3 3 2 14 43 52 100 24 7 100 100 100 5 100 4 3 96 81 98 6 100 17 6 3 1 40 15 40 69 100 66 10 3 100 22 59 100 19 75 100 5 99 7 99 4 2 2 2 16 28 33 100 89 100 67 4 70 80 56 50 100 100 3 97 11 3 2 3 7 2 17 50 16 6 100 100 100 100 3 6 15 37 19 7 20 1 5 2 58 2 2 18 20 83 60 31 7 81 5 8 3 1 8 73 13 93 8 39 92 11 6 19 77 18 12 40 99 7 5 24 5 2 2 99 13 1 34 10 3 93 2 20 69 24 93 23 12 6 2 80 3 5 10 43 11 1 2 2 3 84 1 * see planting scheme chapter 2.3.2

Table 37: Leaf litter production at Cienda

Leaf litter production Cienda subplots 2005 [gDM m-2]

Jan 29- Mar 08- Apr 14- Jun 01- Jun 28- Aug 01- Sep 01- Oct 01- Nov 01- Dec 01- Subplot Mar 08 Apr 14 Jun 01 Jun 28 Aug 01 Sep01 Oct 01 Nov 01 Dec 01 Jan 01 1 8.39 4.30 8.25 4.53 3.53 3.83 4.71 8.67 7.40 4.05 2 19.75 4.04 2.40 1.23 3.78 1.52 1.24 0.95 0.85 2.23 3 7.50 21.97 5.58 1.42 0.57 1.07 2.24 2.64 3.78 2.28 4 1.94 17.83 4.70 3.03 9.16 3.04 2.24 8.53 2.49 1.31 5 9.29 5.05 17.15 6.12 2.65 2.32 3.59 4.67 3.23 6.77 6 1.08 0.91 1.47 0.81 0.15 0.31 0.27 0.62 0.35 0.48 7 0.46 0.02 0.04 0.03 0.63 3.68 0.40 0.07 0.01 3.90 8 16.75 3.77 1.69 0.89 2.09 1.53 2.42 1.89 1.25 2.16 9 0.21 0.07 8.74 0.02 0.11 0.01 0.93 0.04 0.00 1.63 10 23.12 1.95 2.59 0.64 2.67 1.47 2.29 1.68 0.90 0.72

• 222 • 12 Appendices

Table 38: Leaf litter decomposition at Cienda

Litter decomposition Cienda subplots [% mass loss]

