Biodiversity is under pressure worldwide, particularly in the Netherlands under sustained anthropogenic stressors. Governments are responding through policy measures that require impact quantification, creating the need for scalable monitoring methods. Geospatial Foundation Models (GFMs) can learn patterns across space, time, and context, potentially enhancing biodiversity monitoring compared to conventional remote sensing methods. Yet the ecological community has not widely embraced GFMs, likely due to their rigidity to changes in modalities without re-pretraining, misalignment between pretraining modalities and downstream tasks, and the unknown effect of data-gaps during discriminatory phenological periods. This study adapts GFMs for biodiversity monitoring via two research questions, addressing the challenges in reverse order: how data-gaps in GFM embeddings impact vegetation species mapping capacity, and how modular GFMs trained on multimodal data affect this capacity. SpheerFM, a Sentinel-2 pixel-in-time GFM pretrained on Dutch nature reserves, is evaluated on the Veluwe, with height data added as a complementary modality. Embedding sensitivity is assessed through period-based ablation scenarios, and height data is incorporated using a modular combination methodology to test multimodal sensitivity against Sentinel-2 perturbations. Results show spatiotemporal data-gaps are prevalent over the Veluwe, also during discriminatory vegetative periods, with likelihood varying by year and period due to cloud cover and tile availability. Unimodal embeddings show sensitivity to data-gaps; masking discriminatory periods reduces discriminability, while masking assumed redundant periods generally improves it. No existing method satisfied all modular and cross-modal requirements. The resulting multimodal embeddings generally prevented worsened classification following data-gaps when the added modality’s features are discriminatory and complementary. ... Read more

https://github.com/lotvanneerbos/HeLEOsat-S2 ... Read more

Measure Correlate Predict (MCP) is a method used to characterize the future wind resource at potential new wind farm locations. It finds a relationship between the wind data obtained at the target site and a nearby reference site over a short time period. This relationship is applied to long-term historical data from the reference site in order to predict the long-term wind resource at the potential wind farm location. Over the last few decades, new MCP techniques have been proposed, and modelled meteorological data in the form of reanalysis products have emerged. However, little research has been done regarding how the application of modelled data and different regression techniques in MCP affect the achievable wind resource prediction accuracy.

This project set out to investigate the accuracy of the MCP procedure under different configurations. Three comparative studies have been done in this project. Firstly, the attainable accuracy with either nearby MET-station data or ERA5 reanalysis data as a long-term reference source is assessed. Secondly, the use of different regression types for forming the relationship between target and reference data in the MCP procedure is assessed. Lastly, the accuracy achieved with standard MCP is compared to that achieved with a new wind resource estimation method, the method of analogs. The accuracy of the different configurations is assessed through the ability to accurately predict a period of wind speed values measured at 35 sites located in different terrain types. The predictions are evaluated using metrics such as the coefficient of determination, the root mean square error, the mean absolute error and the mean bias error.

This study found that ERA5 reanalysis data can serve as a reliable alternative to observed MET-station data. Generally, using ERA5 reanalysis data as a reference source always led to more accurate predictions then a MET-station reference source if the Pearson correlation between target and MET-station is lower than 0.8, and for offshore targets. If the Pearson correlation between target and MET-station reference is higher than 0.9, the achieved accuracy with either the MET-station or ERA5 data as a long-term reference is similar and depends on specific site conditions. In terms of regression methods it was found that the Matrix method, using the target site sectors for determining the regression parameters, generally outperforms other regression methods in terms of accuracy when determining the mean wind speed. Lastly, the method of analogs, a recently developed wind speed estimation method, yielded a similar prediction accuracy to standard MCP.

It should be noted that the different regression methods in employed in MCP all exhibited very similar prediction outcomes, with an average absolute difference in the predicted mean wind speed between the best and worst performing regression methods of only 0.046 m/s. Furthermore, the performance of the method of analogs in terms of accuracy improves with a longer concurrent period during which the relationships are formed. This project employed relatively short concurrent periods for certain targets, which may have contributed to sub-optimal performances of the method of analogs. ... Read more