Mapping wind-blown mass transport by modeling variability in space and time

Field observations of wind-blown particle transport are often characterized by a considerable spatial variation, which makes quantitative modeling of wind erosion difficult. This study examines how the horizontal distribution, or pattern, of mass transport can be determined from a limited number of point measurements. Twenty-one sediment catchers were installed in an experimental plot in the Sahelian zone of Niger, on a sandy, siliceous, isohyperthermic Psammentic Paleustalf. Mass transport values during four storms ranged from 24.0 to 213.6 kg m(-1), 7.2 to 26.0 kg m(-1), 9.6 to 68.9 kg m(-1), and 68.9 to 282.7 kg m(-1). Geostatistical theory was applied to produce storm based maps by modeling the spatial correlation structure with the variogram. To estimate the variogram from 21 observations, the four storms were treated as independent temporal replicates. Two geostatistical mapping techniques were applied. Kriging (a spatial interpolation technique) produced maps of mass transport providing the best possible estimates of net soil losses from the plot, equal to 12.5, 2.0, 4.6, and 26.8 Mg ha(-1), respectively. To overcome smoothing, possible realizations of actual mass transport were created by stochastic simulations with simulated annealing. The simulated maps reproduced the statistical properties of the observations and allowed a distinction between erosion and deposition areas within the experimental plot.