Water Quality Chl-a Inversion Based on Spatio-Temporal Fusion and Convolutional Neural Network

Abstract

The combination of remote sensing technology and traditional field sampling provides a convenient way to monitor inland water. However, limited by the resolution of remote sensing images and cloud contamination, the current water quality inversion products do not provide both high temporal resolution and high spatial resolution. By using the spatio-temporal fusion (STF) method, high spatial resolution and temporal fusion images were generated with Landsat, Sentinel-2, and GaoFen-2 data. Then, a Chl-a inversion model was designed based on a convolutional neural network (CNN) with the structure of 4-(136-236-340)-1-1. Finally, the results of the Chl-a concentrations were corrected using a pixel correction algorithm. The images generated from STF can maintain the spectral characteristics of the low-resolution images with the R2 between 0.7 and 0.9. The Chl-a inversion results based on the spatio-temporal fused images and CNN were verified with measured data (R2 = 0.803), and then the results were improved (R2 = 0.879) after further combining them with the pixel correction algorithm. The correlation R2 between the Chl-a results of GF2-like and Sentinel-2 were both greater than 0.8. The differences in the spatial distribution of Chl-a concentrations in the BYD lake gradually increased from July to August. Remote sensing water quality inversion based on STF and CNN can effectively achieve high frequency in time and fine resolution in space, which provide a stronger scientific basis for rapid diagnosis of eutrophication in inland lakes.

Meta-Analysis of Satellite Observations for United Nations Sustainable Development Goals: Exploring the Potential of Machine Learning for Water Quality Monitoring

“…Additionally, recurrent neural networks, such as LSTM, and GRU have also gained significant popularity in this field due to their ability to handle sequential data and capture temporal dependencies [22,42,48,194,249,266]. CNNs, such as the well-known convolutional neural network [43,194,220,257,260,262,264,265], the DINCAE [50,255,259], and generative adversarial neural networks (GANs) [277], may not have been widely used, but they have a significant advantage in dealing with spatial dependencies.…”

Long-Term Monitoring of Inland Water Quality Parameters Using Landsat Time-Series and Back-Propagated ANN: Assessment and Usability in a Real-Case Scenario

“…[64] used UAV images to build a CNN model and obtained an R 2 of 0.79 (RMSE: 8.76), while [65] used Sentinel 2 images and Ada boost regression resulting in a R 2 of 0.90 (RMSE: 1.48). Hafeez et al [20] used ANN achieving an NRMSE of 5.1% (R 2 : 0.87, RMSE: 1.4) while [66] reached an R 2 of 0.88 using CNN and Sentinel 2 and Geo-Fan 2.…”

“…However, the training of NN models requires a large training dataset, otherwise, they may lead to overfitting or underfitting, which greatly limits the extraction of general rules and the generalization ability of the model [69]. Taking into account that most of the previously analyzed papers used small training data sets, such as 125 [22], 60 [25], 155 [64], and 92 [66] samples, it was expected that the proposed method would have a higher accuracy. In addition, the selection of optimal input data as well as the usage of the large time series covering a wide variety of conditions and an early stopping function [70] to avoid overfitting probably had an impact on the increase in the model accuracy.…”

Mapping the Continuous Cover of Invasive Noxious Weed Species Using Sentinel-2 Imagery and a Novel Convolutional Neural Regression Network

“…Therefore, as vegetation is more appropriately described by continuous metrics, flexible and robust methods that can map target vegetation variables (invasive species cover) to a continuous scale rather than a discrete class are urgently needed. Recently, the convolutional neural network-based regression (CNNR) model has also been widely applied in the estimation of continuous metrics across multiple disciplines, achieving promising results and outperforming the traditional regression model [46,[69][70][71][72].…”