A rapid questionnaire survey was carried out during
January–February 2006 sampling 176 villages (covering 50%
of the total villages and farmers, representing both periphery and
enclave) located within and around the 19 forest divisions, to
assess the intensity of elephant–human conflict. At each village,
farmers were interviewed for basic information on extent of
cultivated land and location, type of crops cultivated, and elephant
damage to crop and properties for the year 2005 [see Annexure 1].
Human deaths due to elephants and elephant mortality/capture
as a result of conflict were also noted. Secondary data on compensatory
(ex gratia) payments toward crop and property damage,
and human deaths, were collected for the period from April 2000
to March 2006 for various forest divisions from official records.
Differences in number of farmers and villages affected by elephants
as well as the extent of annual crops cultivated between
the eastern and western part of the landscape were tested using
a chi-square test. Further, the relationship between the extent
of elephant–human conflict with various habitat attributes was
explored using multiple regression. In the multiple regression
framework, the dependent variable was the extent of conflict
(percentage of farmers affected by elephants) in each division
while the independent variables were the perimeter of the forest
division, two simple metrices of habitat fragmentation (ratio of
perimeter of non-forest elements to area of elephant habitat;
number of forest patches within each division), moisture levels
(proportion of evergreen forest), elephant density, and type ofcrops cultivated (percentage of farmer cultivating annual crops).
At first the relationship between the dependent variable and
independent variables were tested using scatter plots. Based on
the relationship of independent variables, the variable was entered
either in linear form or non-linear form with quadratic term. When
the relationship was quadratic, both independent variable and its
square term were entered into the multiple regression model. If
the quadratic term turned out to be insignificant, it was dropped.
At the end, only significant independent variables were retained
in the equation.
Results