Spiny Forest Heterogeneity: Implications for Regeneration and Its Detection

Spiny Forest Heterogeneity: Implications for Regeneration and its Detection

Catherine Reuter

Advisor: Jules Ramangalahy Academic Director: Jim Hansen Spring 2009

1

Acknowledgements

I would like to thank Barry Ferguson for helping me originate the idea for this project and for innumerable resources from compasses to aerial maps. I would also like to thank Christian, translator and invaluable field assistant, without whose innovative thinking and possibly photographic memory this project would not of succeeded.

2

Table of Contents

Section Page 88

Acknowledgements ______2

Abstract ______4

Introduction ______5

Methods ______7

Results ______12

Discussion ______17

Conclusion ______26

Appendix 1: Comprehensive Species List ______27

Appendix 2: FTM 1954 Map of Forest Cover ______30

Sources Cited ______31

3

ABSTRACT

This study sought to verify claims made in a recently published paper by Thomas Elmqvist that certain portions of Madagascar’s spiny forest are rapidly regenerating. The study took place in the forest around the village of Manavy located in Central Antandroy, where historical and current images of the land cover do not collaborate Elmqvist’s conclusions. Using maps derived from remote sensed images of the area, quadrats were established within Elmqvist’s ‘regenerating’ area. Within these sites detailed vegetative analyses of species composition and regenerative process were performed as well as qualitative assessment of disturbance level. The results of this study indicate that density of plots as ascertained from remote sensed images did not correspond level of perturbation. The vegetative surveys showed extreme heterogeneity in the forest due to both natural and human causes, however, attributing this heterogeneity to any specific factor proved difficult. In light of these findings, many of Elmqvist’s methods appear inappropriate and his finally interpretation of the landscape as undergoing extensive regeneration unlikely. Future study that sought to untangle the web of human and natural forces informing forest quality is necessary before a statement about the regenerative capacity of the forest can be made.

4

INTRODUCTION

Scientists and conservationists have identified Madagascar as one of the world’s priority

Biodiversity hotspots (8). This categorization has its basis in the combined threatened nature of

Madagascar’s natural resources and their extreme diversity and uniqueness. Madagascar boasts an extreme diversity of “ecological communities and associated flora” (4), remarkable in its variety considering the size of the island (8). Estimates of species endemism reach as high as 83% for plants and in many cases still higher for animals (8).

No where is this diversity more apparent than in the southern spiny forests of Madagascar’s

Tandroy region, home to some of the jewels of Madagascar’s unparalleled evolutionary history. 95%

of the plants species of the southern dry forests are endemic and 48% of the genera (2). Its animal

diversity though less varied presents equally fascinating endemics with the notable Verreaux’s sifaka

and the radiated tortoise.

Despite its special nature, government authorities have historically neglected the region such

that as of 1995 less than 3.2% of the spiny forest is formally protected (3). Luckily, local traditions

have preserved many segments of forest, limiting resource use in areas of spiritual significance (2).

However, this local protection can not combat the human pressures on forest resources, which

consist primarily of the collection of firewood, charcoal, and construction materials for urban areas

(5). Though deforestation rates remain contested, it is generally believed that dry forest cover has

been in decline since the 1970s (2).

Sadly, the prognosis for recovery appears equally bleak, with many experts believing the

spiny forest, like many arid ecosystems, has little potential for large-scale natural regeneration due to

the region’s harsh climate and related slow growth rate of many of its endemic plants. However, a

recent paper presented in PLOS by the Swedish researcher Thomas Elmqvist and a team of

Malagasy scientists has placed this common belief in question, asserting that the spiny forest is

5 indeed capable of rapid natural regeneration (2). In their 2007 paper, Elmqvist investigated changes

in forest cover and their institutional context in the dry forests of southern Madagascar. Using remote sensed images spanning two decades, they tracked the presence or absence of forest looking

at three patterns: 1) loss of forest cover, 2) increase in forest cover, 3) and stable forest cover. As

part of his conclusions from this study, Elmqvist emphasized the capacity of the spiny forest for

“large-scale spontaneous regeneration dominated by native endemic species” and identified specific

regions where such regeneration had occurred (Figure 1, 2). He presents a scenario in which

extensive forest cover, noted on aerial maps of the region predating his images, was completely clear

cut and has since regenerated. This assessment of regeneration potential resulted in a flurry of

positive media which asserted that the spiny forest was recovering from human degradation (5, 6)

Unfortunately, “inferring historical patterns of forest cover is always likely to be subject to a

high degree of error” (9) and recent Google Earth images showing patterns of heavy vegetative

cover similar to those in the 1950s makes Elmqvist regenerative scenario seem improbable

(Appendix 2). Furthermore, questions surround the appropriateness of the methods Elmqvist used to

establish this scenario, satellite remote sensing of forest regeneration assessment being uncommon

(9). This study therefore sought to verify Elmqvist’s claims and determine the true state of the

forests of Central Androy between the towns Antanimora, Ambohimalaza, and Ifotaka. By exploring

the authenticity and process of forest composition and structure, this study investigated the present

state of the forests of Central Tandroy, their potential for regeneration, and how Elmqvist may have

reached his conclusions of increasing forest cover.

