Military training land management, including locating land disturbances, quantifying disturbance severity, and monitoring disturbance mitigation efforts are resource intensive and costly for military land managers. This project investigated the cost, accuracy, cost-efficiency, value of information, and scale appropriateness for several remotely sensed data sources. 

Cost-efficiency was evaluated as the difference between predicted and field-observed percent vegetative cover (PVC). More specifically, remote sensing methods were demonstrated to map and detect change in PVC using pre-military training and post-military training unmanned aerial system (UAS) and satellite imagery at two military installations in different ecosystems (Ft. Riley, KS and Ft. McCoy, WI) and at five different spatial resolutions: 1) Native resolution UAS imagery (~2 cm and ~8 cm) for assessing PVC at large/site-specific scales; 2) Resampled resolution (5 m) for assessing PVC at medium/landscape scales; 3) Resampled resolution (10 m) also for assessing PVC at medium/landscape scales; 4) Native resolution (Sentinel-2) (10 m) satellite imagery for assessing PVC at medium/landscape scales; and 5) Native resolution (MODIS) (250 m) satellite imagery for assessing PVC at small/regional scales. Two camera systems were used for the analysis, the RedEdge-M sensor and the Sony a7RIII camera.

In all cases, the remotely sensed data were correlated with field observations of PVC collected at 5 m by 5 m and 10 m by 10 m field plots. Based on our analysis of the native resolution UAS imagery, the RedEdge-M sensor was more cost efficient than the Sony a7RIII camera for most uses.  Resampling UAS imagery 10 m was more cost efficient that resampling to 5 m. For satellite images, the Sentinel-2 imagery was less cost efficient than MODIS imagery.

The use of native resolution RedEdge-M multispectral UAS imagery (~ 8 cm) resulted in the highest cost-efficiency for mapping PVC and PVC change. The capability to detect individual tracks after training where PVC had decreased significantly or where there were new patches of bare ground was demonstrated for both UAS image types at their native resolutions. The Sony a7RIII image analysis resulted in much lower cost-efficiency values due to much larger file sizes which increased processing time substantially, as well as, less accurate estimates of PVC due to the limitations of color cameras when compared to multispectral sensors. Although the cost-efficiency of using MODIS data was only slightly lower than the cost-efficiency of the native resolution RedEdge-M imagery, the value of the information that can be extracted from the RedEdge-M imagery was much higher than the MODIS imagery. MODIS analysis is appropriate for identification of disturbance but not appropriate for site specific analysis. Overall, this study demonstrated the value of using UAS images to map PVC and its dynamics due to military training activities and provided useful implications for developing a near real-time system of monitoring such changes.