And Rainforest Conversion on River Discharge in Central Sulawesi

Brigham Young University BYU ScholarsArchive 4th International Congress on Environmental International Congress on Environmental Modelling and Software - Barcelona, Catalonia, Modelling and Software Spain - July 2008

Jul 1st, 12:00 AM Effects of „ENSO-events” and rainforest conversion on river discharge in Central Sulawesi (Indonesia) – problems and solutions with coarse spatial parameter distribution for water balance simulation G. Gerold

C. Leemhuis

Follow this and additional works at: https://scholarsarchive.byu.edu/iemssconference

Gerold, G. and Leemhuis, C., "Effects of „ENSO-events” and rainforest conversion on river discharge in Central Sulawesi (Indonesia) – problems and solutions with coarse spatial parameter distribution for water balance simulation" (2008). International Congress on Environmental Modelling and Software. 183. https://scholarsarchive.byu.edu/iemssconference/2008/all/183

This Event is brought to you for free and open access by the Civil and Environmental Engineering at BYU ScholarsArchive. It has been accepted for inclusion in International Congress on Environmental Modelling and Software by an authorized administrator of BYU ScholarsArchive. For more information, please contact [email protected], [email protected]. iEMSs 2008: International Congress on Environmental Modelling and Software Integrating Sciences and Information Technology for Environmental Assessment and Decision Making 4th Biennial Meeting of iEMSs, http://www.iemss.org/iemss2008/index.php?n=Main.Proceedings M. Sànchez-Marré, J. Béjar, J. Comas, A. Rizzoli and G. Guarisco (Eds.) International Environmental Modelling and Software Society (iEMSs), 2008

Effects of „ENSO-events” and rainforest conversion on river discharge in Central Sulawesi (Indonesia) – problems and solutions with coarse spatial parameter distribution for water balance simulation

G. Gerolda and C. Leemhuisb

a Landscape Ecology Unit, Institute of Geography, University of Göttingen, Goldschmidtstr. 5, 37077 Göttingen, Germany ([email protected]) b Center for Development Research, University of Bonn, Walter-Flex-Str. 3, 53113 Bonn, Germany ([email protected])

Abstract: Natural interannual climate fluctuation as ENSO-events and forest conversion play an important role for the water balance of tropical catchments. Distributed hydrological modelling that relate land cover changes and climate changes with river discharge changes for humid tropical catchment areas at the mesoscale level are rare. This article applies a hydrological modelling approach to analyse the impact of land cover changes and ENSO-events on the water resources of a mesoscale humid tropical catchment. Because of a lack of detailed spatial soil information, a topographic based nested scale approach to parameterise soil physical properties was used for the hydrological model WASIM-ETH (Schulla and Jasper, 1999). The PHA (potential homogeneous soil texture areas) approach produces reasonable results for the water balance components for the whole catchment, but is too coarse for flow component differentiation within the catchment. Thereafter the distributed hydrological model WASIM-ETH (“Water balance Simulation Model”) was calibrated and validated for the current land use (2001/2002, Landsat/ETM+ scene). Model results were generally consistent with the observed discharge data and reproduced the seasonal discharge dynamics well. The implications of possible future climate and land use conditions on the water balance of the Gumbasa River sample catchment were assessed by a scenario analysis. The results of the scenario simulations clearly demonstrate a strong relationship between deforestation rates and increasing discharge variability. Especially a significant increase of high water discharges was simulated for the applied land use scenarios. The main results of the scenario analysis are: (1) that ENSO anomalies of precipitation (monthly decrease for an average ENSO -21 until -80% from June to October) lead to an increase of discharge variability. (2) that strong ENSO-events (El Nino) lead to 30% reduction of the water yield and annual crop scenarios up to 1,200 m a.s.l. showed a 42% and the perennial crop scenario (cacao) a 23% increase in river discharge with high increase in overland flow and flooding risks. (3) that ENSO-events (El Nino) decrease the potential (flooding) area of paddy rice cultivation to one third in the second half of the year. With regard to the high deforestation rates of the research catchment and increase of El Nino frequency with climate change in the future, one can expect further negative changes for the water resources in Central Sulawesi.

Keywords : Hydrological Modelling; ENSO; Tropical Deforestation; Spatial Parameter Generation.

553 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

1. INTRODUCTION

Tropical catchment areas are mainly located within developing countries and are predominantly ungauged or poorly gauged. The vast majority of tropical catchment studies have been conducted at the lower meso spatial (<10 km²) and time (<5 years) scale (Bruijnzeel, 1996). Costa et al. (2003) confirm that studies relating changes in land cover with changes in river discharge of tropical catchments are abundant especially on an experimental scale (< 1 km²). The global and local scale, which is applied in the vast majority of tropical hydrological studies (e.g. Güntner, 2002; Döll et al., 2003; Godsey et al., 2004; Fleischbein et al., 2005) is far from the regional reality where decisions on water resource management are developed and implemented (Falkenmark and Rockström, 2004). Data availability is one of the main difficulties for the analysis of tropical catchment processes on a mesoscale level. An appropriate database, which is urgently needed to enable water resources to be developed and managed, is lacking (IAHS 2003). Therefore long term records, which are required to study hydrological trends, are not widely available (Manley and Askew, 1993). Considering the increasing stress on water resources in humid tropical developing countries due to rapid deforestation rates and climate change there is an urgent global research need in humid tropical hydrology and its associated mesoscale catchment processes. After Abbot and Reefsgard (1996) the spatially-distributed and time- dependent hydrological model becomes the “conditio sine qua non” for investigations in this area.

