Home > Geospatial Application Papers > Environment > Forest Management


Abstract


Delineation of northern boundary of Caspian forest using multitemporal satellite images


Ali Darvishsefat
Associate Professor
University of Tehran,
Iran
Email: adarvish@ut.ac.ir


Farokh Pourshakouri
MS in Forestry
University of Tehran
Email: shakoori@nrf.ut.ac.ir

Ali khlilpour
MS in Remote Sensing
FRO
Email: khalil pour@hotmail.com


Because in the most cases separation of forest boundary with monotemporal images is not possible, this study was performed in order to delineate Caspian forests boundaries with multitemporal images in Iran (case study in Chaboksar). In addition to deciduous broad-leaved forests there are afforested coniferous evergreen stands, tea and citrus orchards and farms and orchards with deciduas species in the study area. Spot-HRG and Terra-ASTER images related to 2002/8/14 and 2002/2/28 were selected for this study. Radiometric errors were negligible and geometric correction (orthorectification) was precisely applied by a digital elevation model. Two images were registered with each other. After defining forest and non-forest classes, the suitable training areas were determined and revised. Separation of forest and non-forest classes was carried out by monotemporal classification, multitemporal classification, hierarchical and digital-visual hybrid approaches. In order to determine the accuracy of maps resulted from interpretation and classifications a ground truth (forest boundary) was prepared using surveying with GPS. The length of surveying route was 64.5 km. The results of different classification were compared with ground truth and their accuracy was determined. Kappa coefficient for classification of growing season and leafless season images, multitemporal classification, hierarchical approach and digital-visual hybrid interpretation were %28, %43, %57, %62 and %71 respectively. The most and least accuracy were related to digital-visual interpretation and classification of growing season image respectively. Overall accuracy was %87.30 for best map.Based on results forest boundary can be separated from non-forests by multitemporal images with high and acceptable. Based upon knowledge, experiences and the current study it is strongly suggested that leafless season images accompanied by growing season image, are used in updating process of topographic maps of Caspian forests in such region.