6

Figure 1. “Time-series analysis of changes in forest cover based on satellite images (dry season synoptic views from 25 June 1984 (Landsat 5 TM), 15 April 1993 (Landsat TM) and 28 May 2000 (Landsat 7 ETM+) from Androy, southern Madagascar. Three patterns of forest cover change between 1984 and 2000 is presented: 1) red-reddish areas-loss of forest, 2) blue-bluish areas–increase offorest cover and 3) green areas–stable forest. doi:10.1371/journal.pone.0000402.g005”; (Figure 5, Elmqvist et al. 2007)

METHODS

Study Area This study took place in the forests surrounding the fokotany of Manavy, a small community

of several small scattered villages. It is located just east of Route Nationale 13 between the town of

Antanimora to the north and the urban center of Ambovombe to the south. This site lies within reach

of several patches of forest denominated as regenerating by Elmqvist (Figure 2). As is the subject of

this paper, the landscape is a mosaic of fields and forests in a several stages of degradation and regrowth. Naturally, the area supports heterogeneous forest characterized by assorted endemic

species in the Didieraceae family and of Euphorbia. The canopy typically ranges between 4-5m with

protruding Dideraceae reaching 15m in some instances. The remaining vegetation consists of plants

well adapted to the region’s harsh semi-arid climate, displaying assorted water-saving modifications

7 such as minimized leaves, conversion of leaves to spine and photosynthetic stems as well as the

ability to store water in trunks and roots (8). Such modifications are necessary in an area that

receives on average less than 500 mm of rain per year and where drought is a constant threat (2). In

these forests the emblematic inhabitants of the south, radiated tortoises and lemurs such as

Verreaux’s sifaka and ring-tailed maki, abound.

Figure 2. This map segment is taken from Elmqvist map of changing forest cover (see Figure 1). The rectangle delineates the study area in relation to Elmqvist’s

The same semi-arid conditions that have shaped the natural environment play an important

role in the lives of the local human population. Villagers engage primarily in agriculture, cultivating corn, manioc, and sweet potatoes as well as raising livestock. The expansion of agricultural fields necessitates the slash and burn of forest cover and animals are grazed directly in the forest. In times of famine, the population heavily supplements their livelihoods via commercial harvest of forest resources, an option facilitated by the fokotany’s proximity to the national route and nearby urban areas. The villagers of Manavy exploit principally two endemic tree species: ‘katrafay’ (Cedrelopsis greveii) the preferred firewood and ‘fantiolotse’ (Alluaudia procera) used in the construction of boards.

In the past few months, Manavy underwent a “transfer de gestion” facilitated by ONG

Sokake in the early 2000s. This process placed responsibility for the management of local forests

directly into the hands of the community rather than the government. The WWF is currently

8 involved in the delineation of the new conservation and exploitation zones within the forest; this projects quadrats fall within three of these zones.

Field Research

Using Google Earth images from 2008 and 2009 (source file 2005), the vegetative cover of the land surrounding Manavy was divided into strata based on relative density; fields and villages were excluded from this categorization (Figure 3). This area falls in Area II, designated by Elmqvist as “Regenerating” (Figure 2). From these strata, quadrats were chosen using random stratified sampling in an attempt to investigate all land cover types. Each 30x30m plot was established using maps prepared from the Google Earth images and GPS. This plot size was chosen to be larger than

Elmqvist’s in order to capture a fuller picture of each forest parcel but still be manageable

considering constraints of time and expertise.

Least Dense In each plot, all woody plants over 2m were counted and

Medium Dense identified to species. Further measurements were taken for three Dense specific trees to provide a picture of age distribution and thus the

Figure 3. Original density classifications from Google possibility of regeneration in each plot: ‘fantiolotse’ (A. procera), Earth images (eye alt. 1.6km) used to esgablish radom ‘katrafay’ (C. greveii), and two species of ‘daro’ (Burceraceae stratified plots. family, Commiphora species). Fantiolotse and katrafay were chosen because they are generally most abundant, are targeted for exploitation, and to overlap with

Elmqvist’s study. Daro was chosen as a third indicator species because it has not been known to be targeted for commercial exploitation, is abundant and widespread, and should therefore indicate a stable population unchanged by human activity if the forest has remained at all intact.

For fantiolotse, height was estimated and DBH at 130cm was recorded for all individuals over 2m in height. A tally was taken of all juveniles less than 2m. For katrafay, height and DBH were recorded for all individuals greater than 20cm in girth, this being the size at which they become useful for construction. A tally was taken for all individuals less than 2m, including seedlings. For

9 the daros, two species, darofoty and daromena, were grouped together as they are undistinguishable with out a high level of familiarity and exhibit similar growth patterns. The same methods employed for katrafay were used for daro. No tally of young plants was taken as none were apparent during initial methodological trial.

In addition to these measurements, the canopy cover was determined for each plot. Two

diagonal transects connecting the four corners of each plot were performed. Every three meters

looking 90° from the ground, the percentage of sky obscured by vegetation was estimated. The

species of plant responsible for the obstruction was noted.

Finally, levels of disturbance from wood exploitation and grazing were qualitatively noted as

well as proximity to villages, fields and paths. Local guides were consulted to determine the social

history of each plot.

10

er

pond to A-D respectively. The Mark

led as such, plots 11-13 corres

of each corner plot .

this study. Plots 1-10 are labe

The 14 plots investigated for

Figure 4. corresponds to the NE GPS coordinates

11 RESULTS

Fourteen quadrats were explored in the area north and southwest of the town of Manavy

(Figure 4). These quadrats covered a range of land cover types, from largely untouched primary

forest to abandoned, regenerating agricultural fields Assessment of these plots indicates a wide

heterogeneity of forest composition and structure.

As will be discussed below, quantifying and ranking the levels of disturbance as well as attributing differences in forest quality to natural or human causes in each quadrat proved extremely difficult. Yet in the hope to allow for some comparison the following very general grouping of quadrats can be made as follows:

Table 1. General Grouping of Quadrats based on Disturbance Regime. Type Description Plots 1 Unaltered, Intact 7, 8, 9 Primary Forest 2 Disturbed by 1, 3, 4, Exploitation 5,10, 11, and Grazing 12, 13 3 Former field or 2*, 6, 14 village site *accounts of social history vary for Plot 2 see Appendix

These 3 groupings were used for gross comparison of numerical data but comparison is

hesitant because of the natural variation in plant composition particularly between Type 1 and 2. The

segments of undisturbed forest show a marked difference in the plants present such as the noticeable

decrease of fantiolotse in Type 1 and the increased presence of hazonta in Type 2.