The El Niño/ Southern Oscillation (ENSO) phenomenon is the strongest known natural interannual climate fluctuation. The two most recent extreme ENSO events during 1982/83 and 1997/98 had severe impacts on major parts of Indonesia’s socio-economy. Understanding the hydrology of humid tropical catchments is an essential prerequisite to investigate the impact of climate variability. Additionally, land use changes by forest conversion alter the components of the water cycle in catchments of Indonesia. Rainforest conversion predominately into annual cultures and cacao-systems was intensified in Central Sulawesi during the last decade. Within STORMA (DFG-project) and IMPENSO (BMBF- project) project work we investigate and quantify the impact of ENSO induced climate variability and forest conversion effects on the water balance of a mesoscalic tropical catchment. Main goals are: how ENSO-caused precipitation anomalies and forest conversion effects the main water balance components and river discharge in a mesoscale tropical catchment.

2. STUDY AREA – GUMBASA RIVER CATCHMENT

The Gumbasa River catchment (Fig. 1) is a sub-basin of the Palu River catchment (2,694 km²), which is located in Central Sulawesi, Indonesia (1°S, 120°E). The Gumbasa River catchment has a total drainage area of 1,275 km² and is characterized by a gross annual areal precipitation of 2,000 mm/yr and a mean annual discharge of 22 m³s-1. The seasonal hydrological regime of the catchment is described by two peaks and strongly corresponds with the bimodal monsoonal pattern of the yearly precipitation distribution of Central Sulawesi, Indonesia. Like other river basins that are located in monsoon regions also the Gumbasa River shows a great variation of seasonal and annual flow (Oyebande and Balek, 1987). ENSO years are characterized by reduced precipitation from July to October, which represents the dry period of the monsoonal setting of Central Sulawesi. The elevation ranges from 99-2,491 m.a.s.l with a mean elevation of 1,119 m.a.s.l. and a mean slope of 10.37°. About two-thirds of the catchments area is classified as slope with moderate and strong inclination (69 %), which emphasizes the mountainous character of the watershed. The irrigated paddy fields and other agricultural land use systems are situated mainly in the valleys. Mountainous tropical forest with a total percentage of 82 % forms the main land cover type within the watershed.

2 554 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

INDONESIA study area

Palu River catchment Gumbasa River sub-catchment Lake Lindu Experimental Nopu catchment

010205 Kilometers

Figure 1. The Palu River watershed, the Gumbasa River modelling catchment and the experimental Nopu catchment, Central Sulawesi, Indonesia (1°S, 120°E)

3. MATERIAL AND METHODS

3. 1 Temporal and spatial data base

To study the impact of forest conversion and ENSO effects on the water balance of the Gumbasa River catchment we applied the Water Flow and Balance Simulation Model WASIM-ETH (Schulla and Jasper, 1999). WASIM-ETH is a process-based fully distributed catchment model. The spatial resolution is given by a grid and the time resolution can vary from minutes to days. The main processes of water flux, storage and the phase transition of water are simulated by physically-based simplified process descriptions (Schulla, 1997). The meteorological input data of the model are interpolated for each grid cell and are followed by the simulation of the main hydrological processes like evapotranspiration, interception, infiltration and the separation of discharge into direct flow, interflow and base flow. These calculations are modularly built and can be adapted to the physical characteristics of the catchments area. For the set up of the Gumbasa River catchment a 500 m x 500 m raster spatial, and a daily temporal resolution was chosen. As spatial input data WASIM-ETH requires a Digital Terrain model, a land cover map and a soil map. All spatial data is required in a raster data format (grid) with unique spatial resolution and extension. Starting from the primary grid of the topography secondary grids were generated with a topography analysis program.

To obtain a feasible temporal data set of meteorological forcing and hydrological calibration data for the performance of a distributed hydrological model the catchment was equipped with four automatic stage recorders and eight automatic climate stations monitoring precipitation, temperature and humidity. Additionally five automatic climate stations measure the complete meteorological parameters including wind speed and global radiation (Fig. 2). We measured hydrological and meteorological data for the period Sept. 2002 – Sept. 2004. Yearly areal precipitation of Gumbasa catchment (2,150 mm) is comparable with yearly precipitation averages from climate stations (e.g. Kulawi 1916- 1941: 2,446 mm). For the calibration (01/09/2002–31/08/2003) and validation (01/09/2003–31/08/2004) of the hydrological model WASIM-ETH we chose the land cover classification of the Landsat/ETM+ scene from 2002. The land cover class concept is based

3 555 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge on a global hierarchical class definition system (LCCS) and has been adapted to the project region in Central Sulawesi. Each land cover type refers to corresponding physical vegetation parameters like stomata resistance, Leaf Area Index (LAI), albedo, vegetation height, rooting depth and vegetation cover. The parameterization of the vegetation physical properties was mainly derived from a global vegetation guide by Matthews (1999), single studies (Mo et al., 2004; Dijk and Bruijnzeel, 2001; Körner, 1994) and local studies (Bohmann, 2004; Falk, 2004) which were conducted within the research area.

