In the field of agriculture, understanding crop phenology is crucial for decision makers to effectively plan and adapt agricultural management strategies. With the increasing availability of big Earth observation data, efforts have been made to accurately detect crop phenology using Remote Sensing (RS) and high resolution weather data. However, many existing studies have focused on large scale predictions or developed methods that do not fully support crop modeler communities in leveraging Sentinel-1 and Sentinel-2 data fused with high resolution climate data. To address this gap, a Machine Learning (ML) LightGBM model was trained to predict 13 phenological stages for eight major crops across Germany at a 20-meter scale. Observed phenologies from the German national phenology network (German Meteorological Service; DWD) between 2017 and 2021 were used for training and validation. A thorough feature selection analysis was conducted to identify the best combination of RS and climate data for detecting phenological stages. The results showed promising outcomes at a national scale, with predicted phenology achieving a reasonable precision of R2 > 0.43 and a low Mean Absolute Error of 6 days across all phenological stages and crops. The spatio-temporal analysis demonstrated the model's transferability across different spatial and temporal contexts in Germany. By combining radar sensors with climate data, the model showed significant improvements that can benefit various practical applications. Furthermore, insights from previous studies highlighted the importance of data volume in influencing ML model accuracy, as well as the impact of specific feature sets on predicting different phenological stages of crops. For instance, radar backscatter coefficients combined with optical raw band data were found to be valuable for identifying early stages in winter wheat crops. Overall, this novel fusion approach of Sentinel-1 and Sentinel-2 data with climate information holds great potential for enhancing crop phenology estimation accuracy, supporting informed agricultural decisions, contributing to sustainable food production practices, and aiding in crop model calibrations and evaluations to meet the growing global food demand.
- - Understanding crop phenology is crucial for effective agricultural management strategies.
- - Efforts have been made to detect crop phenology using Remote Sensing (RS) and high resolution weather data.
- - A Machine Learning (ML) LightGBM model was trained to predict 13 phenological stages for eight major crops across Germany at a 20-meter scale.
- - The model achieved promising outcomes with predicted phenology having an R2 > 0.43 and a low Mean Absolute Error of 6 days across all phenological stages and crops.
- - Combining radar sensors with climate data led to significant improvements in the model's performance.
- - The fusion approach of Sentinel-1 and Sentinel-2 data with climate information has great potential for enhancing crop phenology estimation accuracy and supporting informed agricultural decisions.
SummaryUnderstanding crop phenology means knowing the different stages of plant growth, which is important for farmers to plan their crops. Scientists use technology like Remote Sensing and weather data to track these stages from far away. They trained a special computer program called LightGBM to predict when crops will grow in Germany. The program did a good job at guessing when plants would reach each stage of growth. By combining radar sensors and climate information, they made the program even better at predicting crop growth.
Definitions- Crop phenology: The study of different stages in the life cycle of plants.
- Remote Sensing (RS): Using technology to gather information about objects or areas from a distance.
- Machine Learning (ML): A type of artificial intelligence that allows computers to learn and improve from experience.
- R2: A statistical measure that shows how well data fits a regression model.
- Mean Absolute Error: A measure of how close predictions are to the actual values.
- Fusion approach: Combining different sources of data or information for better results.
Introduction
In the field of agriculture, understanding crop phenology is crucial for decision makers to effectively plan and adapt agricultural management strategies. Crop phenology refers to the timing of various stages in a plant's life cycle, such as germination, flowering, and maturity. These stages are influenced by environmental factors such as temperature, precipitation, and sunlight.
Traditionally, crop phenology has been monitored through on-site observations or manual surveys. However, with the increasing availability of big Earth observation data and advancements in technology, remote sensing (RS) techniques have become a valuable tool for accurately detecting crop phenology. RS involves using sensors mounted on satellites or aircraft to collect data about the Earth's surface.
In recent years, efforts have been made to develop methods that can accurately detect crop phenology using RS data combined with high resolution weather data. This has led to improved predictions at large scales but has not fully supported the needs of crop modeler communities who require more detailed information for their work.
To address this gap, a team of researchers developed a Machine Learning (ML) LightGBM model trained on Sentinel-1 and Sentinel-2 satellite data fused with high resolution climate data. The goal was to predict 13 different phenological stages for eight major crops across Germany at a 20-meter scale.
Methodology
The study used observed phenologies from the German national phenology network (German Meteorological Service; DWD) between 2017 and 2021 for training and validation purposes. The ML model was trained using LightGBM algorithm which is known for its fast processing speed and ability to handle large datasets.
A thorough feature selection analysis was conducted to identify the best combination of RS and climate data for detecting different phenological stages. This involved testing various combinations of features such as radar backscatter coefficients from Sentinel-1 images and optical raw band data from Sentinel-2 images along with climate variables like temperature and precipitation.
Results
The results of the study showed promising outcomes at a national scale, with predicted phenology achieving a reasonable precision of R2 > 0.43 and a low Mean Absolute Error of 6 days across all phenological stages and crops. This indicates that the ML model was able to accurately predict the timing of different crop stages.
The spatio-temporal analysis demonstrated the model's transferability across different spatial and temporal contexts in Germany. This means that the model can be applied to other regions or time periods with similar environmental conditions and still produce accurate predictions.
Furthermore, by combining radar sensors with climate data, the researchers were able to achieve significant improvements in predicting crop phenology compared to using RS data alone. This highlights the potential benefits of fusing multiple data sources for agricultural applications.
Insights from previous studies were also taken into consideration during feature selection analysis. These studies have shown that data volume plays an important role in influencing ML model accuracy, as well as the impact of specific feature sets on predicting different phenological stages of crops. For instance, radar backscatter coefficients combined with optical raw band data were found to be valuable for identifying early stages in winter wheat crops.
Conclusion
In conclusion, this novel fusion approach of Sentinel-1 and Sentinel-2 data with climate information holds great potential for enhancing crop phenology estimation accuracy. The results from this study can support informed agricultural decisions, contribute to sustainable food production practices, and aid in crop model calibrations and evaluations to meet the growing global food demand.
This research has shown that by leveraging big Earth observation data and advanced ML techniques, it is possible to accurately predict crop phenology at a fine scale. The use of satellite imagery combined with climate variables has proven to be a powerful tool for monitoring agriculture on a large scale.
Future research could focus on expanding this approach to other regions or countries where agriculture plays a significant role in their economy. Additionally, incorporating more diverse data sources and improving the ML model's ability to handle complex features could further enhance its accuracy.
Overall, this study highlights the importance of utilizing advanced technologies and data fusion techniques in agriculture to support sustainable food production and decision making. With the increasing availability of big Earth observation data, there is great potential for further advancements in crop phenology prediction methods that can benefit farmers, researchers, and decision makers alike.