Subplot 1 2 3 6 8 Slope 11 Mesh [mm] 0.1 4 0.1 4 0.1 4 0.1 4 0.1 4 0.1 4 0.1 4 48.0 55.1 49.3 73.1 53.1 38.4 58.2 100.0 61.0 47.8 42.5 36.9 60.7 46.7 54.8 43.8 47.7 78.3 55.3 53.8 67.7 82.2 56.3 54.4 35.4 37.1 46.1 54.1 61.0 50.7 44.1 70.0 56.1 55.1 64.1 100.0 55.9 54.2 39.3 34.0 56.3 44.4 66.1 50.8 53.3 88.7 55.1 51.1 50.5 83.3 57.9 45.9 34.1 40.7 45.1 41.5 60.1 58.1 49.2 77.2 53.7 55.3 61.5 85.8 61.1 44.5 44.8 43.1 49.9 61.4 53.1 53.3 53.5 78.9 47.8 50.6 57.6 72.5 45.5 55.4 38.7 31.7 42.2 63.1 62.6 54.5 61.3 56.2 54.3 45.6 65.7 62.6 51.3 50.7 40.7 54.5 49.7 11.3 55.8 54.5 60.1 57.1 51.7 45.3 62.7 68.3 50.6 58.1 35.4 45.9 52.1 43.3 52.7 53.5 57.5 57.2 48.7 46.9 58.0 65.1 53.5 53.6 38.2 39.9 52.6 27.1 Replicates 57.6 51.2 53.3 56.7 47.2 49.5 64.7 64.3 42.3 45.6 35.1 36.0 50.5 65.8 63.1 48.9 58.4 71.5 58.5 55.1 63.4 62.7 55.6 56.8 40.3 41.5 52.1 44.9 45.5 57.9 58.3 78.5 53.5 48.9 57.7 69.2 59.0 54.9 42.5 39.4 55.7 54.3 63.8 47.3 62.9 53.5 49.4 39.0 67.3 65.6 40.0 52.5 41.3 36.1 50.9 35.5 57.1 97.8 58.8 64.2 59.5 46.6 66.2 100.0 47.7 60.0 39.1 40.7 55.5 59.3 58.7 95.7 52.5 57.3 54.5 19.7 67.1 89.3 63.5 62.8 34.8 42.5 54.4 72.3 65.5 64.1 57.2 55.5 60.6 49.6 71.3 100.0 60.5 53.9 40.5 34.3 50.0 56.1 59.7 68.3 52.7 47.1 58.2 57.1 71.3 100.0 62.3 47.6 38.1 40.4 52.3 37.3 60.2 95.9 54.9 56.1 53.1 53.1 66.6 42.9 60.5 48.1 39.0 68.8 55.8 72.1 Average 58.1 61.2 54.7 65.4 55.4 64.0 61.3 81.3 47.5 46.4 38.9 41.3 51.8 49.5 St.Dev. 5.7 17.2 5.0 11.7 57.9 47.9 48.5 73.6 62.1 49.8 3.0 8.6 4.4 15.8 Coeff.Var. 9.8 28.1 9.2 18.0 58.3 56.5 63.1 65.8 58.4 60.3 7.7 20.8 8.6 32.0 56.8 50.6 54.9 58.3 58.9 44.3 59.5 45.9 63.4 72.7 43.5 42.1 48.0 48.0 56.7 94.9 46.3 40.7 58.5 56.1 64.1 62.8 53.7 43.8 57.1 56.8 66.1 60.3 58.3 47.0 45.6 56.5 58.8 46.2 50.8 48.0 49.9 42.1 61.7 47.3 53.0 44.3 47.7 54.5 60.7 49.7 50.4 53.3 52.1 44.2 56.5 58.3 65.9 58.1 58.3 53.3 65.1 85.7 53.2 51.5 50.1 54.6 56.5 94.4 62.9 44.8 52.6 63.1 60.9 100.0 50.5 49.7 nd 43.2 57.2 87.2 50.3 53.3 nd 42.7 65.5 69.8 45.5 52.8 nd nd 62.5 98.2 47.6 48.7 Average 53.9 49.7 61.8 75.6 54.0 50.7 St.Dev. 4.2 8.0 5.1 17.6 6.7 5.5 Coeff.Var. 7.8 16.2 8.3 23.2 12.5 10.9

• 223 Table 39: Correlations between environmental parameters, Cienda subplots

PAR Corg LoI Corg LoI Corg EA Cmic Cmic / BR 1d BR 3d BR 6d BR 12d BR 18d BR 24d BR 30d qCO2 Soil pH CLOM NLOM PLOM Root Root Leaf RWD RWD RWD resp. dec. dec. litter 0-15 0-15 15-30 Pearson 0-5cm 7-12cm 32° C Corg 24d correlations 0.1mm 4mm prod. fine thick fine

Corg LoI 0-5cm -0.366

Corg LoI 7-12cm -0.290 .705*

Corg EA -0.364 .960** .875**

Cmic 32° C -0.361 0.520 .650* 0.609

Cmic/Corg -0.316 0.317 0.454 0.393 .967** BR 1d -.788** .643* .708* .716* .718* 0.601 BR 3d -.831** 0.491 0.466 0.517 .632* 0.578 .799** BR 6d -.674* 0.545 0.579 0.596 .675* 0.585 .850** .766** BR 12d -.816** 0.568 0.551 0.603 .651* 0.563 .932** .861** .896** BR 18d -.702* 0.400 0.588 0.505 0.526 0.449 0.625 .833** .654* .681* BR 24d -0.607 0.562 0.611 0.624 .760* .686* .767** .844** .760* .861** .805** BR 30d -0.613 0.590 0.572 0.630 0.607 0.517 .668* .817** 0.580 .748* .831** .944**

