Methodological sensitivities when identifying dryland degradation: case studies from two Australian regions

Quantifying dryland degradation is crucial for effective land management. Remote sensing data offers continuous observational records of dryland vegetation over recent decades, aiding in the identification of degraded regions. However, the sensitivities introduced by different satellite vegetation products and varying resolution data in detecting dryland degradation remain unclear. Additionally, the extent of spatial variability that might be overlooked when using gridded climate data is also uncertain. This study quantifies these sensitivities and provides recommendations for more robust identification of dryland degradation areas using the Time Series Segmented Residual Trend (TSS-RESTREND) V0.3.2 method. Our findings reveal that the spatial variability of rainfall in gridded rainfall products is low, potentially leading to errors in the precipitation-vegetation relationship. While GIMMS and MODIS NDVI generally align, variations in the timing of peak NDVI can result in differing degradation assessments, particularly those caused by climate. We recommend using MODIS NDVI when long-term time series data is not required. Specifically, the higher spatial resolution of MODIS NDVI identified degradation areas covering 1.16% of Fowlers Gap and 4.26% of Boolcoomatta, whereas coarse resolution data showed no signs of degradation in these regions. This underscores the importance of higher resolution data for assessing dryland degradation. Conversely, higher temporal resolution provided little additional information, suggesting that monthly data is sufficient for identifying degradation. Additionally, different vegetation indices exhibit similar degradation detection. • AGCD and CHIRPS rainfall datasets (0.05°) did not fully capture the fine-scale spatial rainfall variability observed within the study areas. • Differences in peak NDVI timing influenced the quantification of climate-related degradation when applying the TSS-RESTREND method. • Higher-spatial-resolution data revealed degradation signals undetectable with coarser-resolution products, highlighting the importance of scale matching. • Aggregating native 16-day MODIS data to monthly composites did not reduce trend detection, supporting computationally sufficient analyses when fine temporal detail is not critical. • All vegetation indices used in this paper (NDVI, EVI, and PV) show similar patterns of vegetation change.