ONLINE RESOURCES: REGIONAL ENVIRONMENTAL CHANGE
Title: Changes in the dry tropical forests in Central India with human use
Agarwala, M.1,2, DeFries, R. S.1, Qureshi, Q.2, Jhala, Y. V.2
1 Department of Ecology, Evolution, and Environmental Biology, Columbia University, 1200 Amsterdam Avenue, New York, NY 10027
2 Wildlife Institute of India, Chandrabani, Dehradun 248001, India.
Email address of corresponding order:
S1. QUANTIFYING FOREST COVER IN STUDY REGION
We used Landsat imagery (Path: 143, Row: 045, acquired on Dec 07, 2009; Path 143, Row: 044, acquired on Dec 7, 2009, and Path: 144, Row: 045, acquired on Jan 31, 2010) from winter 2010 to classify forest cover in the study region. We used iMAD to radiometrically normalize these images (Canty and Nielsen 2008), and classified forest and non-forest using supervised classification. Training data was based on seventy-six 1-km long transects conducted in the area in 2009-2010, where land cover (forest, non-forest) and forest type was recorded every 100 meters. Accuracy of classification based on a validation dataset randomly selected from training data was 93.02%. Following this, we used ArcGIS 9.3 to quantify available forest cover in 2-km buffers around each village using ArcGIS 9.3.
Figure S1: Study area and sampling to quantify forest cover
S2. SAMPLING STRATEGY
To select representative villages for sampling, data on 1125 villages in the study region were compiled from various sources: village-level livestock population (Department of Animal Husbandry 2012), human population, distance to market (Wildlife Institute of India 2011), distance between village and forest, available forest area (explained in S1), forest type and its historical legacy of management (Madhya Pradesh Forest Department 2011). For historical legacy of management, we obtained detailed compartment history of each forest compartment from the Divisional Forest Offices, and classified forest compartments based on management plan followed and the year of implementation in the management cycle. For distance between village and forest, we did not use metric units but based distance on the number of villages between forest and village. A village surrounded by forests on all sides was given a value of 0, and a village where half its perimeter was forest was given a value of 1. For villages that were not themselves adjacent to forests, a village that had only one village between it and the forest was given a value of 2, and those with more than one village between it and the forest were given a value of 3. Distance to market was obtained from existing census (Wildlife Institute of India 2011), and this distance was based on distance to market by road.
Cluster analysis (using Ward’s method) was used to cluster villages according to similarity. Six villages were randomly selected from three of the largest clusters (two villages from each cluster: details in Table S2).
Table S2: Details on Villages Selected, based on data collected by Wildlife Institute of India (Total Population, Nearest Town, Distance to Town) and Department of Animal Husbandry (Cattle and Buffalo populations).
Cluster / Village# / TotalPopulation / Cattle
+Buffalo / Nearest
Town / Dist. To
Town (km)
Cluster DH / Village1-Dh / 1007 / 500 / BAMHNI / 7
Cluster DH / Village6-S / 846 / 519 / BAMHNI / 12
Cluster HA / Village5-M / 881 / 373 / BAIHAR / 15
Cluster HA / Village3-H / 781 / 588 / BAIHAR / 14
Cluster DU / Village2-Du / 403 / 737 / BAMHNI / 21
Cluster DU / Village4-K / 755 / 670 / BAIHAR / 41
S3: QUANTIFYING FOREST USE
To understand sustainability of forest use, it is important to understand the patterns and processes of forest use as well as distribution of existing forests. At present, many studies quantify forest usage by relying on ephemeral and seasonal signs such as dung abundance or focusing on other signs of human use such as logging, lopping and trails, where one-time use leaves a long-term signature on the forest. The forest still retains signs of use several years after a section of the forest is used intensely for a short period. Depending on such signs may lead to inaccurate understanding of forest use patterns, and it is necessary to accurately quantify forest use to better assess its impact on the forest. Therefore, to quantify forest use, we used trackers to locate movement of cattle and people in forests (Section 2.2). We used these estimates to locate our sample plots within each treatment. Forest area available for each treatment are listed in Table S3.1.