qCO2 24d -0.068 -0.345 -0.271 -0.339 -.709* -.726* -0.302 -0.105 -0.243 -0.135 0.126 -0.124 0.037 Soil respiration -0.191 0.542 0.252 0.465 0.242 0.141 0.321 0.487 0.488 0.339 0.201 0.252 0.199 -0.246 pH 0.292 -.777* -0.139 -0.567 -0.446 -0.418 -0.472 -0.358 -0.349 -0.421 0.205 -0.166 -0.115 .817* -0.557

CLOM -0.424 .802** 0.613 .788** .744* .636* .683* 0.551 0.439 0.514 0.350 0.502 0.502 -.701* 0.398 -.805*

NLOM -0.480 0.616 .634* .669* .721* .639* 0.564 0.628 0.528 0.452 .704* 0.544 0.545 -0.499 0.336 -0.442 .775**

PLOM -.799** 0.376 0.455 0.433 .714* .689* .877** .820** .772** .823** 0.568 0.613 0.476 -0.435 0.346 -0.508 .635* 0.607 Root dec. 0.1mm 0.299 -.829** -0.390 -.722* -0.121 0.058 -0.409 -0.275 -0.148 -0.352 -0.138 -0.325 -0.481 0.039 -0.304 .977** -0.609 -0.241 -0.096 Root dec. 4mm 0.160 -0.494 -0.270 -0.442 -.648* -.640* -0.233 -0.479 -0.410 -0.349 -0.512 -0.619 -0.583 0.422 -0.319 0.420 -0.500 -.683* -0.256 0.178 Leaf litter prod. -.960** 0.252 0.149 0.230 0.306 0.297 .717* .713* .645* .793** 0.597 0.558 0.529 0.091 0.023 -0.240 0.299 0.347 .733* -0.214 -0.142 RWD 0-15, fine -0.630 0.242 0.005 0.170 0.327 0.371 0.328 0.582 0.161 0.310 0.518 0.413 0.562 -0.109 0.016 -0.412 0.490 0.554 0.414 -0.299 -0.417 0.565 RWD 0-15, th. -0.556 0.050 -0.055 0.014 0.302 0.380 0.257 0.534 0.197 0.202 0.472 0.278 0.372 -0.149 0.106 -0.279 0.355 0.526 0.436 -0.005 -0.299 0.467 .911** RWD 15-30, fine -0.389 -0.032 -0.031 -0.033 0.401 0.492 0.362 0.464 0.056 0.263 0.117 0.257 0.278 -0.355 0.058 -0.398 0.444 0.188 0.546 -0.061 0.003 0.302 0.609 .632* RWD 15-30, th. -0.588 0.003 -0.065 -0.023 0.350 0.433 0.428 0.604 0.226 0.355 0.262 0.264 0.282 -0.275 0.232 -0.421 0.426 0.302 .662* -0.030 -0.032 0.486 .705* .788** .928** * = significant at  = 0.05 ** = significant at  = 0.01 • 12 Appendices