Species Count

Species richness varied widely among plots with a low of 15 and a high of 39. Average

species richness showed a marked inverse relationship with level of disturbance, decreasing with

increased level of disturbance (Table 2). Total Density of woody plant likewise varied widely

12 among plots with no significant difference between Type 1 and 2, but both Type 1 and 2 were

significantly greater in plant density than Type 3.

Table 2. Species Richness and Total Density of Woody Plants Total In the plots surveyed, katrafay proved to be the most Density of Total Woody ___abundant species with 4694 individuals, followed by daro Number of Plants (h Woody >2m) per ___ (foty and mena) at 992, and fantiolotse with 813 (Appx 1). Species Ha-1 Type 1 Species Measurements Plot 7 39 9478 ______The measurements of specific trees, Fantiolotse, Plot 8 37 8478 Plot 9 30 8844 Average 35 8933 ____Katrafay, and Daro, also showed extreme variation (Tables Type 2 ____3, 4, 5 respectively). Plot 1 33 11489 Plot 3 15 9411 Plot 4 24 5611 ______To begin with, the total number of adult fantiolotse Plot 5 18 3589 Plot 12 23 11533 ____ranged from 73 to 0 in Type 1 plots, from 176 to 0 in Type lot 13 31 8411 Plot 10 20 12767 ____2 plots from 176 to 0, from 34 to 0 in Type 3. Average Plot 11 34 11367 Average 25 9272 ____numbers of juvenile fantiolotse (<2m) were equally variable Type 3 Plot 2 17 3244 ____ranging from 0 to 61 across the three plot types. Average Plot 6 19 3044 Plot 14 18 303 ____DBH and height were similar for Type 1 and 2 plots, but Average 18 2197 significantly lower in Type 3 plots when fantiolotse was

present (Table 3).

Adult katrafay were present in all Type 1 and Type 2 plots, though on average, Type 2 plots

exhibited twice as many as adult katrafay as in Type 1. Katrafay populations were greatly reduced or

not present in Type 3 plots. There was extreme variation in the population of juvenile katrafay (<2m)

between plots even in the same disturbance type. Finally, the numbers of katrafay with DBH >20cm, their, DBH and height was similar between both plot and disturbance type (Table 4).

13 Table 3. Measurments of Fantiolose (Alluadia procera) by disturbance type and plot. Type FANTIOLOSE Type 1 Type 2 3 Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot 7 8 9 Avg 1 3 4 5 10 11 12 13 Avg 2 6 14 Avg # Individuals > 2m 73 10 0 28 109 112 106 86 0 0 176 87 85 42 2 10 18 # Individuals < 2m 60 6 N/M 33 45 61 45 52 N/M N/M 25 21 42 17 N/M 45 31

Avg DBH (cm) 36 36 N/M 36 33 32 40 38 N/M N/M 45 36 37 N/M N/M 25 25 Avg Height (m) 6 6 N/M 6 6 5 6 5 N/M N/M 5 6 6 N/M N/M 4 4 N/M – Not measured N/P – Not Present

Table 4. Measurments of Katrafay (Cedrelopsis greveii) by disturbance type and plot. KATRAFAY Type 1 Type 2 Type 3 Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot 7 8 9 Avg 1 3 4 5 10 11 12 13 Avg 2 6 14 Avg # Individuals > 2m 217 180 338 245 442 556 251 86 977 685 673 246 490 9 0 34 14 # Individuals < 2m 125 547 817 496 182 291 122 151 1233 754 447 31 401 N/M N/P 106 106 # Individuals > 20cm DBH 9 34 9 17 12 12 14 12 3 7 18 30 14 N/M N/P 5 5 Individuals > 20cm DBH Avg DBH (m) 24 25 23 24 227 22 24 25 21 22 22 23 48 N/M N/P 22 22 Individuals > 20cm DBH Avg Height (m) 5,1 5,2 4,3 4,9 5,8 4,5 4,4 4,6 3,8 4 4,8 5 4,61 N/M N/P 3,5 3,5 N/M – Not measured N/P – Not Present

14 Table 5. Measurments of Daro by disturbance type and plot. DARO Type 1 Type 2 Type 3 Plot Plot Plot Plot Plot Plot Plot 8 9 Avg 10 11 12 13 Avg 14 # Individuals > 2m 92 64 78 11 41 87 149 72 9 # Individuals > 20cm DBH 35 13 24 5 9 48 48 28 4 Individuals > 20cm DBH, Avg DBH (m) 29 28 28 27 27 30 30 29 22 Individuals > 20cm DBH, Avg Height (m) 3,6 3,4 3,5 3,1 3,2 4,1 3 3 3.5

Finally, for daro, measurements were only taken for Plots 8 through 14, limiting comparative

value (Table 5). Total population of adult daro varied widely even between plots with similar

disturbance types. The same was true for the numbers of individuals with DBH >20cm. However,

the average height and DBH of these individuals was comparable among plots and disturbance types.

Canopy Cover

Canopy cover for Type 1 plots proved almost twice as dense as for Type 2 plots, with more than half the sky being obscured by vegetation 62 % of the time compared to 35 %. However, Type

3 plots averaged greater than 50% cover 45% of the time (Table 6). There does not appear to be a consistent relationship between initial density categorization from remote sensed maps and canopy cover. Plots originally delineated as ‘dense’ averaging 55% for their final canopy cover percentage and ‘medium dense’ plots averaging 50%. The ‘least dense’ plots however did average a significantly lower final canopy cover value of 22%.

Among all plots, katrafay proved by far the most frequent source of canopy cover appearing

112 times, followed by darofoty (11), and then sognogne (15). Fantiolotse appeared only 9 times.

Otherwise a large variety of species were responsible for the canopy cover, most appearing less than

15 five times. Between disturbance types, katrafay was still appeared most frequently in Types 1 and 2 though only once in Type 3 where famanta was noted most frequently. Fantiolotse was most responsible for canopy cover in Type 2 plots, being recorded once in Type 3 and not at all in Type 1.

