Large-scale remote sensing analysis reveals an increasing coupling of grassland vitality to atmospheric water demand.

Grasslands provide important ecosystem services to society, including biodiversity, water security, erosion control, and forage production. Grasslands are also vulnerable to droughts, rendering their future vitality under climate change uncertain. Yet, the grassland response to drought is not well understood, especially for heterogeneous Central European grasslands. We here fill this gap by quantifying the spatiotemporal sensitivity of grasslands to drought using a novel remote sensing dataset from Landsat/Sentinel-2 paired with climate re-analysis data. Specifically, we quantified annual grassland vitality at fine spatial scale and national extent (Germany) from 1985 to 2021. We analyzed grassland sensitivity to drought by testing for statistically robust links between grassland vitality and common drought indices. We furthermore explored the spatiotemporal variability of drought sensitivity for 12 grassland habitat types given their different biotic and abiotic features. Grassland vitality maps revealed a large-scale reduction of grassland vitality during past droughts. The unprecedented drought of 2018-2019 stood out as the largest multi-year vitality decline since the mid-1980s. Grassland vitality was consistently coupled to drought (R2 = .09-.22) with Vapor Pressure Deficit explaining vitality best. This suggests that high atmospheric water demand, as observed during recent compounding drought and heatwave events, has major impacts on grassland vitality in Central Europe. We found a significant increase in drought sensitivity over time with highest sensitivities detected in periods of extremely high atmospheric water demand, suggesting that drought impacts on grasslands are becoming more severe with ongoing climate change. The spatial variability of grassland drought sensitivity was linked to different habitat types, with declining sensitivity from dry and mesic to wet habitats. Our study provides the first large-scale, long-term, and spatially explicit evidence of increasing drought sensitivities of Central European grasslands. With rising compound droughts and heatwaves under climate change, large-scale grassland vitality loss, as in 2018-2019, will thus become more likely in the future.

National-scale mapping of building height using Sentinel-1 and Sentinel-2 time series.

Urban areas and their vertical characteristics have a manifold and far-reaching impact on our environment. However, openly accessible information at high spatial resolution is still missing at large for complete countries or regions. In this study, we combined Sentinel-1A/B and Sentinel-2A/B time series to map building heights for entire Germany on a 10 m grid resolving built-up structures in rural and urban contexts. We utilized information from the spectral/polarization, temporal and spatial dimensions by combining band-wise temporal aggregation statistics with morphological metrics. We trained machine learning regression models with highly accurate building height information from several 3D building models. The novelty of this method lies in the very fine resolution yet large spatial extent to which it can be applied, as well as in the use of building shadows in optical imagery. Results indicate that both radar-only and optical-only models can be used to predict building height, but the synergistic combination of both data sources leads to superior results. When testing the model against independent datasets, very consistent performance was achieved (frequency-weighted RMSE of 2.9 m to 3.5 m), which suggests that the prediction of the most frequently occurring buildings was robust. The average building height varies considerably across Germany with lower buildings in Eastern and South-Eastern Germany and taller ones along the highly urbanized areas in Western Germany. We emphasize the straightforward applicability of this approach on the national scale. It mostly relies on freely available satellite imagery and open source software, which potentially permit frequent update cycles and cost-effective mapping that may be relevant for a plethora of different applications, e.g. physical analysis of structural features or mapping society's resource usage.

Mapping urban-rural gradients of settlements and vegetation at national scale using Sentinel-2 spectral-temporal metrics and regression-based unmixing with synthetic training data.

The increasing impact of humans on land and ongoing global population growth requires an improved understanding of land cover (LC) and land use (LU) processes related to settlements. The heterogeneity of built-up areas and infrastructures as well as the importance of not only mapping, but also characterizing anthropogenic structures suggests using a sub-pixel mapping approach for analysing related LC from space. We implement a regression-based unmixing approach for mapping built-up surfaces and infrastructure, woody vegetation and non-woody vegetation for all of Germany and Austria at 10 m resolution to demonstrate the potential of sub-pixel mapping. We map LC fractions for one point in time, using all available Sentinel-2 data from 2017 and 2018 (<70% cloud cover). We combine the concept of synthetically mixed training data with statistical aggregations from spectral-temporal metrics (STM) derived from Sentinel-2 reflectance time series. We specifically examine how STM can be used for creating synthetically mixed training data. STM are known to facilitate large area mapping by being largely independent of image acquisition dates and inherently incorporate phenological information. Vegetation is an important part of settlements and time series information supports its mapping. Synthetically mixed training data facilitates a streamlined training by using pure reference spectra to generate artificial mixtures as input to regression modelling of LC fractions in mixed pixels. We here show how combining both offers great potential for wall-to-wall LC fraction mapping. We further investigate the positive effect of STM on map results by comparing the performance of different subsets of STM combinations. Our results indicate that many STM combinations containing spectral variability and vegetation indices provide suitable input to creating synthetic training data for regression-based fraction mapping. Results for built-up surfaces and infrastructure (MAE 0.13/RMSE 0.18 at 20 m resolution), woody vegetation (0.18, 0.22) and non-woody vegetation (0.14, 0.19) are highly consistent across Germany and Austria. Only a few surface types were not accurately predicted in our nation-wide mapping. Further research is required to optimize mapping of temporally invariant bare soil and rock surfaces that show spectral similarity to built-up surfaces and infrastructure. The proposed methodology combines benefits of both regression-based modelling with synthetically mixed training data and STM, and thus facilitates mapping of LC fractions on a national scale and at high resolution. Such information will allow to better characterize settlements and identifying processes such as densification that are best represented by continuous LC mapping.