Time series analysis of satellite greenness indices for assessing vegetation response to community based rangeland management

Includes bibliographical references.Presented at the Building resilience of Mongolian rangelands: a trans-disciplinary research conference held on June 9-10, 2015 in Ulaanbaatar, Mongolia.After the transition of Mongolia's agriculture sector to a market economy in the early 1990's, community-based rangeland management (CBRM) organizations have been established across Mongolia to cooperatively manage rangeland resources. We hypothesized that rangeland ecoregions under CBRM would have greater biomass than ecoregions managed using traditional herder practices. We used time series analysis of AVHRR (8-km resolution, 1982 to 2012) and MODIS Normalized Difference Vegetation Index (NDVI) (250-m, 2000 to 2013) to calculate integrated NDVI (iNDVI) as a proxy for vegetation biomass. To address whether CBRM response is scale related, we created buffers of increasing distance around livestock winter shelter locations in soums where CBRM programs had been initiated and soums without formal programs. Spatial averages of iNDVI were calculated within buffer boundaries for each location, stratified by ecological zone. A repeated measures mixed model with yearly rainfall as a covariate was used to test for differences in iNDVI for CBRM status over time for buffer distances of 1, 2, 5, 10, and 30 for MODIS, and 10 and 30 km for AVHRR. In general, results were similar across buffer distances indicating that average vegetation response was similar for distances greater than 1 km around sampling sites. For MODIS NDVI, sites in the Desert Steppe and Eastern Steppe did not have significantly higher productivity in CBRM managed soums over time, regardless of buffer size. Mountain and Forest Steppe (MFS) locations had higher iNDVI in non-CBRM sites throughout the time series for both NDVI data sets, although these differences were not statistically significant. CBRM sites in the Steppe zone had higher iNDVI throughout the time series for both MODIS and AVHRR. Given that these differences occur throughout the AVHRR time series, they do not appear to be the result of CBRM activities. Our findings indicate that differences in vegetation response as a result of CBRM activities were not detected during the time series using productivity proxies from satellite imagery. In addition, the MODIS time series may be too short for detecting CBRM differences since it does not include data prior to when most CBRM programs were implemented