Table 6. Canopy Cover Original Remote-sensed Remote-sensed Original

Disturbance Disturbance Type Plot Categorization Density 0 1 2 3 4 5 TotalforCategories coverage or <50% 0-2 TotalforCategories coverage or >50% 3-5 is canopy of time % 50% covered than more Dense 2/22 1 7 4 1 3 6 6 7 15 68 Med. 8 Dense 4 3 2 2 4 7 9 13 59 Med. 9 Dense 4 4 1 3 4 6 9 13 59 Average 62 (Table2 7).1 Dense 5 7 0 4 6 0 12 10 45 3 Dense 8 5 1 3 5 0 14 8 36 Least 4 Dense 11 6 1 3 1 0 18 4 18 Least 5 Dense 15 3 1 0 3 0 19 3 14 Med. 10 Dense 8 4 2 4 1 3 14 8 36 Med. 11 Dense 7 3 2 4 4 2 12 10 45 12 Dense 3 6 1 3 4 5 10 12 55 Least 13 Dense 4 9 2 0 4 3 15 7 32

Average 35 Least 3 2 Dense 10 6 1 3 2 0 17 5 23 6 Dense 6 1 0 1 9 5 7 15 68 Average 45

Canopy cover was estimated along two transects for each plot. Looking 90° from the ground, the percentage of sky obscured by vegetation was estimated using the following scale. 0 = 0% 3 = 50-75% 1 = 0-25% 4 = 75-95% 2 = 25-50% 5 = 95-100% This table shows the frequency of each canopy cover category by plot and a final canopy cover value that equals the average percent of time that the sky was more than 50% obscured by vegetation.

16 Table 7. Species Responsible for Canopy Cover

Total by Disturbance Type Species 1 2 3 Total Darofoty 5 17 0 22 Famanta 0 2 7 9 Fantiolotse 0 8 1 9 Hazomena 7 2 0 9 Hazonta 0 4 6 10 Katrafay 26 85 1 112 Rohondroho 4 9 0 13 Sognogne 9 6 0 15 Zanampoly 0 2 4 6

This table shows the frequency individuals of a species were responsible for canopy cover. The species included are those that appeared over 5 times and were found in more than a single plot.

DISCUSSION

Heterogeneity and Disturbance

Any discussion of the numerical results from this study must be undertaken with a full

appreciation for the context in which they were recorded. The main sources of anthropogenic

disturbance in the area of study are selective logging of trees and grazing of livestock. Evidence of

these activities was ubiquitous for all plots; not a single site demonstrated a complete absence of

human presence and manipulation. This evidence ranged from a couple of chopped trees stumps or a

couple of cow patties to dozens of felled trees and large paths running through sites. Moreover,

regardless of the level of activity within the site, heavy disturbance and human activity was never far away. A village was never more than 2km distant and a path never more than 50m. The human fingerprint on the landscape was inescapable.

Quantifying this human footprint, identifying how much cutting or grazing has occurred, is

largely impossible. Knowledge of the timing of disturbance events is also highly limited to the

immediate. Local information on the social history of the areas in question can be helpful but here of

course comes the fallacy and the limited length of human memory. It is thus very difficult to

determine whether the quality of the forest in question, its openness for example or diversity, is the

17 result of natural or human pressures. This is a key observation in a discussion of process as it

complicates the number variables that may be responsible for an observed trend. For example,

elucidating whether a high number of juvenile trees in the result of natural factors, such as soil type

or water availability, or human degradation prove problematic when human presence is so pervasive.

Instead of saying therefore what knowledge of patterns of human perturbation can provide, I

will begin with what it fails to explain. Our results indicate that level of disturbance is not visibly a

cause for decrease correlate to a specific level of species richness or abundance. Referring to Table 2

and Appendix 1, we see extreme variability in the number as well as kind of species present between

all quadrats. Fantiolotse is an excellent example of this. Though one of the most highly exploited

species in the forest it remained the most prolific in all but the most disturbed sites. Yet its population decreased or disappeared in the most intact sites. This variation appears to be natural, with little signs of wood exploitation to suggest the former existence of fantiolotse in the area.

This kind of variation hinders any assessment of the authenticity of species composition of

the forest disturbed or not. For example, it was the opinion of French colonial administrators in the

1950s, that the “Formations de degradation” were treed savannahs characterized by bushy vegetation

and the following species: halomboro, famoty [possibly famanta], hazonta, filofilo, komptse,

pisopiso, lamonty, sakoa; and fano (277). While a worker of the WWF indicated that highly

disturbed regenerating land is characterized by hazonta (Rhigozum madagascarensis), kirava

(mimosa delicatula), andrapary and sely (Grevia spp.) and famanta (Euphorbia laro) (7). Yet, many

of these species, such as hazonta and andrapary, were found in significant numbers in the less

disturbed plots (Appendix 1). The only species that appears to follow expectation is famanta, which

exhibited a greater density in each of the Type 3 plots than in any of the Type 1 or Type 2 plots.

Thus, authenticity forest sites is extremely difficult to elucidate and level of disturbance is no

guarantee of what species will be found on a site and reciprocally, no species is an indication of the

quality of a forest plot.

18

Process Considering the variability of the data understanding of the regeneration capacity of the spiny

for remains ambiguous yet some statements about the potential for regeneration can be made from

our investigation of the age distribution of daro, fantiolotse, and katrafay.

Katrafay remains the most abundant tree in the forest despite its being one of the most

heavily exploited. It was highly prolific in all types of plots, with juveniles numbering in the

thousands per hectare in some cases. However, considering the variability of the data obtained, no

statement can be made about, disturbance and these high values. However, the small proportion of

large katrafay >20cm in DBH and smaller adults appears natural and not the result of human activity

thee total population with DBH >20cm, their average DBH and their height was similar between all

plots and disturbance type. The disproportion is likely due to the trees ability to quickly obtain

height but to slowly grow in girth. Katrafay was present in reduced numbers in Type 3 plots

indicating that its ability to grow after this type of disturbance may be limited. This will be discussed

more fully below.

