Effect of spatial image support in detecting long-term vegetation change from satellite time-series

Authors: Maynard, Jonathan; Karl, Jason; Browning, Dawn

Submitted to: Landscape Ecology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/8/2016
Publication Date: 5/10/2016
Publication URL: http://handle.nal.usda.gov/10113/5555872
Citation: Maynard, J.J., Karl, J.W., Browning, D.M. 2016. Effect of spatial image support in detecting long-term vegetation change from satellite time-series. Landscape Ecology. 31:2045-2062. doi:10.1007/s10980-016-0381-y.

Interpretive Summary: Arid rangelands within the southwestern United States have been severely degraded over the past century due to intensive land-use practices and the increasing effects of drought and climate change. Consequently, there is a critical need to develop monitoring approaches that can detect significant changes in vegetation health and distribution across vast spatial extents. Multi-temporal remote sensing techniques are ideally suited to address these challenges; however, considerable uncertainty exists regarding the effects of changing image resolution on the ability to detect ecologically meaningful change from satellite time-series. The main objective of this study was to explicitly test the effects of changing image resolution on the ability of NDVI time-series to detect observed long-term changes in plant biomass. Resutls from this study show that both moderate (Landsat) and coarse (MODIS) scale imagery were effective in modeling temporal dynamics in vegetation structure and composition, although MODIS was more strongly correlated to biomass due to its cleaner (i.e., fewer artifacts/data gaps) 16-day temporal signal. While the results presented in this study are likely specific to arid shrub-grassland ecosystems, the approach presented here is generally applicableand future analysis is needed in other ecosystems to assess how scaling relationships will change under different vegetation communities.

Technical Abstract: Context Arid rangelands have been severely degraded over the past century. Multi-temporal remote sensing techniques are ideally suited to detect significant changes in ecosystem state; however, considerable uncertainty exists regarding the effects of changing image resolution on their ability to detect ecologically meaningful change from satellite time-series. Objectives (1) Assess the effects of image resolution in detecting landscape spatial heterogeneity. (2) Compare and evaluate the efficacy of coarse (MODIS) and moderate (Landsat) resolution satellite time-series for detecting ecosystem change. Methods Using long-term vegetation monitoring data from grassland and scrubland sites in southern New Mexico, USA, we evaluated the effects of changing image support using MODIS (250-m) and Landsat (30-m) time-series in modeling and detecting significant changes in vegetative biomass using time-series decomposition techniques. Results Within our study ecosystem, landscape-scale (>20-m) spatial heterogeneity was low, resulting in a similar ability to detect vegetation changes across both satellite sensors and levels of spatial image support. While both Landsat and MODIS imagery were effective in modeling temporal dynamics in vegetation structure and composition, MODIS was more strongly correlated to biomass due to its cleaner (i.e., fewer artifacts/data gaps) 16-day temporal signal. Conclusions The optimization of spatial/temporal scale is critical in ensuring adequate detection of change. While the results presented in this study are likely specific to arid shrub-grassland ecosystems, the approach presented here is generally applicable. Future analysis is needed in other ecosystems to assess how scaling relationships will change under different vegetation communities that range in their degree of landscape heterogeneity.