Pixel Patchwork: Monitoring Temperate Grassland Restoration Success With Sentinel‐2 Phenology Metrics in Urban Australia

ABSTRACT Questions More than 99% of Australia's native temperate grasslands have been lost since European colonisation, creating a need for restoration to retain their biodiversity. Remote sensing is a powerful tool for monitoring restoration, yet accessible approaches are needed to extract meaningful, restoration‐relevant information for diverse practitioners. Here we ask if Sentinel‐2 –derived time series can monitor the success of grassland restoration by identifying significant phenological differences between non‐native and native grassland vegetation in time and space. Location The study spanned six geographically distinct regions in southern Australia, strategically positioned along a landscape and management gradient extending from unrestored to restored to remnant intact sites, covering a spatial continuum from the coastal mediterranean climate of Adelaide to the semi‐arid landlocked Riverland region, which supports remnant temperate grasslands. Methods We analysed temperate grasslands using Sentinel‐2 EVI2 (Enhanced Vegetation Index) time series within the web‐based Google Earth Engine platform and using R software. The study included three management categories (remnant, restored and control) with vegetation cover assessed qualitatively by experienced practitioners to confirm these classifications. Phenology modelling was used to fit seasonal growth curves and extract two key metrics: the length of the growing season (LOS) and peak of season EVI2 values (POS). Modelled LOS and POS values were compared across management categories using ANOVA (analysis of variance). Results While we did not identify significant differences in the LOS, we did find significantly different POS values between grasslands dominated by non‐native plants and by native plants. Control sites had POS values 44% higher than restored sites, likely due to the greater leaf area of non‐native vegetation. In contrast, restored grasslands had POS values only 8% lower than remnant reference sites, indicating restored grasslands closely resembled remnant grasslands. This suggests that POS is a reliable metric for distinguishing between native and non‐native grasslands. Conclusions The results demonstrate that phenological metrics derived from Sentinel‐2 pixels effectively capture differences in vegetation composition, providing a practical tool for assessing restoration success. Readily available remote sensing tools like this can support diverse restoration practitioners in prioritising conservation efforts and improving restoration strategies, particularly in urban contexts where grasslands are small and fragmented and continue to face significant anthropogenic pressures.