Remote sensing data imputation using deep learning for multispectral imagery

2026-05-19

AI Summary

This study addresses the challenge of spectral data gaps in optical remote sensing imagery caused by cloud cover, which hinders effective monitoring of aquatic events such as algal blooms. To mitigate this issue, the authors propose a deep learning–based approach for reconstructing missing multispectral bands. They systematically evaluate the gap-filling performance of several architectures—including CNN, Inception ResNet, Autoencoder, and hybrid models integrating LSTM—on PlanetScope SuperDove imagery, and apply the reconstructed data for the first time to estimate algal bloom indicators. Experimental results demonstrate that the proposed methods significantly outperform conventional linear interpolation, with the CNN-based model achieving the best performance. Furthermore, Green/Red ratios and Normalized Difference Chlorophyll Index (NDCI) derived from the reconstructed data show strong agreement with in situ measurements, confirming the method’s effectiveness in enhancing the reliability of water quality monitoring via remote sensing.

Abstract

Remote sensing techniques have been increasingly utilised in aquatic applications in recent years. A common challenge in using optical satellite data is the presence of missing observations due to cloud cover. These data gaps can lead to missed detection of critical events, such as algal blooms, in lakes of high interest to water authorities. As a result, enhancing the completeness of optical satellite datasets is crucial for improving the monitoring and prediction of algal blooms. In this study, we compared a traditional data imputation method (i.e., linear interpolation) with deep learning models for reconstructing missing spectral bands across four lakes with historical records of algal blooms. The deep learning models adopted include CNN-based architectures (i.e., CNN, Inception Resnet, and Autoencoder) and CNN-LSTM-based architectures (i.e., CNN-LSTM, Resnet-LSTM, and Autoencoder-LSTM). Our results demonstrated that deep learning models substantially outperformed the baseline linear interpolation method in imputing spectral band values within artificially masked regions. Among these models, CNN delivered the best performance across most lakes. Furthermore, we evaluated the performance of algal bloom indices (i.e., Green/Red and NDCI) derived from the imputed imagery by comparing them with the observed data. Our results demonstrate that deep learning models are effective for imputing missing data in PlanetScope SuperDove imagery, enabling more reliable applications in water monitoring.

Problem

• remote sensing
• data imputation
• cloud cover
• algal blooms
• multispectral imagery

Innovation

• deep learning
• data imputation
• multispectral imagery
• algal bloom monitoring
• CNN-LSTM