To understand seasonal and spatial differences in forest use by cattle, we used raster package in R package to calculate maximum and mean distance travelled by cattle each day. We also used differences in elevation for consecutive readings in a day to calculate slope traversed by cattle. To calculate average ruggedness traversed per day, we used the absolute values of slopes and totaled these values for a day. To understand factors associated with greater distance traversed by cattle into the forest, we used generalized linear mixed models (lme4 package in R) and constructed models that used maximum and mean distance travelled by cattle into the forest as response variables and used season, distance to town, cattle populations, forest cover available in a 2 km radius, number of villages in 5 kilometer radius, distance to nearest village and ruggedness of terrain traversed per day as predictors. Since many of these predictors were correlated (Table S3.2), we constructed models (Table S3.3) that excluded predictors with correlation coefficient, r, above a certain threshold (r>0.5). Best model was selected using Akaike Information Criterion (AIC) (Burnham and Anderson 2002).
We analyzed factors associated with differences in mean and maximum distance travelled into the forest by people using an analysis that was identical to the one for cattle. However, there were a few differences due to differences in the data. People in villages that had the highest population density and were the closest to major towns reported that they did not visit the forest to collect any produce, which left us with only 4 villages for which we could conduct the analysis. For the analysis, we constructed competing models (Table S3.4) and excluded predictors that had a correlation coefficient, r, greater than 0.5 (Table S3.2). The best model was selected using AIC (Burnham and Anderson 2002).
For reported use of species by people, we calculated proportion of people who voluntarily recalled a certain species or forest product as important. For cattle use of species, we quantified the proportion of sites that a species had been browsed to the number of sites that a species was present. We compared these with the literature on use of species by local residents and cattle (Troup 1983; Brandeis 2007)(Table S3.5).
Table S3.1: Total forest area available for each treatment in each village.
Village / Total Forest Area (km2) / Forest Treatment 2 (km2) / Forest Treatment 1 (km2) / Forest Treatment 0 (km2) / Proportion of Forest in Treatment 2 / Proportion of Forest in Treatment 1 / Proportion of Forest in Treatment 0Vill 1-Dh / 1.75 / 1.16 / 0.59 / 0 / 0.66 / 0.34 / 0
Vill 6-S / 4.534 / 1.49 / 0.15 / 2.89 / 0.33 / 0.03 / 0.64
Vill 5-M / 2.4 / 0.44 / 0.38 / 1.59 / 0.18 / 0.16 / 0.66
Vill 3-H / 4.42 / 0.92 / 1.24 / 2.26 / 0.21 / 0.28 / 0.51
Vill 2-Du / 7.94 / 1.75 / 2.55 / 3.64 / 0.22 / 0.32 / 0.46
Vill 4-K / 8.41 / 1.25 / 2.43 / 4.73 / 0.15 / 0.29 / 0.56
Table S3.2: Correlation coefficients for models
Human Population / Livestock Population / Distance to Town / Total Forest Area / Number of Villages (5-km radius)Human Population / 1 / -0.7884 / -0.40009 / -0.78916 / 0.619864
Livestock Population / -0.7884 / 1 / 0.56579 / 0.882542 / -0.88121
Distance to Town / -0.40009 / 0.565793 / 1 / 0.82212 / -0.83634
Total Forest Area / -0.78916 / 0.882542 / 0.82212 / 1 / -0.95464
Number of Villages (5-km radius) / 0.619864 / -0.88121 / -0.83634 / -0.95464 / 1
Table S3.3: AIC models for prediction of mean distance travelled by cattle
SNo / Model / AIC1 / log(Mean Distance Travelled)~ Total Forest Area+ Ruggedness
+(1|as.factor(Season)) / 114.7859
2 / log(Mean Distance Travelled)~ Total Forest Area
+(1|as.factor(Season)) / 112.7859
3 / log(Mean Distance Travelled)~ Number of villages
+(1|as.factor(Season)) / 113.1850
4 / log(Mean Distance Travelled)~ Livestock Population
+(1|as.factor(Season)) / 114.3483
5 / log(Mean Distance Travelled)~ Total Forest Area+ as.factor(Season)) / 106.3154
Table S3.4: AIC models for prediction of mean distance travelled by people
SNo / Model / AIC1 / log(Mean Distance Travelled)~ Total Forest Area+ Ruggedness+ (1|as.factor(Season)) / 91.00691
2 / log(Mean Distance Travelled)~ Total Forest Area+ (1|as.factor(Season)) / 89.57250
3 / log(Mean Distance Travelled)~ Number of villages + (1|as.factor(Season)) / 89.58907
4 / log(Mean Distance Travelled)~ Human Population+ (1|as.factor(Season)) / 88.89647
5 / log(Mean Distance Travelled)~ Human Population+ Distance to Town+ (1|as.factor(Season)) / 89.72356
7 / log(Mean Distance Travelled)~ Human Population+ Distance to Town+ as.factor(Season) / 81.45967
Table S3.5: Column 2 lists the percentage of residents who voluntarily recalled the species when asked about the use of the forest. Columns 3 lists % of species browsed. Columns 4-6 list use of species.
