Speaker
Description
Introduction
Monitoring crops, including their phenological stages, plays a crucial role in agronomic applications, providing essential insights into temporal dynamics and crop conditions throughout the growing season. These insights are particularly valuable for informing crop models used in agricultural management and resource allocation, especially in the face of climate change. Remote sensing (RS) data, acquired over large areas, are fundamental in driving process-based simulation models. These models, in turn, generate simulations that are invaluable for informing strategic decision-making in agriculture. While the majority of agricultural studies rely on either optical or radar satellite sensors (Blickensdorfer et al. 2022; Meroni et al. 2021), there is still a notable gap in the application of fused data products to benefit from both systems (Cheng et al. 2023; Chen and Zhang 2023).
In our study, we evaluate the potential of dense time series of multimodal fused data obtained from Sentinel-1 (S1), Sentinel-2 (S2), and PlanetScope (PS) satellites to assess Land Surface Phenology (LSP) metrics at 10m resolution for two crops (maize, winter wheat) in Germany from 2017–2023. The specific objectives are (i) to establish and validate patterns that indicate a phenological stage of Start-, Peak- and End-of-Season (respectively SoS, PoS and EoS) within the time series of both optical and backscatter data, (ii) to prove the added value of particular gap-filling and fusion strategy, and (iii) to obtain regional spatially explicit information of crop phenology.
Materials and methods
The methodology (Fig. 1) encompasses radar (S1) and optical (S2, PS) data pre- and processing (filtering, cloud screening, calculation of several vegetation indices e.g., NDVI, RVI), time-series processing (interpolation, gap filling, smoothing and fitting, data fusion), the LSP metrics estimation (threshold-, slope and derivative-based methods), and accuracy assessment (based on phenology from German Weather Service, Copernicus HR-VPP, farmers reports). Analysis and comparison is performed over several federal states of Germany (Brandenburg, Lower Saxony, Bavaria).
Results
Generally, the results indicate a strong temporal correlation between dense RS time series obtained from multimodal satellite sources and significant level of agreement between the LSP metrics at a field level for the analyzed crops. The fusion of optical S2 and PS data improves the characterization of phenology variations at a sub-field level across multiple sites and years, and reduce uncertainties related to mixed pixels. Nevertheless, the obvious impact of clouds on optical systems, insufficient data availability and uncertainty related to image co-registration, persist the critical points in accurate LSP retrieval. The S1-derived LSP metrics, such as SoS, PoS and EoS, show high accuracy and reliable results within the expected range compared to other sources like S2 derived metrics, DWD, HR-VPP and farmers reports. Using backscatter and cross-polarization ratio, the temporal patterns of crop phenological development were clearly identified and LPS metrics accurately derived.
Discussion
The utilization of multimodal RS time series from combined optical and radar systems significantly broadens the scope of crop phenological monitoring beyond conventional field observations. The fusion of datasets through advanced algorithms and machine learning techniques promises comprehensive insights into crop development stages with denser time series and higher spatiotemporal resolution. This would facilitate timely intervention for crop management, and more effective responses to climate variability as well as can inform process-based crop models, e.g. for early-season prediction of yields.
References
Blickensdorfer, L., M. Schwieder, D. Pflugmacher, C. Nendel, S. Erasmi, and P. Hostert. 2022. 'Mapping of crop types and crop sequences with combined time series of Sentinel-1, Sentinel-2 and Landsat 8 data for Germany', Remote Sensing of Environment, 269.
Chen, J., and Z. Zhang. 2023. 'An improved fusion of Landsat-7/8, Sentinel-2, and Sentinel-1 data for monitoring alfalfa: Implications for crop remote sensing', International Journal of Applied Earth Observation and Geoinformation, 124.
Cheng, G., H. Ding, J. Yang, and Y. S. Cheng. 2023. 'Crop type classification with combined spectral, texture, and radar features of time-series Sentinel-1 and Sentinel-2 data', International Journal of Remote Sensing, 44: 1215-37.
Meroni, M., R. d'Andrimont, A. Vrieling, D. Fasbender, G. Lemoine, F. Rembold, L. Seguini, and A. Verhegghen. 2021. 'Comparing land surface phenology of major European crops as derived from SAR and multispectral data of Sentinel-1 and-2', Remote Sensing of Environment, 253.
| Keywords | Time-series; Land Surface Phenology (LSP); Fusion; Optical; Radar |
|---|