#³ c

c /# c c c c c #³ c#³ #³c

#³ c c c c

/#c

0 10 20 Kilometers

#³ automaticautomatic stage stage recorder recorder mainmain irrigation irrigati channeonl chanel

/# stagestage recorderrecorder - Forestry – F departmenorestry t LoreLore Lindu Lindu National ParcNational Park c cliclimmateate station sta tion PALUPalu RIVER river watershed watershed c cliclimmateate station station - STORMA STORM project A

Figure 2. Location of the climate and hydrological stations within the Palu river watershed, Central Sulawesi, Indonesia

3. 2 Application of the topographic based soil parameterisation approach

The only available soil map of the research area is the soil map of Indonesia ( 1 : 1,000,000), which is based on available exploration soil maps that were published in various years. Therefore the “Pedotransfer Function” approach (Bohne, 1993) for the parameterisation of soil hydraulic properties could not be applied at all. PTFs are the most widely used method to estimate soil hydraulic properties for larger areas (Hodenett and Tomasella, 2002). Sobieraj et al. (2001) evaluated the performance of nine published PTFs for estimating the soil hydraulic properties of a rainforest catchment for modelling stormflow generation. They conclude that the published PDFs are inadequate to model stormflow generation, because runoff was overestimated for all events. The main differences between van Genuchten soil water-retention parameters for temperate and

4 556 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge tropical soils were investigated by Hodenett & Tomasella (2002). Their survey showed that most of the PTFs have been developed using databases for the soils of temperate regions, which are non-transferable for tropical soils.

Based on an experimental small headwater catchment of the Gumbasa River (Nopu 2.3 km2, Fig. 1), studied for the effects of forest conversion on water balance and nutrient output since 2001 within the STORMA project (Kleinhans, 2004), a topographic based nested scale approach was developed. The topographic based nested scale approach to parameterise soil physical properties for mesoscale hydrological modelling consists of an experimental hillslope soil analysis and a classification of the mesoscale catchment area in potential homogeneous soil texture areas (PHA) according to their topographic characteristics. The surveyed catena of the experimental catchment located within the mesoscale catchment area serves as a leitcatena assuming a similar soil hydraulic behaviour for equal PHAs. The topographic based nested approach assumes that texture classes and bulk density mainly affect the soil physical properties and depend on the degree of slope and its geomorphologic location (basin, valley, slope, summit area). For the classification of PHA`s complex morphometric terrain factors are calculated on the basis of the Digital Terrain model (DTM). The relevant classification factors are (a) the relative altitude above channel line (a.a.c.l.), (b) catchment area and (c) slope angle. Table 1 lists the morphometric terrain description of a PHA classification.

Table 1: PHA classification and its corresponding morphometric terrain factors

Morphometric description mean values m.a.s.l. a.a.c.l. slope (°) (m) (m) I Intramontane basin (low altitude above channel 591.46 2.68 1.60 line) II Intramontane basin (moderate altitude above 453.80 8.00 2.50 channel line) III Valleys outside the intramontane basin catchment 802.05 5.17 9.31 (large catchment area) IV Valleys outside the intramontane basin catchment 1154.31 3.71 12.14 (moderate catchment area) V Slope with moderate inclination 1191.37 88.13 17.43 VI Slope with strong inclination 1143.57 121.84 30.77 VII Summit area with low inclination 1337.96 87.88 5.31 VIII Summit area with moderate inclination 1340.72 168.00 12.74

Table 2: PHA classes and its associated soil physical parameters

PHA Pot. texture α n ψ k l l θ s θ r s krec t n I sand 0.45 0 7.36 1.23 385 9.0E-4 0.1 1.0 10 II sand 0.45 0 7.36 1.23 385 5.0E-4 0.1 1.0 10 III loamy sand 0.41 0 1.86 1.26 375 4.0E-4 0.1 0.9 10 IV loamy sand 0.41 0 1.86 1.26 375 5.0E-4 0.1 0.9 10 V sandy loam 0.45 0 4.01 1.2 345 1.5E-3 0.1 0.9 10 VI loam 0.49 0 4.01 1.2 350 5.3E-3 0.1 0.7 10 VII sandy clay loam 0.51 0 2.0 1.13 290 5.0E-4 0.1 0.3 10 VIII sandy clay loam 0.51 0 2.0 1.13 290 4.0E-4 0.1 0.9 10

The surveyed catena of the experimental catchment served as a leitcatena for the whole mesoscale catchment area. Hence a similar soil hydraulic behaviour was assumed for equal PHAs. Table 2 lists the required soil hydraulic parameters for the soil model of WASIM-

5 557 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

ETH according to the PHAs classes. Instead of using the measured values of the saturated hydraulic conductivity ks we increased the value consistently. With the hydrological model application of the experimental catchment study using the initial measured saturated hydraulic conductivity values no satisfactorily calibration modelling results were achieved. This effect is related to the modelling concept of the soil model using the Richards Equation. Here the soil is represented by a homogeneous matrix and therefore the flux of water through the matrix is characterised by a homogeneous matrix flow. But in reality the soil has an inhomogeneous structure, which implies a heterogeneous flux of water with preferential flow through macropores. Thus if we assume a flux of water through a homogeneous soil matrix the saturated soil hydraulic conductivity is relatively higher than the measured value. The recession constant krec represents an adjustable calibration parameter that determines the varying saturated hydraulic conductivity with increasing soil depth. Initially this parameter is set to 0.1. The parameter values of the suction ψ were transferred from the study of Niehoff (2001). The parameters describing the van Genuchten equation were computed from measured values. Empirical studies (Woesten et al., 2001)

showed a good parameter adjustment if the residual water content θ r was set to zero.

