The effect of model structure and data in modeling land conditions in disturbed complex ecosystems

Prediction and uncertainty source analysis of the spatial and temporal disturbance from off-road vehicular traffic in a complex ecosystem

“…Disturbance model quality measured using R-Square was similar for Fort Riley (0.344 to 0.569) and Fort Hood (0.337 to 0.637) (Fang et al, 2007). The range in total variance was less for Fort Hood (0.00 to 0.195 variance units) than Fort Riley (0.00 to 0.119 variance units) despite higher levels of predicted disturbance (Fang et al, 2002).…”

“…Anderson et al (2005) using the same methodology, GIS data and field data, found that incorporating TM imagery improved model performance. Fang et al (2007) subsequently demonstrated that incorporating interaction terms in the model could improve model performance. Fang et al (2007) also demonstrated that the amount and quality of data available for model development significantly affected model predictions.…”

“…Fang et al (2007) subsequently demonstrated that incorporating interaction terms in the model could improve model performance. Fang et al (2007) also demonstrated that the amount and quality of data available for model development significantly affected model predictions. Depending on the type and amount of input data used to develop the model, the resulting models accounted for 39 to 54% of the variation in observed disturbance.…”