Fantiolotse does not possess the same prolificacy as katrafay with fewer individuals per plot.

Here again the variation in numbers of adults and young allows and the lack of correspondence with

disturbance regime allows no statement to be made about the regenerative process. As those sites least disturbed exhibited what appears to be a naturally lower population of fantiolotse, it would be

misleading to draw conclusions by comparison with other disturbed plots.

According to local knowledge, both of these species require nearby presence of adult individuals for the substantial propagation of seedlings. In a clear cut site only those seeds that happen to remain in the seed bank will grow. This necessity for an established tree to facilitate regeneration of a population is supported by the pattern of young katrafay in Type 3 plots. There was an appreciable number of young katrafay in Plot 14, where a remnant of adult primary forest trees existed, unlike in Plots 2 and 6.

19 Reassessment of Elmqvist et al.

The data collected from this study supports careful reevaluation of Elmqvist's methods and

interpretation of his results.

Remote-Sensing

Elmqvist based his claims for regeneration on a series of Landsat TM/ETM images from the years 1984, 1985, and 2000. Researchers are increasingly using remote sensing as a tool for characterizing land cover, however, it application is not appropriate for all forest types and goals (9).

Based on my study, Elmqvist’s use of remote sensing to identify regenerating forest appears problematic.

Firstly, his team visually categorized forest land cover by density, using NDVI as an indicator. However when we used remote sensed Google Earth images which are of a higher resolution than Landsat images, apparent density of the plots did not provide a good indication of

forest quality. As has been discussed above, the forest in question demonstrates gradient of composition and structure both from natural and human processes, a range which is not easily perceivable from satellite imagery. For example, both Plots 1 and 7 were originally categorized as dense from Google Earth images but upon ground-truthing Plot 1 exhibited high levels of wood exploitation and Plot 7 next to none. Both contained very different dominant plant species as well.

Another striking example of the potential for mischaracterization based on satellite imagery alone lies in Plot 6. Quadrat 6 is a prime example of regenerating vegetation but instead of appearing

“patchy” as described by Elmqvist (2), it appeared extremely dense via remote sensing. Furthermore, the species regenerating were not primary forest species, a fact undetectable from remote sensed images.

There are several possible explanations for this disconnect between image density and forest

quality. The first is the general patchiness of the spiny forest. The canopy is not dense with frequent

20 gaps. This openness is largely the result of the tree species that characterize the spiny forest, the

great majority of which have leaves less than 2 cm2. This is true for two of the three most abundant

trees surveyed, fantiolotse and daro. The majority of canopy cover that is viewed is the result of

katrafay, the most abundant tree surveyed, which also has slightly larger leaves and a sizable crown,

around 4cm2. In short there is not a lot of green vegetative density which allows imagery to easily

penetrate the canopy. While vegetation less than 2m and therefore not part of the canopy was not

counted, general observation and opinion of guides shows that bush cover increases with disturbance

thus implying that perhaps disturbed areas, which should correlate with regeneration, would exhibit

a greater density of green vegetation obscuring the ground, not increased patchiness. Also even in intact forest the soil is frequently exposed, the leaf litter of the spiny forest is very thin and quickly disappearing with perturbation whether natural or manmade. Thus the chances of imagery reaching the ground are very high and resultant patchiness thus not necessarily symptomatic of regeneration.

Moreover, as mentioned above, even highly degraded forest may appear dense via satellite

imagery. This is perhaps again a result of the open canopy cover and the character of the forest’s

plants. For example, a forest parcel with extensive fantiolotse exploitation may still appear dense.

This is possible because fantiolotse, despite its abundance in certain settings, does not make up a

high percentage of the canopy cover due to his small leaves and near vertical branches. Felling of

these trees does not drastically change the openness of the canopy though it may drastically change

the character of the forest. Interpreting the dense patch of forest thus depends heavily on what

species are responsible for the apparent density. This question can only be answered via ground-

truthing and will be discussed below

Elmqvist used NDVI values (Normalized Difference Vegetative Index) to support his remote

sensed forest classifications. However, this type of analysis is complicated by many of the same

variables as image density. NDVI values use “reflectivity in the red and near infrared parts of the

electromagnetic spectrum. These bands are particularly sensitive to the density of green vegetation

21 and the absorption of chlorophyll; and provide a high degree of contrast between vegetation cover and the soil surface” (9). As discussed above natural spiny forest with its trees adapted to dry climate has a very patchy canopy cover, characterized by small leaves and photosynthetic stems. Such composition and structure would result in a much lower NDVI values. Meanwhile, as mentioned, disturbed areas tend to exhibit a greater density of bushy plants which have greater numbers of leaves. Thus disturbed or regenerating areas may exhibit increased NDVI values. A great example of this is again Plot 6 which appeared very dense and very green via remote sensing and would most likely have a high NDVI. Following Elmqvist’s definition, such a plot should be stable forest, but instead it is regenerating secondary vegetation with no resemblance to a natural spiny forest community. The example of exploitation in a fantiolotse stand also undermines the use of NDVI values to characterize forest quality. Because these trees make up such a small percentage of the forest canopy and the green vegetative cover, their presence, non-presence, and size may not contribute significantly to the NDVI value.

Ground-Truthing

Fantiolotse as an Indicator Elmqvist was not unaware of the possibility of misinterpretation of forest if based solely on remote sensing and performed a number of plots to verify his classifications. His data supported his regenerating/stable classifications, but unfortunately, in light of what was seen in my study, his methods and results fail to appreciate the variation and nuance of the spiny forest landscape and therefore his results do not provide an accurate picture of the landscape.