Accordingly θ r was set to zero for all PHAs classes. In general it should be reminded that a calibration parameter no longer describes the measured point value but rather an effective parameter that represents the average value for the element (Grayson, 2000). Hence they reproduce the bulk behaviour of a finite volume and cannot directly be related to point measurements at all.

3. 3 Calibration and validation

The calibration period was set for the period 01/09/2002-31/08/2003, which corresponds to the first year of the measurement period. To achieve realistic soil moisture conditions for the initial conditions of the calibration run, the start of the whole model run was predated to the 01/09/2001. The required temporal data for this period of model initialisation was copied from the first year of measurements. The validation period was split into the validation period (01/09/2003-31/08/2004), and the total period of simulation (01/09/2002- 31/08/2004). Table 3 list the complete calibration and validation results for the model efficiency of the Danau Lindu, Sopu and Takkelemo gauging sites simulation results on a daily and weekly resolution. Further the graphical evaluation of observed and simulated discharge for the Lake Lindu, Takkelemo and Sopu sub-catchment (Fig. 3, Leemhuis, 2005) shows a very good agreement. The baseflow is simulated within realistic values and also the discharge fluctuation during the rainy season from March till May is well reproduced. The flat hydrograph of observed discharge for the Lake Lindu sub-catchment, which occurs end of April, is due to a faulty design of the hydrological station.

Table 3: Coefficients of efficiency R2 for the sub-catchments Lake Lindu, Sopu and Takkelemo on a daily and weekly resolution for the calibration, validation-split sample and validation whole period

R² Lake Lindu Sopu Takkelemo daily weekly daily weekly daily Weekly calibration period 0.83 0.86 0.79 0.77 0.58 0.77 validation period - split 0.61 0.62 0.58 0.44 0.41 0.50 sample validation period - whole 0.74 0.76 0.69 0.68 0.45 0.61

6 558 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

observed discharge

) 8

-1 simulated discharge simulated baseflow 6

discharge (mm d (mm discharge 2

3 3 3 1/0 7/0 /0 09 1/11/02 /0 1/05/03 /0 / 01/09/02 0 01 01/03/03 0 01 01

Figure 3. Calibration of the Lake Lindu sub-catchment (daily resolution) : comparison between observed and simulated discharge (R2 = 0.83)

3. 4 ENSO scenario generation

A statistical scenario approach was applied to generate scenarios of spatially and temporally variable ENSO-caused rainfall anomalies as input data for a hydrological scenario run of the Gumbasa River watershed. Five climate stations with a long term historical record (1981-1999) were located within the area around the climate stations, which were used for the calculation of areal precipitation. The interdisciplinary climate research group of the IMPENSO project calculated the monthly precipitation anomaly of ENSO years in percent related to the mean monthly precipitation of the observed years. Therefore two graded ENSO scenarios which reflect the mean anomaly of all observed ENSO events (1987, 1991, 1994, and 1997) and on the other hand the extreme ENSO event of 1997, were generated. Thereafter the climate stations used for the interpolation of areal precipitation were allocated to one of the long term stations according to their elevation and topographic position. The allocation is mainly basin related although each of the long-term climate stations is located within one of the main basins,. The study of Aldrian (2003) has shown that ENSO events have a significant influence on the climate of Central Sulawesi between the dry period from June till October. The ENSO scenario was applied for this time frame which is affected by ENSO related precipitation anomalies. For the generation of the two graded ENSO scenarios the modelling year 2003 was taken as a base year.

4. DISCUSSION AND RESULTS

4. 1 Land use change

The application of the hydrological model WASIM-ETH on the Gumbasa River catchment in Central Sulawesi, Indonesia is the model’s first simulation of the water balance of a mesoscale humid tropical catchment. The achieved model efficiencies can be classified among other WASIM-ETH model applications in various catchment studies of different climatic regions (Schulla, 1997; Jasper et al., 2002; Piepho, 2003). The modularly based character of WASIM-ETH allows a general adaptation of the hydrological model to different climatic regions and catchment data bases in space and time. If the overall model uncertainty is assessed with regard to the available spatial data set the land use classification and the soil classification represent in general uncertainty sources. With regard to land use classification derived from satellite images like Landsat ETM+ it should

7 559 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge be further analysed if different land use classification methods result in different modelling results.

