of soil series was too large to permit examination of the
relationship of each individual series with the density of
trees in paddy fields, they were grouped together into soil
series associated with three land form categories: (1) flood
plains and low terraces, (2) middle terraces, and (3) high
terraces and hills. Each type of land form is associated with
certain specific soil series (Table 2). In doing this analysis,
the various soil series occurring in each grid cell were
identified in order to determine the landforms occurring
within the cell. In those cases, where sample grid cells
included two or more different physiographic units, the cell
was assigned to the dominant land form within it. We
found that 66 cells contained only one type of topography
(32.5 %), 85 cells (41.9 %) contained more than one type
of topography but with more than 70 % of the area occupied by the representative topography, and only 52 cells
(25.6 %) contained more than one type of topography with
70 % of the area occupied by the representative topography. The 203 grid cells were categorized into 73 cells of
flood plain and low terraces, 101 cells of middle terraces,
and 15 cells of high terraces and hills. Fourteen grid cells
were excluded from the analysis because they could not be
assigned to any land form category. The statistical significance of differences in tree density among different
landforms was tested with the Kruskal–Wallis test.