Change Detection Based on Fractional Tree Cover Derived from Modis Data

Knowledge about the land cover of large areas is important for monitoring and modeling of ecological and environmental processes and for land cover change detection. Considerable efforts have recently resulted in the development of global continuous fields for different land cover types at large spatial scales based on NOAA-AVHRR, SPOT-VEGETATION, and TERRA-MODIS data. Researchers have applied a range of techniques to depict the sub-pixel fraction of land cover types from remotely sensed data. As a result, such products have a high potential to accurately monitor land cover and its change over time. In this study, we assess the potential of MODIS-derived fractional tree cover for change detection. Based on the uncertainty-model of our method to map land cover continuous fields we describe a methodology to quantify the statistical certainty of detected changes, thus adding statistical significance levels. The presented study is carried out in Switzerland a central part of the European Alps with a complex topography and a highly fragmented landscape. Based on MODIS data (MOD09) at a spatial resolution of 500m we derived fractional tree cover using a generalized linear model (GLM). The advantage of this model is quantitative availability of the model error for all predicted observations. This information can be used to calculate the interval of tolerance for each observation and for each year individually. Based on these intervals the probability of an occurred change of the fractional cover between two years can be calculated. For purpose of evaluation we first modelled the fractional tree-cover for the years 2001 and 2003 on MODIS 500m data using Landsatdata for training resampled to the spatial resolution of 500m. In a second step we calculated the fractional tree cover change for Landsat and MODIS and compared it to each other. Finally, we mapped the fractional tree cover of 1999 based on a Landsat-image of summer 1999. The calculated change between 1999 and the Landsat respectively MODIS-based fractional tree cover were compared to severe storm damages occurred in December 1999 over large parts of Europe. In Switzerland these damages were mapped by visual air photo interpretation. The correlation between Landsat based and MODIS based fractional tree cover change was 0.52.The correlation between air-photo based storm damages and the calculated fractional change was 0.43 for the Landsat respectively 0.45 for the MODIS fractional tree cover. The automatic classification of significantly changed pixels based on the probability revealed that on the 95% confidence interval only observations with a change of approx. 30% were detected. A decrease of the confidence level to 75% revealed only a few misclassifications and showed that the approach is very promising. Generally we conclude that our approach based on MODIS MOD09 data is appropriate for 1) mapping fractional tree cover and 2) monitoring its temporal change.