Simulating spatial pattern and dynamics of military training impacts for allocation of land repair using images

The land management of US Army installations requires information on land conditions and their history for planning future military training activities and allocation of land repair. There is thus a strong need for methodology development to estimate the land conditions and cumulative military training impacts for the purpose of repair and restoration. In this study, we simulated at Fort Riley, USA, spatial patterns and temporal dynamics of military training impacts on land conditions quantified as percent ground cover using an image-aided spatial conditional co-simulation algorithm. Moreover, we estimated the historical percent ground cover as a measure of the cumulative impacts, and then calculated the allocation of land repair and restoration based on both current and historical land conditions. In addition, we developed a loss function method for allocation of land repair and restoration. The results showed: (1) this co-simulation algorithm reproduced spatial and temporal variability of percent ground cover and provided estimates of uncertainties with the correlation coefficients and root mean square errors between the simulated and observed values varying from 0.63 to 0.88 and from 23% to 78%, respectively; (2) with and without the cumulative impacts, the obtained spatial patterns of the land repair categories were similar, but their land areas differed by 5% to 40% in some years; (3) the combination of the loss function with the co-simulation made it possible to estimate and computationally propagate the uncertainties of land conditions into the uncertainties of expected cost loss for misallocation of land repair and restoration; and (4) the loss function, physical threshold, and probability threshold methods led to similar spatial patterns and temporal dynamics of the land repair categories, however, the loss function increased the land area by 5% to 30% for intense and moderate repairs and decreased the area by 5% to 30% for no repairs and light repairs for most of the years. This approach provided the potential to improve and automate the existing land rehabilitation and maintenance (LRAM) system used for the land management of the U.S. Army installations, and it can be applied to the management of other civil lands and environments. In conclusion, this study overcame the important gaps that exist in the methodological development and application for simulating land conditions and cumulative impacts due to human activities, and also in the methods for the allocation of land for repair and restoration.