Table 40: Mann-Whitney tests abaca mean biomass per subplot. * Different at α = 0.05; ** different at α = 0.01 Inventory 240704 S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S1 0.006** 0.004** 0.267 0.287 0.000** 0.018* 0.108 0.068 0.000** S2 0.006** 0.000** 0.002** 0.076 0.000** 0.000** 0.094 0.000** 0.853 S3 0.004** 0.000** 0.100 0.001** 0.156 0.965 0.000** 0.346 0.000** S4 0.267 0.002** 0.100 0.055 0.008** 0.180 0.008** 0.539 0.000** S5 0.287 0.076 0.001** 0.055 0.000** 0.002** 0.769 0.012* 0.033* S6 0.000** 0.000** 0.156 0.008** 0.000** 0.237 0.000** 0.032* 0.000** S7 0.018* 0.000** 0.965 0.180 0.002** 0.237 0.000** 0.410 0.000** S8 0.108 0.094 0.000** 0.008** 0.769 0.000** 0.000** 0.001** 0.016* S9 0.068 0.000** 0.346 0.539 0.012* 0.032* 0.410 0.001** 0.000** S10 0.000** 0.853 0.000** 0.000** 0.033* 0.000** 0.000** 0.016* 0.000**

Inventory 050505 S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S1 0.008** 0.569 0.813 0.135 0.090 0.842 0.000** 0.716 0.000** S2 0.008** 0.019* 0.017* 0.151 0.000** 0.041* 0.488 0.008** 0.003** S3 0.569 0.019* 0.779 0.311 0.021* 0.871 0.000** 0.406 0.000** S4 0.813 0.017* 0.779 0.221 0.033* 0.921 0.000** 0.549 0.000** S5 0.135 0.151 0.311 0.221 0.009** 0.342 0.010* 0.113 0.000** S6 0.090 0.000** 0.021* 0.033* 0.009** 0.104 0.000** 0.291 0.000** S7 0.842 0.041* 0.871 0.921 0.342 0.104 0.001** 0.568 0.000** S8 0.000** 0.488 0.000** 0.000** 0.010* 0.000** 0.001** 0.000** 0.000** S9 0.716 0.008** 0.406 0.549 0.113 0.291 0.568 0.000** 0.000** S10 0.000** 0.003** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000**

Inventory 040705 S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S1 0.334 0.121 0.023* 0.611 0.018* 0.465 0.000** 0.330 0.000** S2 0.334 0.902 0.512 0.203 0.003** 0.738 0.054 0.090 0.000** S3 0.121 0.902 0.431 0.061 0.001** 0.547 0.002** 0.018* 0.000** S4 0.023* 0.512 0.431 0.012* 0.000** 0.218 0.025* 0.003** 0.000** S5 0.611 0.203 0.061 0.012* 0.057 0.273 0.000** 0.644 0.000** S6 0.018* 0.003** 0.001** 0.000** 0.057 0.005** 0.000** 0.126 0.000** S7 0.465 0.738 0.547 0.218 0.273 0.005** 0.004** 0.125 0.000** S8 0.000** 0.054 0.002** 0.025* 0.000** 0.000** 0.004** 0.000** 0.000** S9 0.330 0.090 0.018* 0.003** 0.644 0.126 0.125 0.000** 0.000** S10 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000**

Inventory 300406 S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S1 0.374 0.103 0.167 0.438 0.010* 0.618 0.006** 0.029* 0.000** S2 0.374 0.764 0.905 0.152 0.002** 0.237 0.260 0.006** 0.001** S3 0.103 0.764 0.777 0.036* 0.001** 0.081 0.158 0.001** 0.000** S4 0.167 0.905 0.777 0.057 0.001** 0.119 0.124 0.002** 0.000** S5 0.438 0.152 0.036* 0.057 0.046* 0.814 0.003** 0.148 0.000** S6 0.010* 0.002** 0.001** 0.001** 0.046* 0.033* 0.000** 0.340 0.000** S7 0.618 0.237 0.081 0.119 0.814 0.033* 0.009** 0.107 0.000** S8 0.006** 0.260 0.158 0.124 0.003** 0.000** 0.009** 0.000** 0.000** S9 0.029* 0.006** 0.001** 0.002** 0.148 0.340 0.107 0.000** 0.000** S10 0.000** 0.001** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000** 0.000**