1 / 2 / 3 / 4 / 5 / 6Species / Recalled / Browsed / Use: fruit and flower / Use: leaf and wood
Use: bark and sap
Anogeissus latifolia / None / Leaf: tanning
Wood: charcoal, agricultural implements / None
Buchanania latifolia / 2.15 / Fruit: eaten , trade / Leaf: plates
Wood: light construction / Bark: tanning
Sap: pellucid gum
Casearia graveolens / 37.5 / Fruit: poison / Leaf: browse
Cassia fistula / Fruit: purgative, medicine / Leaf: cattle fodder (in UP)
Wood: agricultural implements / Bark: tanning, dying
Sap: gum
Diospyros melanoxylon / 6.45 / 16.67 / Fruit: eaten
Flower: None / Leaf: commerce / None
Garuga pinnata / Fruit: eaten, pickled / Leaf: fodder / Bark: tanning
Sap: used
Wood: fuel, indoor work
Lagerstroemia parviflora / <1 / None / Leaf: tanning / Bark: tanning
Sap: sweet gum eaten
Wood: agricultural implements, construction
Miliusa tomentosa / Leaf: cattle fodder
22.2 / Wood: sheds, lopping / None
Madhuca indica / 8.06 / Fruit: oil, poison, insecticide, emetic
Flower: alcohol, trade / Leaf: None / None
50 (n<5) / Wood: protected (even though good for railway sleeper) / Sap: gum
Schleichera trijuga / Fruit: oil (South India)
Flower: None / Leaf: lac, cattle-fodder (UP)
Wood: crushers, ploughs, etc. (hard wood) / None
Terminalia alata / <1 / None / Leaf: silkworm, lac pollarded
Wood: fuelwood (excellent), potash, heavy construction / Bark: tanning, chewed with betel leaf
53.13
Terminalia chebula / 3.23 / Fruit: commerce, tanning, dying, medicine
Flower: None / Leaf: galls-ink, dye
Wood: furniture, agricultural implements, house building / Bark: tanning and dying
Ziziphus xylopyrus / Fruit: dye for trade
Flower: None / Leaf: fodder / Bark: tanning
Wood: torches, cart-building / Sap: None
S4: PATTERNS OF FOREST USE
In all villages and both seasons, cattle were not recorded moving further than 2.8 kilometers from their starting point in the village. The average distance travelled by cattle was 652 meters, although cattle travelled longer distances in the winter (Fig S4.1), and travelled further into the forest where there was more forest available (Fig S4.2, Table S3.3). However, available forest (forest cover) was correlated with distance to town, livestock population, number of villages within a 5-km radius and distance to nearest village (Table S3.2), which suggests that these factors could be associated with distance travelled. Either way, it appears that cattle travel further into the forest where there is more forest available, and more forest is available where there are fewer villages, where the villages are far from towns and where livestock populations are also higher (residents probably own more cattle where there is more forest as there are more resources available for the cattle).
Contrary to livestock, local residents visited a larger area of the forest in the summer (Fig S4.1), when they were recorded to go up to 3.85 kilometers from their village to obtain commercially valuable forest produce such as Mahua indica and Diospyros melanoxylon that are available in the summer. Average distance travelled into the forests was 1.06 kilometers, which suggests that residents may need to travel shorter distances to collect other forest produce. Greater distances travelled into forests was also associated with higher population density (perhaps representing greater need) and greater distance to towns (perhaps representing more valuable forest products) (Fig S4.2, Table S3.3).
Of those that considered forests to be useful, most respondents reported fuelwood and timber as the primary use of the forest (Fig S4.3). Where species were named, bamboo, Madhuca indica, teak and Diospyros melanoxylon were recalled most often. Where multiple species were present, some species were browsed more frequently than other species (Fig S4.3), although this under-represents several other species that are used as browse by cattle (Table S3.5).
Discussion: Patterns of Forest Use
Average foraging distance of livestock herds in the study region is similar to previous studies in dry deciduous forests in India (~1.96 km: Banerjee et al 2013). However, unlike previous studies which found greater foraging distance in the summer (Banerjee et al 2013), our study reports greater foraging distance in the post-monsoon season. This may be because cattle are left free to forage in the summer, while they are escorted by a livestock herder in the post-monsoon season to prevent cattle from grazing on the crops in the field.
