Biodiversity from space - optimized grazing and biodiversity conservation with satellite monitoring

Semi-natural grasslands are valuable biotopes in the agricultural landscape, provide habitats for numerous rare plant and animal species and are among the most diverse habitats – both in Sweden and elsewhere. Livestock grazing is vital for the maintenance of grassland biodiversity. Agri-environment schemes (AES) compensate farmers for maintaining pastures, and grazing is a prerequisite for farmers to obtain compensation. At the same time, the grazing intensity needs to be optimally managed as too intense or too little grazing can have detrimental ecosystem effects, amongst other on flower resources for pollinating insects. Monitoring grazing is thus important both for assessing compliance with grazing, and to assess whether grazing regimes are beneficial for grassland biodiversity. However, controlling requirements (eligibility conditions) linked to the AES payments is challenging. In the last EU common agricultural policy (CAP) period, 2015-2022, controls of AES linked requirements were executed by visual inspection of about 5% of the pastures in Sweden annually. In the new CAP period that started this year (2023), new regulations demand that payments for pastures should be controlled for all parcels annually. Surveying grazing intensity in semi-natural pastures on a large spatio-temporal scale is a difficult and very costly task, and there is a great need for new, innovative monitoring methods. This project aims to develop a cost-efficient remote sensing method that will detect grazing and can be used to measure and monitor grazing intensity for semi-natural grasslands from local to national scales, with a high spatio-temporal resolution. Our method will enable grassland management to be quantified over (i) a sequence of years (i.e., time series), (ii) within seasons and (iii) spatially over each pasture. The method will combine spectral and synthetic-aperture radar data to estimate biomass and grazing using a deep temporal window convolutional neural network. The combination of methodological approaches and their application to grazed semi-natural grasslands is novel and fills an important gap. Additionally, we will use an innovative method to validate grazing events, by comparing grazing based on GPS tracked grazing animals and grazing detected by remote sensing. We will demonstrate the usability of the method for agri-environmental purposes with two complementary applications. First, we will link estimated spatio-temporal grazing patterns to the abundance and diversity of flower resources for pollinating insects, as a biodiversity indicator. Secondly, we will provide proof that the method can be used for controls of eligible conditions of AES payments for the maintenance of semi-natural pastures. This will allow for automatic nationwide controls, on a yearly basis. In summary, our project will make a major contribution in harnessing remote-sensing data to solve major challenges at the interface of agriculture and biodiversity conservation, and facilitate achieving national and international commitments for an increased biodiversity.