HR: 0830h
AN: B41C-0889    [PDF]
TI: Analyzing Land Cover Change in Kazakhstan: Land Surface Phenology, Climatic Variation, and Sensor Artifacts
AU: * de Beurs, K M
EM: kdebeurs@calmit.unl.edu
AF: University of Nebraska-Lincoln, School of Natural Resources 113 Nebraska Hall, Lincoln, NE 68588-0517 United States
AU: Henebry, G M
EM: ghenebry@calmit.unl.edu
AF: University of Nebraska-Lincoln, School of Natural Resources 113 Nebraska Hall, Lincoln, NE 68588-0517 United States
AB: The collapse of the economic and political institutions of the Soviet Union in the early 1990s led to widespread agricultural de-intensification, land abandonment, loss of livestock, and decreased grazing pressure. In semi-arid to arid regions dominated by dryland agriculture and grazing, the quantification of land cover change must distinguish anthropogenic forcings from interannual climatic variation and the peculiarities associated with specific sensor systems. Were the land cover changes that occurred in Kazakhstan following independence in 1991 of sufficient magnitude to alter the land surface phenology at resolutions relevant to climate models? To explore this question it is necessary first to partition the sources of variation in the image archive. We used the standard Pathfinder AVHRR Land (PAL) dataset, which consists of global 10 d maximum NDVI composites from 7/1981 to 9/2001 at 8 km resolution. To what extent are the PAL data affected by sensor artifacts that may mask other kinds of change? We evaluated 19 subsets of 1600 sq km, one for each ecoregion of Kazakhstan as delineated by the World Wildlife Fund. To minimize residual cloud contamination in the PAL data, a modified version of the best index slope extraction algorithm was applied. The method filters distortions without altering the seasonal NDVI pattern. We pursued two complementary aspects of change analysis: (1) detection of trends within each sensor's tenure and (2) detection of trends and discontinuities across the entire observational period. Seasonal polynomial models of NDVI phenology were developed to relate accumulated growing degree-day with NDVI. To test for trends within periods, both the residuals and the filtered data were submitted to seasonal Mann-Kendall tests that were modified to correct for serial correlation. To identify discontinuities, the entire series was tested using the standard normal homogeneity test (SNHT) without trend. The Kruskal-Wallis test with Bonferroni correction was applied to compare both the original NDVI data and the residual data among the periods. All ecoregions displayed distinct land surface phenologies. There were no significant trends in the NDVI data within a single sensor period. SNHT detected specific discontinuities that were attributable to sensor changes in certain ecoregions. These results imply that PAL NDVI data are a credible source for land cover change research and that straightforward nonparametric statistical methods are able to distinguish between climatic variation, sensor artifacts, and institutional changes.
UR: http://www.calmit.unl.edu/kz/
DE: 1640 Remote sensing
DE: 1851 Plant ecology
DE: 3322 Land/atmosphere interactions
DE: 9320 Asia
SC: Biogeosciences [B]
MN: 2003 Fall Meeting