Comparison of MERIS and MODIS Time Series Datasets for Land Degradation Analysis in Irrigated Drylands in Central Asia
Tueshaus, Julia1; Dubovyk, Olena2; Khamzina, Asia2; Menz, Gunter3
1Center for Remote Sensing of Land Surfaces, University of Bonn, GERMANY; 2Center for Development Research, University of Bonn, GERMANY; 3Remote Sensing Research Group, University of Bonn, GERMANY

Land degradation is a severe problem in irrigated drylands of Central Asia due to widespread soil salinization caused by decades of intensive irrigated agriculture. As a result, decreased crop yields and subsequent land abandonment negatively affect the socio-ecological systems, as exemplified in the study area in northern Uzbekistan. Monitoring of land degradation through remote sensing is an important prerequisite for implementing the land rehabilitation measures, but there is no consensus on sensors and vegetation indices suitable for the assessment in irrigated agricultural environments.
To this end, time series of vegetation indices calculated from ENVISAT-MERIS and Terra-MODIS datasets were analyzed to assess land degradation in northern Uzbekistan for the period from 2003 to 2011. Mann-Kendall trend analysis was conducted with time series of MERIS- and MODIS-based Normalized Differenced Vegetation Index (NDVI), Soil-Adjusted Vegetation Index (SAVI) and MERIS-based Terrestrial Chlorophyll Index (MTCI). MERIS_FR_2P and MOD09Q1 products with 300 m and 250 m spatial resolution, respectively, were used for this purpose. The two multispectral sensors and three vegetation indices were selected to detect possible variations of the trends and to examine the strengths and weaknesses of each sensor and index. Assumingly, the MERIS data may be well-suited to detect vegetation parameters throughout the time, as it has a similar spatial and temporal resolution compared to MODIS, but with better spectral characteristics. The MTCI, which includes the red edge region, may be another advantage of MERIS. The following research questions were considered: "Which areas experienced land degradation from 2003 to 2011 as analyzed by applying vegetation index time series?" and "What are the differences between MERIS and MODIS based land degradation trends?"
The methodology consisted of several steps: (1) preprocessing of the original images to derive congruent and 8-day composited vegetation indices images, (2) processing and statistical analysis of the corresponding time series datasets and (3) comparison of the resulting trends. Results of the MODIS- and MERIS-based trend analysis confirmed the occurrence of widespread vegetation decline, interpreted as land degradation. The rate of significant negative trends ranged from 5.5% (MERIS MTCI) to 21% (MODIS NDVI) of the total irrigated cropland in the study area. We argue that MODIS NDVI overestimates land degradation due to the large influence of soil reflectance. In contrast, MERIS MTCI may underestimate the trends as they reflect the high sensitivity of the index to chlorophyll and low influence of soil. Generally, all indices detect the same spatial patterns of cropland degradation, e.g., in the south-western and northern parts of the study area. Average vegetation index values of NDVI and SAVI were higher when measured by MERIS than by MODIS, while areas of negative trends were nearly 1.5 times larger than estimated with the MODIS trend analyses. Agreement between MERIS and MODIS was higher with strong Mann-Kendall trends. These contrasting results indicate the remaining need of investigating the differences between MERIS and MODIS data and show the importance of validating the results from multiple vegetation indices in relation to mapping of land degradation trends in irrigated croplands.