Development of technique to characterize TROF

Abstract

Trees Resources Outside Forests (TROF) is an important natural resource and comprises tree formations in agro-forestry systems. However, determining suitable remote sensing technique to represent TROF for semi-urban to rural environment in India is a challenge. A research is undertaken to develop this technique for a study area existing at the outskirts of Rae Bareli, UP. A classification based approach is tried to be evaluated. The study tried to understand importance of resolution and date of data captured for effective classification of TROF. As classification algorithms can significantly alter accuracy of classification, different classification techniques are analyzed as well. Results are compared statistically with ground truth data for accuracy assessment. The images are geocoded using GPS coordinates of control points. The result of accuracy assessment is summarized using Kappa coefficient. Explanatory statistics is used to compare classified data correspond to classification done with different algorithms over images of various resolution and to comment on efficacy of combinations of resolution of satellite image and classification technique.