Feasibility Study on the Use of Medium Resolution Satellite Data for the Detection of Forest Cover Change Caused by Clear Cutting of Coniferous Forests in the Northwest of Russia

Abstract

This feasibility study examines possibilities of identifying and mapping of clear cuts in the boreal coniferous forests of north western Russia based on satellite imagery of medium spatial resolution (MRSD). Different image products obtained from the MODIS and MERIS sensors were visually examined in order to assess their suitability for clear cut detection. MODIS images of 250 m spatial resolution from the years 2001 and 2002 were ultimately selected as the main data source for change analysis. The study compares different detection and mapping approaches, including (i) visual interpretation using principal components, (ii) unsupervised digital classification and (iii) a combination of image differencing and textural analysis, defining statistical thresholds for the identification of clear cuts. The results show that the detection of clear cut locations is feasible by each approach, provided that the size of the clear cut is larger than about 15 ha. For mapping and area estimation the digital approaches are considered more efficient. The best mapping result was achieved based on the combination of image differencing and textural analysis. However, the mapping accuracy is affected by omission and commission errors, mainly caused by the limited spatial resolution of the imagery and by slight geometrical location shifts between different acquisitions. Also atmospheric and seasonal effects can influence the reliability of change indications. Based on a set of reference clear cut areas, delineated from Landsat TM imagery, the best mapping accuracy achieved was at about 65 %.