• 225 • 12 Appendices

Table 41: Leaf area equations for tree species

Species Leaf area equation

Dipterocarpus validus A = 0.7965 l w – 0.0001 (l w)2 Shorea contorta A = 0.6458 l w + 0.0018 (l w)2 Shorea palosapis A = 0.69 l w + 0.0002 (l w)2 Toona calantas A = 0.45 l w Durio zibethinus A = -0.0357 + 0.735 l w Garcinia mangostana A = 0.616 l w + 0.0004 (l w)2 Lansium domesticum A = 13.84 l w + 0.18 (l w)2 Artocarpus heterophyllus A = 0.6894 l w – 0.0006 (l w)2 Artocarpus odoratissimus A = 0.6424 l w Gmelina arborea A = 0.6745 l w

Table 42: Wood density (0% water contents) and carbon contents of different plant tissues for planted tree species, Gmelina and abaca.

Tree Species Wood density [g cm-3] Carbon contents [%] Twig Branch Stem Root Leaf Twig Branch Stem Root Dipterocarpus validus 0.50 0.58 0.65103 0.51 45.1 51.3 50.3 51.8 49.3 Shorea contorta 0.54 0.54 0.58 0.52 43.7 47.3 53.7 58.8 50.3 Shorea palosapis 0.43 0.49 0.43 n.d. 51.2 nd 53.2 56.1 44.2 Toona calantas 0.45 0.51 n.d. 0.54 45.1 47.8 51.3 nd 45.3 Nephelium lappaceum 0.64 0.68 0.76 0.81 48.8 49.8 56.3 56.8 55.3 Durio zibethinus 0.43 0.52 n.d. 0.45 44.6 50.8 52.7 51.7 51.8 Garcinia mangostana 0.64 0.68 n.d. n.d. 43.5 nd nd nd nd Lansium domesticum 0.68 0.70 0.70 0.70 41.8 50.3 51.7 52.3 47.8 Artocarpus heterophyllus 0.49 0.51 0.56 0.53 40.4 49.3 48.3 47.9 49.3 Artocarpus odoratissimus 0.44 0.52 0.53 0.43 36.2 47.8 49.8 53.8 53.8 Gmelina arborea 0.39 0.44 38.1 50.7 51.8 47.4 53.8 Musa textilis - - - 41.5 - - 46.0104 44.6 Cocos nucifera - 45.1 48.3105 - nd 55.8

Average C content of three composite undergrowth samples was 42.8%. Leaf C contents for D. validus are close to the 44.96%C analysed by GÖLTENBOTH (1998).

103Estimated on the basis of twig and branch density and PHILIPPINE COUNCIL FOR AGRICULTURE AND RESOURCES RESEARCH (1977): The Philippines recommends for Dipterocarps. I. Lumber. Los Banos. 104Pseudostem 105Stalk

• 226 • 12 Appendices

Table 43: Total extractable polyphenolics (TEP) contents of fresh leaves and fine roots

TEP [%] Leaves Fine roots D. validus 6.65 8.46 S. contorta 9.75 13.1 S. palosapis 6.33 n.d. T.calantas 9.24 n.d. G. arborea 1.56 2.14 N. lappaceum 9.80 9.17 G. mangostana 5.07 n.d. D. zibethinus 1.36 7.31 A. heterophyllus 3.70 n.d. A. odoratissima 6.48 5.74 L. domesticum 3.20 4.14 M. textilis 1.09 2.51 C. nucifera 9.26 2.56

Table 44: Leaf weight ratio of planted species calculated as dry weight leaves per dry weight shoot

Leaf weight ratio

D. validus 0.31 S. contorta 0.30 G. arborea 0.16 A. heterophyllus 0.28 A. odoratissima 0.43 N. lappaceum 0.30 L. domesticum 0.17 M. textilis 103cm 0.11 M. textilis 184cm 0.42 M. textilis 318cm 0.43

• 227 • 12 Appendices

Table 45: Alternative version of the multiple regression expressing C, N und PLOM as % of entire sample (see text for explanation).