Table 4: Total yearly simulated water balance for the Gumbasa River catchment and the Lake Danau and Takkelemo sub-catchment for 2003, compared to an annual crop scenario, perennial crop scenario (cacao) and ENSO-scenarios with average ENSO-event and 1997 ENSO-event; Δ represents the changes of precipitation, evapotranspiration and total discharge in percent to the control run (current climate and land use 2003)

Scenario catchment precipitation evapo- total discharge forest area tranpiration (%) P Δ P aET Δ aET Q Δ Q (mm) (%) (mm) (%) (mm) (%) current Gumbasa (82) 2150 - 1543 - 590 climate a. land L.Danau (95) 2192 - 1295 - 896 use 2003 Takkelemo (79) 2259 - 1428 - 804 current Gumbasa (82) 2150 - 1265 - 20.0 838 42.0 climate 2003 a. L.Danau (95) 2192 - 1112 - 14.1 1218 35.9 annual crop1 Takkelemo (79) 2259 - 1124 - 21.3 1113 38.4 current Gumbasa (82) 2150 - 1361 - 11.8 724 22.7 climate 2003 a. L.Danau (95) 2192 - 1144 - 11.7 1097 22.4 cacao Takkelemo (79) 2259 - 1222 - 14.4 1000 24.4 plantation2 average ENSO Gumbasa (82) 1614 - 24.9 1496 - 3.0 401 - 32.0 event a. land L.Danau (95) 1710 - 22.0 1281 - 1.1 684 - 23.6 use 2003 Takkelemo (79) 1724 - 23.7 1423 - 0.4 561 - 30.2 ENSO event Gumbasa (82) 1586 - 26.2 1487 - 3.1 397 - 32.7 1997 a. land L.Danau (95) 1699 - 22.5 1279 - 1,2 684 - 23.6 use 2003 Takkelemo (79) 1709 - 24.4 1421 - 0.5 557 - 30.7 1 forest conversion with annual crop until 1200 m a.s.l. 2 forest conversion with cacao plantation until 1200 m a.s.l.

We simulated a land-use history scenario with the Landsat/ETM land cover classification of 2001 as input land cover grid for the analysis of the impact of observed land cover changes from 2001 till 2003 within the Gumbasa River catchment (details on land cover changes see Leemhuis et al., 2007). Furthermore we applied a total change scenario, which assumed that all forest up to 1,200 m a.s.l. is converted in a first conversion phase into annual crops such as maize or dry rice and within a second step into perennial crops (table 4).

14 actual land use

] land use scenario A1 12 -1 land use scenario A2 10

8 6

discharge [mm d discharge [mm 4

2 Figure0 4. Scenario results for daily discharge (Gumbasa catchment) with two different land use developments ( A1 = conversion of rainforest ≤ 1,200 m a.s.l. in 3 annual crops; A2 = coversion of rainforest ≤ 1,200/03 m a.s.l. into cacao 01/03 03/03 05/03 07/03 09 11/0 1/ 01/ 01/ 01/ plantation)0 01/ 01/

Figure 4. Scenario results for daily discharge (Gumbasa catchment) with two different land use developments ( A1 = conversion of rainforest ≤ 1,200 m a.s.l. in annual crops; A2 = coversion of rainforest ≤ 1,200 m a.s.l. into cacao plantation)

8 560 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

Cacao is a cash crop in Central Sulawesi and therefore is most likely to be planted as perennial crop. In Central Sulawesi the cacao plant can be cultivated up to an elevation of 1,200 m a.s.l., since up to this elevation the temperature does not drop beneath 16°C at night (Rehm, 1989). For both scenarios the physical vegetation parameters were altered according to the applied land cover type. The soil physical properties were kept constant, because the changes of soil physical parameters induced by land cover changes were not investigated for the Gumbasa River catchment.

The comparison of the simulated yearly water balance and the applied land cover scenarios gives a general first overview about the impact of land cover change on the hydrological performance of the observed watershed. Table 4 compares the changes induced by land cover change for the total Gumbasa River catchment and the sub-catchments of Danau Lindu (95% forest cover) and Takkelemo (79% forest cover) respectively of the following water balance components: areal precipitation (P), actual evapotranspiration (aET) and total discharge (Q). With respect to the actual land cover changes from 2001 to 2002 the effect on the yearly simulated water balance is rather insignificant. The hydrological model calculated a 0.2% of the total water balance decrease of total yearly discharge for the Gumbasa River catchment (Leemhuis et al., 2007). The annual crop scenario with an increase of 42% of the yearly total run-off for the Gumbasa River catchment and the perennial crop scenario with a rise of 23% of the yearly total discharge indicate a great increase of the yearly total discharge with cumulative deforestation rates (Fig. 4). All changes of the total discharge were induced by an alteration of the yearly evapotranspiration rate, due to a decline of LAI, root depth, vegetation height and stomata conductance. Less water is evaporated by interception and hence throughfall increases and more water can infiltrate into the soil. Plot research results within STORMA shows a reduction in aET for annual crops (maize) with an order of 10-20% and for perennial crops (cacao) with 5-12 % (Falk 2004). The reduction of the evapotranspiration causes the soil to be wetter and therefore more responsive to rainfall (Bruijnzeel, 2004). The graduated impact of annual and perennial crop scenario is related to the more similar physical vegetation parameters of the cacao tree in comparison to the broadleaf evergreen forest of the catchment, comparing relationship between forest cover change and changes of the yearly total run-off. Regarding the crop scenario for the Danau Lindu sub-catchment the lower increase of discharge is due to the less decrease of evapotranspiration.

4.2 ENSO effects

For the observed ENSO-events two scenarios which reflect the mean anomaly of all observed ENSO-events (1987, 1991, 1994 , 1997; = average ENSO event) and the extreme ENSO event of 1997 were generated. The current conditions of 2003 serve as a control run for the evaluation of the proportional changes. During ENSO scenario A the total yearly areal precipitation of the Gumbasa River catchment is reduced by 25 %, whereas the total yearly evapotranspiration rate is only slightly minimised by 3 %. The main difference of the yearly water balance were observed for the total yearly discharge, which is reduced by 32 % (Fig. 5).