To begin with, Elmqvist’s plots “were located randomly within stands of Alluaudia procera”

(2) and used the population of this tree as his primary indicator for regeneration. However, as we have seen, the presence or absence of Alluaudia procera is in no way an indication of forest quality or even of forest. Multiple sites (Plots 8 and 9) situated squarely in his “regenerating forest” zones contained little to no fantiolotse while simultaneously exhibiting the lowest levels of disturbance

22 past or present. General observation of the surrounding land supports that these plots are not anomalous. Other quadrats, though less pristine (Plots 10 and 11), contained a species assemblage

(mostly rohondroho and katrafay) in which fantiolotse was conspicuously absent. Local attribution of this variation to soil needs scientific verification, but the absence of stumps and the dead wood common in other sites of fantiolotse exploitation support the natural existence of communities free of this species. Even in plots where Alluaudia procera was present in mass, the number of individuals varied widely, with no correlation to level of disturbance as discussed above. A notable instance of this involves Plots 3 and 1 which had similar numbers of adult fantiolotse (109 and 112 respectively) and yet drastically different levels of wood exploitation.

Moreover, considering the limited contribution of fantiolotse to forest structure as discussed above section concerning remote-sensing, it seems an unlikely choice for assessing forest quality. In short, targeting Alluaudia procera as an indicator of regeneration and of forest in general ignores the diversity of the habitat in question and creates a skewed image of forest quality. Furthermore, considering the level of heterogeneity observed, relying on any single species appears ill-advised.

Young Trees as an Indicator Another aspect of Elmqvist’s ground-truthing methods is his reliance on the proportion of young trees as an indication of regeneration. Elmqvist asserted that in stable forest “for juvenile (<2 m height) A. procera (mean ±s.e.=2136124 ha21 , n= 4) and C. grevei (mean±s.e. =107694 ha21 ,n

=4) density was significantly lower than in area II [regenerating forest] (Mann-Whitney U-test,

P,0.02, P,0.05, respectively, Table 1). Our data exhibited such variation that we cannot support this trend. Counts of juvenile katrafay were too variable and counts of young fantiolotse too similar to determine any trend in the values. Furthermore, understanding of the natural history of the trees investigated, A. procera and C. greveii, remains limited, and their response to opening of the forest due degradation uncertain. Considering the variation in numbers observed it is possible that

23 increased juvenile trees are not a response to disturbance and the most appropriate indication of

regeneration.

Size of Trees Here again, Elmqvist used fantiolotse as his primary indicator for regeneration. The same

reasoning against using simply fantiolotse as a sole indicator applies here as well. Average height

and DBH was so variable and show no correspondence to disturbance such that we cannot support

the use of such data as indicative of the age or regeneration of forest plots. Analysis of the

fantiolotse population would likely be complicated by it high level of exploitation which

undoubtedly has altered the age distribution.

Looking at the size and age of other trees may help serve as a better index of the age of a

forest as their populations are more likely to reflect a natural distribution if the forest is stable and

younger distribution if the forest is regenerating. This is what we attempted to do by measuring daro

as well. While the number of daro varied drastically between sites to allow an average for

comparison average it was observed that in sites with fantiolotse exploitation, large daro still were

present, indicating that the forest was possible of significant age even if all of its subpopulations

proved not to be.

Possibility of Regeneration This study evinced little support for Elmqvist’s claims of large scale natural regeneration of

the spiny forest. To begin, there is little evidence of substantial change in forest cover in the area

based on local opinion. Elmqvist describes areas of increased forest cover as regenerating, based of

the absence and then presence of forest cover between Landsat images. This pattern of ‘no forest’ to

‘forest’ is not a probable in the area studied for a number of reasons. Firstly, the area in question was designated as forest in the mid-1950S (Appendix 2). Our guides, one of whom was over 75 years old, also asserted no knowledge of areas of forest where there had not been previously and acknowledged rather that the forest was slowly shrinking. According to them, all quadrats visited,

24 save those specifically identified as former fields, were in their natural state, differences in openness or species composition being the result of continuing degradation from primary forest or natural variation. Furthermore, the WWF specifically targets primary forest for its conservation zones and multiple segments of the forest in question fall within the zones recently established for protection.

Finally, our vegetative surveys did not support the possibility of the regeneration of primary

forest species on clear-cut land, at least in such a short time scale as suggested by Elmqvist. As

discussed previously, all plots located on sites known to be former agricultural or village sites,

exhibited species compositions completely different from spiny forest. Locals, the chef de

cantonnement forestier, and a representative of WWF corroborated the inability of primary species

to colonize such land. Moreover, each of these sites was already at least eight years old. Considering

that Elmqvist’s images covered less than twenty years it seems unlikely that he would have observed a different pattern.

Suggestions for Future Study

As has been discussed, this study was informed by the human presence ubiquitous across the

study site. This pervasiveness meant that the study lacked a natural model for regeneration for which

to compare to the patterns observed and therefore could draw few conclusions about how natural

processes are responding to human activity. Future studies should seek to incorporate pristine sites to

provide this basis for comparison. Sacred forests, where human activity has historically been limited

as well as monitored, could serve as appropriate sites for such research.

Correspondingly, with human pressures so omnipresent in the landscape, it would be of value

for future studies to in the area to more rigorously quantify them. This study took note of disturbance

on a qualitative basis which made ranking and comparing plots by degree of perturbation difficult. A more intensive look at specific disturbance activities, such as the exploitation regime of katrafay, would help illuminate their direct effects on forest quality.

25 Finally, a poor understanding of causes for natural variation in the landscape proved a

handicap to this study. It resulted in a limited capacity to differentiate between human and natural

causes of habitat diversity. For example, locals frequently sighted soil as the factor behind species composition at a specific site. Scientific verification of this claim would help explain the heterogeneity observed in this study. Other possible research could do the same with other variables such as water availability or aspect.

CONCLUSIONS

After quibbling about appropriateness of indicators or location of quadrats, the question remains: Is the spring forest really regenerating? Unfortunately, a simple yes or no is not possible.