Dependent variable Equation r2 P

th 106 Biomass July 24 , 2004 y = 2.4863 + 5.6794NLOM – 3.2352Cmic/Corg 0.6696 0.0207

th Biomass May 5 , 2005 y = 385.1339 – 2.8542PAR + 37.9541NLOM - 300.5528PLOM 0.8913 0.0027

th Biomass July 4 , 2005 y = 448.7870 – 3.4329PAR + 43.0689NLOM - 362.5662PLOM 0.8860 0.0031

th April 30 , 2006 y = 1819.8258 – 13.6734PAR + 160.2172NLOM - 1452.4589PLOM 0.8943 0.0025

Growth inventory 1 – 2 y = 358.5835 – 2.6681PAR + 34.7497NLOM - 286.5093PLOM 0.8997 0.0021

Growth inventory 2 – 3 y = 66.8429 – 0.6700PAR + 2.6922CLOM – 83.0886PLOM 0.8029 0.0155

106Intercept not significant at α = 0.05 (P = 0.62)

• 228 • 12 Appendices

Table 46: Example of plant calibration in WaNuLCAS (for site PN3; see model runs on enclosed CD for model settings of more species and at other sites). Relevant values applied for tree calibration (Oil palm, rubber, N fixation and pest impact set zero; root type = 0)

Parameters Units S. contorta D. validus G. arborea D. zibethinus A. heterophyllus

Length of vegetative cycle days 6205 4380 1460 2920 1460

Length of generative cycle days 180 180 30 100 135

Earliest day to flower in a year Julian day 90 90 60 150 30

Latest day to flower in a year Julian day 150 150 120 180 360

Initial stage [] 0.058824 0.04 0.05 0.125 0.25

Stage after pruning [] 0.29411765 0.416667 0.25 0.625 0.25

Max. growth rate kg m-2 0.085 0.075 0.0356 0.0089 0.02

Fraction of growth reserve [] 0.05 0.05 0.05 0.05 0.05

Leaf weight ratio [] 0.6062 0.8573 0.4950 0.6641 0.7

Specific leaf area m2 kg-1 18.9196 20.5217 17.3 10.6696 9.4423 Water requirement for dry matter l kg-1 400 300 367.4235 300 212.1320 production Fraction biomass allocatedn to fruit [] 0 0 0.01 0.1 0.1 Max. canopy height above bare m 15 12 15 18 25 stem Ratio between canopy width and [] 0.5 0.5 0.4 0.333 0.3 height Max. canopy radius m 7.5 6 6 6 7.5

Maximum leaf area index [] 2.6667 4 2.6667 2.6667 4

Ratio leaf area index min. and max. [] 1 1 0.5 1 0.9

Relative light intensity at which [] 0.35 0.3 0.1 0.2 0.15 shading starts to affect tree growth

Extinction light coefficient [] 0.7826 0.7 0.4950 0.7342 0.8573

Rainfall water stored at leaf surface mm 0.25 1 1 1 0.25 Coefficient related to tree root cm day-1 1.0E-05 1.0E-05 1.0E-05 1.0E-05 1.0E-05 conductivity Plant potential for max. transpiration cm -500 -1000 -500 -500 -1500

Plant potential for min. transpiration cm -20000 -30000 -3000 -1500 -4500 N concentration in carbohydrate g g-1 0.22 0.22 0.11 0.22 0.33 reserves N concentration in leaf component g g-1 0.03125 0.025 0.0125 0.0275 0.0375

N concentration in twig component g g-1 0.015 0.015 0.0075 0.015 0.0225

N concentration in wood component g g-1 0.01 0.01 0.005 0.01 0.015

N concentration in fruit component g g-1 0.0075 0.015 0.015 0.0225 0.0225

N concentration in root component g g-1 0.01 0.01 0.01 0.01 0.01 P concentration in carbohydrate g g-1 0.022 0.022 0.011 0.022 0.0495 reserves P concentration in leaf component g g-1 0.003125 0.0025 0.00125 0.00275 0.005625

