Estimation of the rangeland cover by coupling artificial neural network (ANN) and geographic information system (GIS) in Baladeh Ranglands

Rangeland is one of the important natural resources in different aspects such as forage production, livestock, promenade and soil and water conservation. Therefore, it is necessary to study rangelands for their sustaible management and conservation. Since field studies of rangelands are time consuming  and costly, it was common to apply models aimed at estimating rangelands vegetation parameters. In this study, ANN was used to estimate rangelands cover percent and GIS was used as a pre-processing and post-processing in modeling respectively in Baladeh rangelands in Mazandaran Province. Multi-layer percepetron (MLP) network and multivariate regression method were used to estimate rangelands cover percent (training stage). In modeling process, sampling and estimation cover percent was also performed  in the 127 sites. Also, the affecting factors in cover percent such as topography, climatic factors, soil and mankind factors were evaluated.  Multivariate regression and stepwise method were used to simulate rangeland cover in SPSS software. Using cover percent as desired parameter and the affecting factors in cover percent as the network inputs,  an optimal network was presented. Then, optimum network was verified (test stage). The study area was divided with the pixels 1×1 km (raster format) in GIS medium. Then, the model input layers were coupled and a raster layer which included the model inputs values and geographic coordinate was generated. The values of pixels (model inputs) were entered into ANN with geographic coordinate. The results showed that ANN has a higher efficiency and accuracy (model test; Rsqr=0.72) than multivariate regression method (model test; Rsqr=0.6)  in rangeland cover modeling. In the next step, cover percent was simulated using the verified optimum network for all study rangelands. Finally, the results of ANN simulation were entered into GIS and cover percent map was generated based on the simulated results of ANN. The results showed that  coupling of ANN and GIS has a high capability (test stage: Rsr= 0.72) in rangelands cover percent modeling.