With a reduction of the total discharge by 24 %, the Danau Lindu sub-catchment is least affected by the precipitation anomaly, whereas the small Takkelemo catchment reaches similar reduction values like the Gumbasa River catchment. Even though the strong ENSO event of 1997 is applied by ENSO scenario B, total yearly changes of the water balance in comparison with ENSO scenario A are minimal. For example the decrease of the yearly total discharge only is reduced by 0.7 % for the Gumbasa River catchment. The comparison of the yearly water balance of the three different catchment types shows that the ENSO- caused precipitation anomaly leads to a slight variance of the yearly precipitation decline among the catchments of 2.9 % for ENSO scenario A and 3.7 % for ENSO scenario B. For both ENSO scenarios the Danau Lindu subcatchment indicates with 22.0 % (Scenario A) and 22.5 % (Scenario B) the lowest yearly precipitation decline, whereas the Gumbasa River catchment has a yearly average precipitation decline of 24.9 % (Scenario A) and 26.2

9 561 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

10 status quo 2003 E0/L0

] 8 ENSO scenario EA

-1 ENSO scenario EB 6

discharge [mm d discharge [mm 2

3 /03 01/03 1/07/03 01/ 01/03/03 01/05/03 0 01/09 01/11/0

Figure 5. Scenario results for daily discharge (Gumbasa catchment) with two different ENSO-events (EA = average year with rainfall decrease from 1.07. – 30.11.2003 with - 60%; EB = average year with rainfall decrease 90%)

% (Scenario B). The simulated yearly precipitation decline is not linear with the simulated yearly discharge decline, which indicates a far wider variance of the different catchment types with 8.4 % for ENSO scenario A and 9.1 % for ENSO scenario B. Again for both scenarios the Danau Lindu catchment has the lowest discharge decline with 23.6 % for both ENSO scenarios and the Gumbasa River catchment the highest discharge decline among the observed catchments with 32.0 % (ENSO scenario A) and 32.7 % (ENSO scenario B). The simulation for the Takkelemo catchment reaches almost similar precipitation and discharge decline values as for the Gumbasa River catchment. The Danau Lindu catchment is characterized by a low discharge variability due to its retention capability. In general the discharge variability during ENSO years is significantly higher.

5. Conclusions

Common problems in mesoscalic hydrological modelling in developing countries are insufficient time series of input- and output data and insufficient spatial data resolution. For the Palu river catchment only one Indonesian climate station and two automatic stage recorders from the Forestry Department of Palu for two subcatchments existed at the beginning of the IMPENSO-project. Therefore an additional network of six automatic stage recorders and eight additional climate stations have been used to collect data since September 2002 for the application of the water balance model WASIM-ETH. The main problem was the coarse scale of available soil maps (1 : 1 Mill.) and lumped representation of soil data from STORMA (isolated small areas). Therefore a more general topographic- and knowledge-based approach has been applied, that classifies potential homogenous areas (PHA) with morphometric terrain factors (Böhner et al., 2002, Leemhuis, 2005). The PHA approach produces reasonable results for the water balance components for the whole catchment, but is to coarse for flow component differentiation within the catchment. However, the calculated decrease of the yearly total discharge by about 30% during ENSO events corresponds with a cross-correlation analysis of discharge time series for two catchments in Central Sulawesi (Miu and Wuno River; Leemhuis, 2005).

ENSO caused precipitation anomalies lead to an overall increase of the discharge variability. Moreover the ENSO scenario discharge simulations shows that mainly the low water and mean discharges are effected in the second half of the year (August till December) with a shift of one to two months. Strong ENSO-events like 1997 lead to 32%

10 562 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge reduction in water yield. Catchment characteristics, mainly change in forest area, cause different hydrological response magnitude to the ENSO impact. The relative lower discharge decrease of the Danau Lindu catchment is due to the higher retention capability of the Lake Lindu. The higher portion of forest area is coupled with slower drying out of the soil water reservoir and a reduced evapotranspiration rate for the high altitude rainforest. It can be concluded, that for the “Gumbasa Irrigation Scheme”, based on the water resources of the Gumbasa catchment, the conditions of the mountainous rainforest headwater catchments like the Danau Lindu with less disturbed rainforests are important for the input of irrigation water. With average ENSO events scenario results show for the second half of the year a decrease in potential irrigation area of the Gumbasa River Irrigation Scheme by about -23 until -66 % (Leemhuis, 2005).

The results of the applied land use scenarios agree with the findings of the tropical catchment study by Kleinhans (2004)], who applied the hydrological model WASIM-ETH on a small headwater catchment of the Gumbasa River catchment. Both studies used the same hydrological model, therefore it was likely that similar land use scenario results would be also calculated on a bigger scale. Experimental paired catchment studies confirm the simulated increases of peak flows after forest removal (Bonel and Balek, 1993; Chandler & Walter, 1998), which is explained by the associated reduction in ET that causes the soil to be wetter and therefore more responsive to rainfall (Bruijnzeel, 2004). Furthermore the scenario results show that after the complete conversion of all deforested area into cacao plantations still the high water discharge is characterised by enhanced peak flows, which indicates a continued alteration of the water balance for cacao plantations.