Although previous studies have documented shifts in foraging areas of cattle (Bassett and Turner 2007), these studies were focused on migratory herders while the cattle owners in our study region were permanently settled in the villages. This suggests that the foraging range of a village is consistent across years despite variation in annual patterns of forest use as the location of the village is constant. For instance, the distance travelled by cattle is dependent on forest area available or distance to nearest village. Therefore, livestock foraging range is limited by geographical location of the neighboring villages, a fact that does vary across years. This footprint generated by the range of movement of livestock may better represent frequency of forest use than sampling of ephemeral signs or those signs where one-time use leaves a long-term signature on the forest.
In establishing spatial and temporal patterns of cattle use, this analysis meets a critical need for empirical studies on variation in livestock mobility (Turner et al 2014) as strategies and mobility in livestock husbandry and its mobility is recognized as an important mechanism to reduce overgrazing and vulnerability to droughts (Turner et al 2014). Over longer time scales, long-term grazing movements has been documented to affect soil nutrient availability (Turner 1998), and cattleuse can damage saplings, alter rate of tree establishment and change the successional stage of the forest. Species with specific traits that may be able to defend themselves due to phenology, resistance and sapling defenses (Seidl et al 2011) will grow at the expense of others. Since we documented the preferential use of plant species by cattle, it would be important to examine how their regeneration and growth are influenced by intensity of cattle use.
People use forests differently from cattle. This analysis documented that people travel longer distances into the forest, and unlike cattle, travel longer distances in the summer when more commercially viable NTFPs are in season. Average distance travelled by people is very different from maximum distance which suggests that people may be able to meet their primary needs (fuelwood and wood for local construction) in local forests, but may need to travel further to meet their other needs. This suggests that we may see biomass loss closer to villages but species-specific losses further from villages. In forests that are managed by humans, only a few species are retained in each forest type (Crook and Clapp 1998), either because disturbances create conditions that cause increases in some species, or because forest users may actively manage forest to increase abundance of useful products by selectively removing other species (Crook and Clapp 1998). Therefore, there is a need for studies that examine whether it is only species that are specifically targeted that are impacted, or whether other species are also impacted.
Overall, we find that forest use patterns vary seasonally, and depend on resource availability (people may travel further when commercially viable NTFPS are in season, cattle may travel further where more forest is available).
Fig S4.1: -coefficients for distance travelled into the forest by (a) cattle, and (b) people
Fig S4.2: Mean seasonal distance travelled into the forest by (a) cattle, and (b) people
Fig S4.3: Species-specific forest use by (a) people and (b) cattle
S5. VEGETATION SAMPLING AND RESULTS
Our sampling plot design can be seen in Fig S5.1. We measured soil compaction (using agraTronix soil compaction meter: agraTronix, Streetsboro, Ohio, USA), pH and nutrients (using pH/soil analyzer analog: SA2000, Ben Meadows Company, Janesville, Wisconsin, USA) at 10 locations in each plot (two randomly selected points in each quadrat where other field measurements such as canopy cover and understory were also taken).
During measurements, grass and debris were removed from sampling location. For measuring soil compaction, soil compaction meter was held perpendicular to the ground and pressed with full force for reading. For soil and pH, a 15 cm hole was dug, and water was poured into the hole and mixed. Analog was placed in hole to measure pH and wiped before measuring soil.
Temperature data was obtained from MODIS 11A1 (reverb.echo.nasa.gov), and precipitation data was obtained from TRMM 3B43 (trmm.gsfc.nasa.gov ) for the year 2012, and average values for the year were calculated. We obtained 30-meter resolution digital elevation model from ASTER (reverb.echo.nasa.gov) and used it to calculate slope. Average temperature, precipitation, elevation and slope for each of our sample plots was extracted and analyzed using ANOVA.
Analysis of Variance (ANOVA) tests revealed that there were no significant differences in pH, soil nutrients, temperature and precipitation across treatments for any site but soil compaction was significantly higher for villages with high population density (Table S5.1).
Fig S5.1: Field sampling design
Table S5.1: ANOVA test results: p-values for significance of difference in soil compaction, pH and nutrients based on village identity, frequency of use and interaction of village and frequency of use. Only soil compaction is significantly different across villages but not across treatments, and does not differ across treatments for a given village.