Landsat remote sensing approaches for monitoring long-term tree cover dynamics in semi-arid woodlands: Comparison of vegetation indices and spectral mixture analysis

Abstract

Tree canopy cover is a major biophysical attribute of dryland ecosystems. Monitoring its long-term changes over large spatial extents is critical for understanding woody vegetation response to climate variability and global change. However, quantifying tree canopy cover with remotely sensed data remains a challenge for dryland ecosystems where vegetation is sparse and trees, shrubs, and grasses often co-exist at fine spatial scales. In this study, we developed a full SMA (spectral mixture analysis) method that regressed photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), and shade components of the SMA with dryland tree cover to monitor tree cover dynamics on a pinyon–juniper woodland landscape in Nevada, USA using Landsat TM data. We assessed 1) how well this method could estimate tree cover in both disturbed (chained and burned) and non-disturbed woodland patches and 2) how sensitive this method was to the confounding effects of climatic variations. The assessment was conducted in comparison with two other more commonly used methods that regressed NDVI or PV with tree cover. Our results showed that although PV performed better than NDVI, both methods overestimated tree canopy cover within recently disturbed woodland patches where the confounding effects of shrubs on greenness index were higher than in non-disturbed patches. The full SMA efficiently quantified variations within post-chaining patches in addition to non-disturbed patches, but overestimated tree cover within burned patches. Of the three methods tested, only full SMA showed promising capability for mitigating the confounding effects of interannual climatic variations on monitoring the woodland recovery process. Our results are generalizable to other semi-arid landscapes comprising a mosaic of small-statured trees intermixed with shrub steppe vegetation.