P concentration in twig component g g-1 0.0015 0.0015 0.00075 0.0015 0.003375

P concentration in wood component g g-1 0.001 0.001 0.0005 0.001 0.00225

P concentration in fruit component g g-1 0.00075 0.0015 0.0015 0.00225 0.003375

P concentration in root component g g-1 0.001 0.001 0.001 0.001 0.0015

• 229 • 12 Appendices

Parameters Units S. contorta D. validus G. arborea D. zibethinus A. heterophyllus Litterfall caused by drought day-1 0.1 0.1 0.1 0.1 0.03 Treeshold value for litterfall due to [] 0.9 0.7 0.5 0.7 0.9 drought Reducing factor for N of litterfall [] 0.85 0.85 0.85 0.85 0.85

Reducing factor for P of litterfall [] 0.85 0.85 0.85 0.85 0.85

Lignin fraction of litterfall [] 0.2 0.2 0.05 0.1 0.1

Lignin fraction of pruned biomass [] 0.4 0.4 0.1 0.4 0.4

Lignin fraction of root [] 0.2 0.2 0.05 0.15 0.15

Polyphenol fraction of litterfall [] 0.0975 0.0665 0.0156 0.05 0.0370 Polyphenol fraction of pruned [] 0.1310 0.0846 0.0214 0.0731 0.06 biomass Polyphenol fraction of root [] 0.1310 0.0846 0.0214 0.0731 0.06

Intercept for total biomass equation kg 0.35850266 0.42032406 0.82304565 0.55986736 0.2534828

Power for total biomass equation cm-1 2.42852598 2.52702458 2.55602073 1.23504819 1.74094669 Intercept for branch biomass kg 0.00137966 0.00045142 0.00191702 0.00186401 0.00204394 equation Power for branch biomass equation cm-1 3.99136449 5.32374294 3.83690076 2.97686043 3.19836794 Intercept for Leaf&twig biomass kg 0.33659699 0.41555195 0.80775386 0.49947191 0.18450442 equation Power for Leaf&twig biomass cm-1 2.10980295 2.32935183 2.48643129 1.03320421 1.17975306 equation Intercept for litterfall equation kg 0.00274465 0.00055548 0.17793489 -0.0036202 0.00232773

Power for litterfall equation cm-1 3.42852598 3.52702458 3.55602073 2.23504819 2.74094669

Wood density kg m-3 580 650 390 600 560

Root tip diameter cm 0.05 0.05 0.1 0.1 0.1 Max. root length density in layer1- cm cm-3 3 3 3 3 3 zone1 Max. root length density in layer1- cm cm-3 2.2 2 2 1.5 2 zone2 Max. root length density in layer1- cm cm-3 0.8 0.45 0.45 0.45 0.6 zone3 Max. root length density in layer1- cm cm-3 0.4 0 0 0 0.4 zone4 Max. root length density in layer2- cm cm-3 2.5 2.5 2.5 2.5 2.5 zone1 Max. root length density in layer2- cm cm-3 1.833 1.667 1.667 1.25 1.667 zone2 Max. root length density in layer2- cm cm-3 0.489 0.25 0.25 0.188 0.333 zone3 Max. root length density in layer2- cm cm-3 0.333 0 0 0 0.222 zone4 Max. root length density in layer3- cm cm-3 0.729 0.729 0.729 0.729 0.729 zone1 Max. root length density in layer3- cm cm-3 0.534 0.486 0.486 0.364 0.486 zone2 Max. root length density in layer3- cm cm-3 0.194 0.109 0.109 0.109 0.146 zone3 Max. root length density in layer3- cm cm-3 0.097 0 0 0 0.097 zone4 Max. root length density in layer4- cm cm-3 0.3 0.3 0.2 0.2 0.2 zone1 Max. root length density in layer4- cm cm-3 0.22 0.2 0.133 0.1 0.133 zone2 Max. root length density in layer4- cm cm-3 0.08 0.045 0.03 0.03 0.04 zone3 Max. root length density in layer4- cm cm-3 0.04 0 0 0 0.027 zone4 Fraction of roots infected by [] 0.3 0.3 0.3 0.3 0.3 mychorrhiza