Such an un-nuanced question cannot embrace the full spectrum of natural composition and

disturbance present in the spiny forest. A lack of consideration for heterogeneity is the primary flaw

in Elmqvist’s claim of large-scale spontaneous regeneration. His broad definition of regeneration as

non-forest to forest presents a simplified scenario that the social history and numerical data obtain in this study does not support. What is seen is a gradient of gradual forest disturbance and variability of species and age distributions which with the data from this study cannot be linked. Questions of disturbance and subsequent regeneration remain cloudy. Proper methodology for assessment and a solid understanding of natural process of this heterogeneous environment continue as the primary obstacles to determining the current state of the spiny forest. Thus, although promising miraculous regenerative powers of the forest proposed by Elmqvist appear unlikely at best, it remains to be shown whether a more moderated regeneration regime that takes into account the natural and human induced variation of the landscape is possible.

26 APPENDIX 1.

Complete woody species and individual count for all quadrats. Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Plot Species 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Total 1 Amanimbalavo 0 0 0 0 0 4 0 3 0 0 0 0 0 0 7 2 Ambilazo 0 0 0 0 0 0 0 0 64 0 0 0 0 0 64 3 Andrabe 0 0 0 0 0 8 0 0 0 0 0 0 0 0 8 4 Andragnoto 0 0 0 0 0 0 0 0 0 0 1 0 3 0 4 5 Andrapary 0 0 0 2 1 0 36 3 2 0 0 1 3 0 48 6 Aolilolo 13 0 0 0 5 5 0 0 0 0 0 1 0 0 24 7 Befoetse 12 0 0 5 0 0 10 26 0 0 2 11 14 0 80 8 Beholitse 4 0 6 6 13 0 9 4 0 0 0 21 7 0 70 9 Belavenoke 1 0 0 1 0 0 1 0 0 0 0 0 0 0 3 10 Borodoke 0 0 0 0 0 0 3 1 9 0 2 0 0 0 15 11 Dagoa 1 0 0 0 0 0 0 3 0 1 0 1 1 0 7 12 Darofoty 142 25 81 89 70 0 132 92 64 11 41 87 149 9 992 13 Daromena 11 0 0 0 0 0 0 0 0 0 0 0 0 0 11 14 Darotandroke 0 0 0 1 4 0 32 3 11 0 2 3 8 1 65 15 Famanta 0 55 1 0 10 93 0 0 0 5 0 0 2 15 181 16 Fandreandambo 0 1 0 0 0 0 0 10 8 18 31 0 3 0 71 17 Faniatse 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 18 Fantiolotse 109 42 112 106 86 2 73 10 0 0 0 176 87 10 813 19 Farehetse 0 0 0 0 0 0 0 0 2 0 0 0 0 0 2 20 Farohiosy 2 0 0 0 0 0 0 0 8 0 2 0 0 1 13 21 Fatikakoho 0 0 0 0 0 0 0 0 9 0 11 0 5 0 25 22 Fatra 0 0 0 1 0 0 11 0 1 7 1 0 1 1 23 23 Fihagne 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 24 Filofilo 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 25 Forofoke 0 0 0 0 0 0 3 6 0 0 0 0 0 0 9 26 Forokoko 0 0 0 0 0 0 0 8 0 0 0 0 0 0 8 27 Forombitsike 0 0 0 0 0 0 0 0 0 0 0 0 2 0 2 28 Halomboromahalao 0 0 0 0 0 0 0 0 0 0 5 0 0 0 5 29 Harahake 0 0 0 0 0 0 1 1 0 0 0 0 0 0 2 30 Hararafy 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 31 Harihy 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 32 Hataemarandoha 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 33 Hatakatakafoty 0 0 0 0 0 0 0 0 0 0 21 0 0 0 21 34 Hatakatakamena 0 0 0 0 0 0 0 0 0 0 9 0 0 0 9 35 Hatambonagne 0 0 0 0 0 5 59 6 21 1 0 0 0 0 92 36 Hazoavy 0 0 0 3 0 1 1 3 16 3 5 0 0 0 32 37 Hazobe 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 38 Hazofoty 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 39 Hazolava 0 0 0 1 0 0 0 0 0 0 0 0 1 0 2 40 Hazomena 0 0 43 0 0 4 0 133 0 0 34 0 0 0 214 41 Hazonta 55 127 35 2 13 2 5 15 2 14 18 5 9 3 305 42 Hentogne 0 0 0 0 0 0 93 0 7 0 2 0 0 0 102 43 Hiligne 7 0 0 0 0 0 0 0 0 0 0 0 3 0 10 44 Jabihy 5 0 0 2 0 0 37 7 26 22 13 9 11 0 132 45 Katrafay 442 9 556 251 86 0 217 180 338 977 685 673 246 34 4694 46 Kibaigne 1 0 0 0 0 0 0 7 0 0 2 3 0 0 13 47 Kibay 0 0 2 5 0 2 0 0 0 0 0 0 0 0 9 48 Kadrakitse 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 49 Kirava 0 0 0 0 0 6 0 1 0 0 5 0 0 17 29

