Multispectral high resolution sensor fusion for smoothing and gap-filling in the cloud

Area of Science:

• Remote Sensing
• Earth Observation
• Geospatial Analysis

Background:

• Operational satellite optical sensors face trade-offs between spectral, spatial, and temporal resolutions.
• Clouds and aerosols contaminate land surface observations, impacting data quality.

Purpose of the Study:

• To develop a Highly Scalable Temporal Adaptive Reflectance Fusion Model (HISTARFM) algorithm.
• To produce monthly, gap-free, high-resolution (30m) land surface observations by fusing multisensor satellite imagery.
• To reduce noise and improve land surface reflectance estimates.

Main Methods:

• Fusion of Landsat (30m) and MODIS (500m) multispectral images.
• Implementation of a bias-aware Kalman filter method within the Google Earth Engine (GEE) platform.
• Estimation of uncertainty in reflectance, enabling error propagation analysis.

Main Results:

• Generated fused images at Landsat spatial resolution with enhanced spatio-temporal coverage.
• Bias correction in Kalman filter estimates accounted for temporally auto-correlated errors.
• Reliable uncertainty estimation for final reflectance products.

Conclusions:

• The HISTARFM algorithm effectively fuses multisensor data for improved land surface monitoring.
• The approach enables operational applications requiring enhanced spatio-temporal resolutions at continental scales.
• Validated through quantitative and qualitative evaluations against state-of-the-art methods.