• 230 • 12 Appendices

Parameters Units S. contorta D. validus G. arborea D. zibethinus A. heterophyllus Reduction of constant P by root mg cm-1 0 0 0 0 0 activity Relative transfer rate for N pool cm-2 day-1 0 0 0 0 0

Relative transfer rate for P pool cm-2 day- 0 0 0 0 0 Max. fire temperature tree can °C 75 75 75 75 75 tolerate Tree cover efficiency factor [] 0.5 0.5 0.5 0.5 0.5

Stella inputs

T_GroResInit 0.0008 0.00031 0.01 0.01 0.01

T_CanBiomInit 0.0001 0.0001 0.0005 0.0005 0.0005

T_WoodBiomInit 0.00013 0.00014 0.0005 0.0005 0.0005

T_HInit cm 0.36 0.29 0.50 0.86 0.63

• 231 • 12 Appendices

Curriculum vitae

Carsten Nikolaus Marohn

Address Rohrackerstr. 264 70329 Stuttgart Date of Birth Dec 6, 1968 Place of birth Stuttgart, Germany Nationality German

Professional experience Jun 2006 - present Research assistant (50%) at Institute of Plant Production and Agroecology in the Tropics / Subtropics, ReGrIn project, West Aceh / Hohenheim107 Sep 2002 - present Associate partner of terra fusca GbR, engineering & consulting108 Oct 1999-Feb 2002 Development worker German Development Service (DED) in Pucallpa, Peru. Focus on agroforestry and organic agriculture, institutional development. 1993-1998 Free lance gardening, stage hand at concert and theatre venues Nov 1994-May 1995 Internships at Weleda S.A., Buenos Aires and Villa Berna, Argentina, and on a smallholder farm in Mérida, Venezuela 1993-1995 Student research assistant (Wihi) at Institute of Plant Nutrition, Hohenheim Sep1990 - Jun 1992 Apprenticeship landscape gardening incl. tree nursery, Ute Haag GmbH, Stuttgart- Sonnenberg

107 http://www.worldagroforestrycentre.org/sea/projects/regrin/ 108 see www.terra-fusca.de for references

• 232 • 12 Appendices

Education Jan 2004 – present (submitted June 2007) PhD thesis at Institute of Plant Production and Agroecology in the Tropics / Subtropics (Prof. Sauerborn): Soils and carbon sequestration under agroforestry systems in Leyte, Philippines Higher education Studies of agrobiology at University of Oct 1992-Apr 1998 Hohenheim Major subjects: Soil Sciences, Plant Ecology, Ecotoxicology and Environmental Analytics, Microbiology, Agroecology in the Tropics/Subtropics Diploma thesis: Soils and land use in semi- arid Northeast Brazil (WAVES project) School education Fanny Leicht-Gymnasium, Stuttgart. 1979 -1988 Specialised courses German, English, maths, history

Trainings and other Oct-Dec 1999 Trainings in participatory methods and rural appraisal, project management and intercultural communication at DED Aug 1988-Apr 1990 Civil service: Youth Centre Stuttgart- Vaihingen

Languages German: Mother tongue English: Fluent Spanish: Fluent Portuguese: Working knowledge Computer skills Standard office packages under Microsoft and Linux. Graphics: Photoshop. Statistics: SPSS, Sigma, SAS, Minitab. Modelling: WaNuLCAS, FALLOW, PCRaster

Carsten Marohn Stuttgart, June 2007

• 233 • 12 Appendices

Erklärung

Hiermit erkläre ich, die vorliegende Dissertation selbständig angefertigt, nur die angegebenen Hilfsmittel benutzt sowie wörtlich oder inhaltlich aus anderen Quellen übernommene Stellen als solche gekennzeichnet zu haben.

Carsten Marohn Stuttgart, im Juni 2007

• 234