But cacao plantations have a more moderate impact on the water balance of a catchment than annual crops like maize. Interpreting the simulation results of the dry season flow (low water discharges) requires caution. Because numerous studies of smaller tropical catchments produced higher baseflow after forest removal, whereas other studies report a decreased baseflow. Regarding these contradictory catchment study results, Bruijnzeel (2004) raised the discussion about the “low flow problem”. Whether the baseflow is raised or reduced after forest removal is strongly connected to the degree of surface disturbance. Costa (2005) argues that as a result of the decreased hydraulic conductivity due to soil consolidation less water can infiltrate, which again leads to an increased direct discharge and actually leads to a decrease of the low water discharge. The “low flow problem” indicates that information on land cover change (from Landsat/ETM+ satellite images) and data on changes of the soil physical properties after forest removal are necessary to improve simulation correctness for the low water discharges. Further, it would also improve the simulation of the mean and high water discharges. In a recent study in the small headwater Nopu catchment the importance of the change of saturated hydraulic conductivity with age and intensity of traditional smallholder land use is shown (De Vries et al., 2008). There extensive used cacao plantations conserve high rainfall infiltration into the soil and leads to higher groundwater recharge.

We can conclude that green water flow plays a significant role within tropical catchments. Most of the alterations of the water balance due to forest conversion can be explained by changes of the green water flow. Besides monitoring of alterations of physical vegetation parameters, there is urgent need of research within tropical catchment areas to describe the corresponding changes of the soil physical parameters and to have higher spatial resolution for effective physical soil parameters.

ACKNOWLEDGEMENTS

This study was supported by the Federal Ministry of Education and Research (BMBF) and by the German Research Foundation (DFG). Special thanks to the Institut Pertanian Bogor (IBP), Java, Indonesia and the Universitas Universitas Tadulako (UNTAD), Palu, Central Sulawesi, Indonesia for the productive research cooperation.

11 563 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

REFERENCES

Abbott, M.B. and J.P. Refsgaard, Terminology, modelling protocol and classification of hydrological model codes. In: Abbott, M. B. and Refsgaard, J. P. (Eds.): Distributed hydrological modelling. Kluwer Academic Publishers, Dordrecht, Boston, London, 17- 37, 1996. Aldrian, E., Simulations of Indonesian rainfall with a hierarchy of climate models. Diss., Universität Hamburg, 23-45, 2003. Alves, D. S., Escada, M. I. S., Pereira, J. L. G. and C.A. Linhares, Land use intensification and abandonment in Rondônia, Brazilian Amazônia. International Journal of Remote Sensing, 24, 899-903, 2003. Asdak, C., Jarvis, P. G., van Gardingen, P. and A. Fraser, Rainfall interception loss in unlogged and logged forest areas of Central Kalimantan, Indonesia. Journal of Hydrology, 206, 237-244 1998. Barnes,C. and M. Bonell, How to chose an appropiate catchment model. In: Bonell, M. and Bruijnzeel, L. A. (Eds.): Forests, Water and People in the Humid Tropics. University Press, 717-741, 2005. Böhner, J., Köthe, R., Conrad, O., Gross, J., Ringeler, A. and T. Selige, Soil regionalisation by means of terrain analysis and process parameterisation.- EU Soil Bureau Research Report No. 7, EUR 20398 EN, 213-222, 2002. Bohman, K., Functional and morphological diversity of trees in different land use types along a rainforest margin, in Sulawesi, Indonesia. Dissertation, Universität Göttingen, 55-63, 2004. Bohne, K., Horn, R. and T. Baumgartl, Bereitstellung von van - Genuchten Parametern zur Charakterisierung der hydraulischen Bodeneigenschaften. Zeitschrift für Pflanzenernährung und Bodenkunde, 156, 229-233, 1993. Bonell, M. and J. Balek, Physical processes. In: Bonell, M., Hufschmidt, M., Gladwell, J. (eds.): Hydrology and water management in the humid tropics. Cambridge, 176–177, 1993. Bonell, M., Runoff generation in tropical forests. In: Bonell, M. and Bruijnzeel, L. A. (Eds.): Forests, Water and People in the Humid Tropics. University Press, Camebridge, 314-406, 2005. Bruijnzeel, L.A., Hydrology of moist tropical forests and effects of conversion: a state of knowledge review. UNESCO International Hydrological Programme, 20-20, 1996. Bruijnzeel, L.A., Tropical forests and water yield. Laxenburg, Austria, 2000. Bruijnzeel, L.A., Hydrological functions of tropical forests: not seeing the soil for the trees? Agriculture, Ecosystems & Environment, 104, 185-228, 2004. Chandler, D. G. and M.F. Walter, Runoff responses among common land uses in the uplands of Matalom, Leyte, Philippines. In: Transactions of the ASAE 41, 1635–1641, 1998. Costa, M. H., Botta, A. and J.A. Cardille, Effects of large-scale changes in land cover on the discharge of the Tocantins River, Southeastern Amazonia. In: Journal of Hydrology 283, 206–217, 2003. Costa, M. H., Large scale hydrological impacts of tropical forest conversion. In: Bonell, M. and Bruijnzeel, L. A. (Eds.): Forests, Water and People in the Humid Tropics. University Press, Camebridge, 590-598, 2005. De Vries, K. and G. Gerold, The consequences of rainforest conversion to agricultural land for saturated hydraulic conductivity in Cental Sulawesi, Indonesia. Catena (submitted), 2008. Dijk, A. and L.A. Bruijnzeel, Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 2. Model validation for a tropical upland mixed cropping system. Journal of Hydrology, 247, 239-262, 2001. Döll, P., Kaspar, F. and B. Lehner, A global hydrological model for deriving water availability indicators: model tuning and validation. Journal of Hydrology, 270, 105- 134, 2003. Elsenbeer, H., Pedotransfer functions in hydrology. Journal of Hydrology, 251, 121-122, 2001.

