Sensitivity analysis of multi-temporal remote sensing land-use/cover change detection from multiple sensors

Multi-temporal change detection studies were conducted to examine the sensitivity of multiple remote sensing sensors to land cover/land use change. The first study was a comparison of Landsat TM and Landsat MSS imagery from 1995 with a 2003 Landsat TM image for land cover/use change. The objective was to examine the sensitivity of a Landsat MSS/Landsat TM change detection scenario versus a change detection of purely Landsat TM. MSS was chosen for its historical value and TM for its longevity. The results of the post-classification change detection study showed that the change detection of Landsat MSS and Landsat TM performed very similarly to the purely Landsat TM study in both type and amount of land cover change.;The second change detection study was the comparison of the sensitivity of MODIS 250 meter Band 1, 2 and Thermal Bands to that of Landsat TM data at 30 meter spatial resolution. The study dates were from June and October, 2003 and compared the two sensors for their ability to capture seasonal land cover changes. Post-classification change detection was used using Bands 1, 2 and Thermal Bands. NDVI image differencing was also used for comparison on account of its popularity for regional and global vegetation studies. The different change detection studies were compared at the global, subset (approximately 850 km2) and pixel levels. Results showed that MODIS post-classification change detection over states the amount of change compared to that of TM post classification change detection at the global and subset levels. The type of change at both the global and subset level examined through the post-classification analysis showed poor agreement between TM and MODIS. The per-pixel examination revealed that the higher the percentage of changed TM pixels within a MODIS pixel, the more likely that pixel is to be classified as change in both post-classification and threshold change detection methods. This study also demonstrated that MODIS classification accuracy can be improved by incorporating the 1 km thermal bands with Bands 1 and 2.