Detection of grassland mowing events with optical satellite time series data

Abstract

Grassland management - in particular the use intensity - determines its ecosystem services,like fodder production, carbon storage, freshwater generation and biodiversity. However, large-scale and spatially explicit information on grassland use intensity is often unknown. Here, an annual time series of high-resolution optical satellite data (Sentinel-2) for the year 2019 was used to detect mowing events in southern Germany. The pre-processed satellite time series was interpolated, smoothed and filtered and the daily Enhanced Vegetation Index (EVI) was calculated. Afterwards, mowing events were detected by applying an algorithm, which locates strong minima within the EVI time series per pixel. The results were validated by comparing them to mowing events extracted from daily pictures of grassland parcels on field scale. The number and dates of mowing events showed good results as 79% of the observed harvests were successfully detected. Mowing events were missed when the dense time series was disrupted by cloud conditions as the EVI response after mowing events usually lasted not longer than 14 days.Falsely detected mowing events were mostly related to grazing activities.