12 564 G. Gerold & C. Leemhuis/ Effects of „ENSO-events“ and rainforest conversion on river discharge

Falk, U. , Turbulent fluxes of CO2, H2O and energy in the atmospheric boundary layer above tropical vegetation investigated by Eddy-Covariance measurements. Diss. Göttingen., 2004. Falkenmark, M. and J. Rockström, Balancing water for humans and nature. Earthscan, London, 181-199, 2004. Fleischbein, K., Wilcke, W., Goller, R., Boy, R., Valarezo, C., Zech, W. and K. Knoblich, Rainfall interception in a lower montane forest in Ecuador: effects of canopy properties. Hydrological Processes, 19, 1355-1371, 2005. Godsey, S., Elsenbeer, H. and R. Stallard, Overland flow generation in two lithologically distinct rainforest catchments. Journal of Hydrology, 295, 276-290, 2004. Grayson, R. and G. Blöschl, Spatial modelling of catchment dynamics. In: Grayson, R. and Blöschl, G. (Eds.): Spatial Patterns in Catchment Hydrology. Camebridge University Press, Camebridge, 51-81, 2000. Güntner, A., Large-Scale hydrological modelling in the Semi-Arid North-East of Brasil. Dissertation, Universität Potsdam, 2002. Hodnett, M.G. and J. Tomasella, Marked differences between van Genuchten soil water- retention parameters for temperate and tropical soils: a new water retention pedo-transfer functions developed for tropical soils. Geoderma, 108, 155-180, 2002. IAHS, Prediction in ungauged basins. International Hydrology today, 26-28, 2003. Jackson, R. B., Canadell, J., Ehleringer, J. R., Mooney, H. A., Sala, O. E. and E.D. Schulze, A global analysis of root distributions for terrestrial biomes. Oecologia, 108, 389-411, 1996. Jasper, K., Gurtz, J. and H. Lang, Advances flood forecasting in Alpine watersheds by coupling meteorological observations and forecasts with a distributed hydrological model. Journal of Hydrology, 267, 40-52, 2002. Kleinhans, A., Einfluss der Waldkonversion auf den Wasserhaushalt eines tropischen Regenwaldeinzugsgebietes in Zentral Sulawesi (Indonesien). Diss. SUB Göttingen 2004. Körner, C., Leaf diffusive conductances in the major vegetation types of the globe. In: Schulze, E. D. and Caldwell, M. M. (Eds.): Ecophysiology of Photosynthesis. Springer, Berlin, 463-490, 1994. Leemhuis, C., The impact of El Nino Southern Oscillation Events on water resource availability in Central Sulawesi, Indonesia. EcoRegio 21, 2005. Leemhuis, C., Erasmi, S., Twele, A., Kreilein, H., Oltchev, A. and G. Gerold, Rainforest conversion in Central Sulawesi, Indonesia: Recent development and consequences for river discharge and water resources. Erdkunde, 61, 284-293, 2007. Matthews, E., Global vegetation guide (1972-1983). http://www.daac.ornl.gov., 1999. Mo, X., Liu, S., Lin, Z. and W. Zhao, Simulating temporal and spatial variation of evapotranspiration over the Lushi basin. Journal of Hydrology, 285, 125-142, 2004. Piepho, B., Untersuchungen zum hydrologischen Reaktionsverhalten eines kleinen Einzugsgebietes unter Verwendung des Wasserhaushaltsmodells WaSiM-ETH. Dissertation, TU Braunschweig, 2003. Rehm, S., Spezieller Pflanzenbau in den Tropen und Subtropen. Ulmer, Stuttgart, 437-446, 1989. Schulla, J., Hydrologische Modellierung von Flussgebieten zur Abschätzung der Folgen von Klimaänderungen. Züricher Geographische Schriften, Verlag Geographisches Institut ETH Zürich, Verlag Geographisches Institut ETH Zürich, 1997. Schulla, J., Jasper, K., Model Description WASIM-ETH, 1999. Scurlock, J. M. O., Assner, G. P. and S.T. Gower, Global Leaf Area Index from Field Measurements, 1932-2000. http://www.daac.ornl.gov., 2001. Sobieraj, J. A., Elsenbeer, H. and R.A. Vertessy, Pedotransferfunctions for estimating saturated hydraulic conductivity: implications for modeling storm flow generation. Journal of Hydrology, 251, 202-220, 2001. Whitten, T. and G.S.M.M. Henderson, The ecology of Sulawesi.Periplus, Singapore, 2002. Wösten, J. H. M., Pachepsky, Y. A. and W.J. Rawls, Pedotransfer functions: bridging the gap between available basic soil data and missing soil hydraulic characteristics. Journal of Hydrology, 251, 123-150, 2001.

13 565