COS 245-2 - High-resolution image time series to exploit phenological profiles for mapping ecological states in drylands

Abstract: In arid rangelands, knowledge of soil and climate-determined land potential (land types or ecological sites) and current vegetation state is used to inform site-specific management decisions. Remote sensing methods are commonly used to map land types and states, but accurate classification in arid ecosystems is challenging due to sparse vegetation cover, high spatial heterogeneity, and high inter-annual variability in cover and production. This study aims to evaluate whether remote sensing combined with image-derived phenology metrics can distinguish land types and ecological states in arid rangelands. We evaluate two image sources: near-surface PhenoCams and Harmonized Landsat Sentinel-2 (HLS), collected between 2014 and 2022 at 12 locations at the Jornada Experimental Range in New Mexico, USA. At each location, we extracted 3-day Green Chromatic Coordinate (GCC) time-series data from PhenoCam imagery and Enhanced Vegetation Index (EVI) time-series data from HLS imagery. Phenology metrics (start and end of season, day and value of peak EVI) were extracted from the time-series data for each year and location. Ground data were used to identify land type, ecological state, and dominant plant cover. Monsoon season data from nearby rain gauges were used to determine ‘wet’ (average to above-average precipitation) and ‘dry’ (below-average precipitation) years.

Current results show differences in image-derived phenology metrics between grass- and shrub-dominated ecological states and between wet and dry years. On average, shrub-dominated locations had a later day of year (DOY) start of season (SOS) during wet years (PhenoCam: 119 ± 41, HLS: 109 ± 56) compared to dry years (PhenoCam: 82 ± 19, HLS: 87 ± 26). Grass-dominated locations also had a later SOS in wet years (PhenoCam: 155 ± 27, HLS: 138 ± 50) compared to dry years (PhenoCam: 88 ± 32, HLS: 76 ± 14). The peak value of EVI was also significantly higher for grass-dominated locations during wet years (0.47 ± 0.12) than for grass-dominated locations during dry years (0.32 ± 0.04) or shrub-dominated locations during wet (0.32 ± 0.09) or dry (0.30 ± 0.03) years. These findings suggest that the use of phenological metrics with remotely sensed imagery has potential to detect ecological state change in arid rangelands to complement field monitoring data, particularly when used in conjunction with precipitation data to identify wet and dry years.