Estimates the longitudinal (or profile), normal and planar curvatures for each site through a finite difference schema.
The curvatures represent the deviations of the gradient vector for unit length (in radiants) along particular curves plotted on the surface under consideration. In particular, the presence of non-zero curvatures has relevant effects on the representation of the properties of the surfaces discretized. For example, if the surface has a negative normal curvature, then the gradients have diverging directions at the extremes of the pixel, P, and the contributing area in P is spread over several adjacent pixels: in this case topography is called locally divergent. Vice versa, the surface is locally converging (negative curvature) and the contributing area in P tends to be spread over a limited set of adjacent pixels and almost certainly on a single pixel.
Roughly speaking, the convex zones are hillslope zones, the concave zones are valleys. As it is known, the latter contain the channel network. Then, the curvature tends to discriminate the points across the basin with greater humidity content (the concave ones). This fact has relevant consequences on the overall hydrologic behavior of basins and, in particular, on the production of runoff and on the evapotranspiration distribution.
The longitudinal curvature represent the deviation of the gradient along
the the flow (it is negative if the gradient increase), the normal
and planar curvatures are locally proportional and measure the
convergence/divergence of the flow (the curvature is positive for
convergent flow).