27 50 Kopaipoty 0 0 0 0 0 0 2 1 0 0 0 0 0 0 3 51 Lahivozake 1 0 0 0 0 0 0 2 0 0 0 0 0 1 4 52 Lamontinamboa 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 53 Laza 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 54 Lazagne 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 55 Mafelefogne 0 0 0 0 0 0 2 0 0 0 0 0 0 0 2 56 Magne 0 0 0 0 0 0 9 1 0 0 0 0 0 0 10 57 Magnendrake 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 58 Mampandre 0 0 0 0 0 0 0 4 2 0 0 0 0 0 6 59 Manateza 0 1 0 0 1 0 7 3 3 1 8 0 0 0 24 60 Manongo 0 0 0 1 0 0 3 0 0 0 0 0 0 0 4 61 Mantsake 3 5 1 3 0 0 11 3 19 5 8 0 11 0 69 62 Marandoha 0 0 0 0 0 0 0 0 0 0 20 8 2 0 30 63 Maroambake 0 0 0 0 0 0 0 0 0 0 2 1 0 0 3 64 Mendoravy 1 0 1 2 2 0 3 0 5 0 0 2 2 4 22 65 Mompandry 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 66 Mozotse 0 0 0 0 0 0 0 0 15 0 0 0 0 0 15 67 Nato 0 0 0 0 0 0 0 0 2 3 2 0 0 0 7 68 Rarafy 0 0 0 0 0 0 0 0 2 0 0 0 0 0 2 69 Raketanosy 0 4 0 0 0 4 0 0 0 0 0 0 0 1 9 70 Robika 0 0 0 0 0 32 0 0 0 0 0 0 0 0 32 71 Rohavitse 6 0 2 0 0 0 0 0 0 0 0 0 0 0 8 72 Rohondroho 40 0 0 1 5 0 25 32 22 70 54 13 51 9 322 73 Sakoandalitse 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 74 Saranga 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 75 Samosa 0 0 0 0 0 0 0 0 13 0 0 0 0 0 13 76 Sasavy 0 1 0 0 0 0 0 0 15 0 2 0 0 0 18 77 Sely 2 0 0 0 0 0 3 0 6 0 0 0 0 0 11 78 Sengatse 0 0 0 0 0 0 0 0 0 0 1 0 6 0 7 79 Sognogne 57 0 1 4 4 0 31 169 8 3 2 7 62 1 349 80 Solohotse 0 0 0 0 0 0 3 3 0 0 0 0 0 0 6 81 Somangy 3 0 0 1 0 0 8 0 6 3 0 2 0 0 23 82 Somontroy 0 0 0 0 4 0 0 0 0 0 7 0 0 0 11 83 Tabarike 21 11 0 0 7 3 4 6 28 0 0 4 0 9 93 84 Tadrakitse 0 0 2 7 2 3 4 2 15 0 11 3 0 0 49 85 Tagnatagnanala 68 0 0 2 0 3 1 0 0 0 11 3 4 0 92 86 Taholakanfotse 4 0 2 0 0 1 0 0 0 1 0 0 0 0 8 87 Taimbitike 0 0 0 8 0 0 1 0 0 0 0 0 16 0 25 88 Tainorogne 0 0 0 0 0 0 0 8 33 0 0 0 0 0 41 89 Talintivoke 0 1 0 0 0 0 0 0 0 0 0 2 12 0 15 90 Taly 0 4 0 0 0 0 0 1 0 0 0 0 0 0 5 91 Taliporokoko 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 92 Tandalilaly 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 93 Tatavy 0 0 0 0 0 0 0 0 12 0 0 0 0 0 12 94 Taolankafotse 0 0 0 0 0 1 0 0 0 0 0 0 3 0 4 95 Tratrioltse 5 0 0 0 0 0 0 0 0 0 0 0 0 0 5 96 Tsiambara 2 0 0 0 0 0 0 0 0 0 0 0 0 0 2 97 Tsiirimbitike 5 0 0 0 0 0 0 0 0 0 0 0 0 0 5 98 Tsintsigne 0 0 0 0 9 0 1 2 0 0 0 0 0 0 12 99 Tsingena 1 2 0 0 0 0 0 0 0 0 0 0 0 0 3 100 UNKNOWN 3 0 0 0 0 0 0 0 0 0 2 0 1 2 8 101 Vahogne 0 2 0 0 0 0 0 0 0 0 0 0 0 0 2 102 Vahombe 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 103 Valeandro 0 0 0 0 0 0 3 2 0 0 0 0 0 0 5

28 104 Vontake 0 0 0 0 0 0 0 0 0 0 1 0 1 0 2 105 Vaovy 0 0 0 0 0 0 6 1 0 0 0 0 0 0 7 106 Varemamoa 0 0 0 0 0 0 0 0 0 0 0 2 0 2 4 107 Vavatoa 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 108 Zanampoly 0 0 0 0 0 94 0 0 0 0 0 0 31 0 125 109 Zira 5 0 0 0 0 0 0 0 0 0 0 0 0 0 5 TOTAL INDIVIDUALS 1034 292 847 505 323 274 853 763 796 1149 1023 1038 757 121 9775 TOTAL SPECIES 33 17 15 24 18 19 39 37 30 20 34 23 31 18

29 APPENDIX 2

FTM Map of Forest Cover, feuille j-62 de 1955. Map is based off of ground-truthed aerial photos.

30 SOURCES CITED

1) Chef de Cantonnement Forestier. L’Etude Monographique de la station de Reboisement d’Antanimora. Antanimora, Madagascar, 1955.

2) Elmqvist T, Pyyko¨nen M, Tengo¨ M, Rakotondrasoa F, Rabakonandrianina E, et al. « Patterns of Loss and Regeneration of Tropical Dry Forest in Madagascar: The Social Institutional Context.” PLoS ONE. Vol. II; No. 5: e402. 2007. <>.

3) Fenn, M.D. “The Spiny Forest Ecoregion.” in Goodman, Steven M. and J.P. Benstead.The Natural History of Madagascar. Univeristy of Chicago Press. 1525-1530.

4) Gautier L. and S.M. Goodman. “Introduction to the Flora of Madagascar.” in Goodman, Steven M. and J.P. Benstead. The Natural History of Madagascar. Univeristy of Chicago Press. 229-250.

5) “Madagascar's Forests are Recovering.” 2008. <>.

6) “Madagascar’s Forests Regenerate Against Expectations.” 2007. <>.

7) Mahawotahy, Sylvian. WWF worker. Manavy. 16 April 2009.

8) Mittermeier, Russel A. et al. “Wilderness Conservation in a Biodiversity Hotspot.” International Journal of Wilderness. Vol II : No. 3. 2005. 42-46.

9) Newton, Adrian C. Forest Ecology and Conservation: A Handbook of Techniques. Oxford University Press: Oxford, 2007.

10) Villagers. Manavy, Madagascar. 6 April thru 23 April 2009.

31