研究目的
To detect anomalies or interventions in forest monitoring through spatial and temporal forecast of satellite time series data.
研究成果
The results demonstrate the applicability of dense time series of remote sensing data for large area forest monitoring with spatial-explicit wall-to-wall coverage. The methods allow fully automatic processing, which is a specific requirement for future forest monitoring applications based on satellite imagery with high spatial and high temporal resolution such as Sentinel2 time series.
研究不足
Processing time is a drawback for functional time series analysis methods, requiring about 8 hours per study area on a usual PC. The ARIMA method is less sensitive to sub-pixel changes compared to the functional time series analysis approach.
1：Experimental Design and Method Selection:
The study uses MODIS product MCD43A4, Version 5, for reflectance data. Methods include functional data analysis, functional principal component analysis, random regression forests with online learning, functional time series analysis, and autoregressive integrated moving average model.
2：Sample Selection and Data Sources:
Two study areas were selected, one in Germany for monitoring forest damages caused by wind-storm, and another in Spain for monitoring forest fires. MODIS time series data from 2000 to 2013 were used.
3：List of Experimental Equipment and Materials:
MODIS product MCD43A4, Version 5, for reflectance data.
4：Experimental Procedures and Operational Workflow:
Data pre-processing included Savitzky-Golay filtering for noise reduction and gap-filling. Three types of forecasts (spatial, temporal, and spatio-temporal) were provided for each pixel.
5：Data Analysis Methods:
Functional time series analysis, autoregressive integrated moving average model, and online random regression forests were used for data analysis.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
