Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Estimation of Fish Catch Using Sentinel-2, 3 and XGBoost-Kernel-Based Kernel Ridge Regression

Kanu Mohammed, Vaishnavi Joshi, Pranjali Diliprao Patil, Sandipan Mondal, Ming-An Lee, Subhadip Dey

TL;DR

This work addresses estimating fish catch from satellite data by linking oceanographic conditions to fish distributions using a non-linear kernel method. It introduces a novel XGBoost-derived kernelized Kernel Ridge Regression (KRR) that leverages one-hot embeddings from an ensemble of trees to form a data-driven kernel, enabling effective capture of environment–fish relationships. Across Sentinel-2 and Sentinel-3 imagery, the proposed XGBoost-KRR approach achieves superior accuracy (e.g., RMSE reductions and high correlations with in-situ data, $\rho$ of 0.924 for S2 and 0.731 for S3) and reveals complementary spectral sensitivities: Sentinel-2 provides fine-scale variability while Sentinel-3 reflects bulk water-quality signals. The resulting spatial fish catch maps align with known oceanographic processes, supporting scalable, sensor-agnostic fisheries monitoring and contributing to SDG targets on Zero Hunger and Life Below Water.

Abstract

Oceanographic factors, such as sea surface temperature and upper-ocean dynamics, have a significant impact on fish distribution. Maintaining fisheries that contribute to global food security requires quantifying these connections. This study uses multispectral images from Sentinel-2 MSI and Sentinel-3 OLCI to estimate fish catch using an Extreme Gradient Boosting (XGBoost)-kernelized Kernel Ridge Regression (KRR) technique. According to model evaluation, the XGBoost-KRR framework achieves the strongest correlation and the lowest prediction error across both sensors, suggesting improved capacity to capture nonlinear ocean-fish connections. While Sentinel-2 MSI resolves finer-scale spatial variability, emphasizing localized ecological interactions, Sentinel-3 OLCI displays smoother spectral responses associated with poorer spatial resolution. By supporting sustainable ecosystem management and strengthening satellite-based fisheries assessment, the proposed approach advances SDGs 2 (Zero Hunger) and 14 (Life Below Water).

Estimation of Fish Catch Using Sentinel-2, 3 and XGBoost-Kernel-Based Kernel Ridge Regression

TL;DR

This work addresses estimating fish catch from satellite data by linking oceanographic conditions to fish distributions using a non-linear kernel method. It introduces a novel XGBoost-derived kernelized Kernel Ridge Regression (KRR) that leverages one-hot embeddings from an ensemble of trees to form a data-driven kernel, enabling effective capture of environment–fish relationships. Across Sentinel-2 and Sentinel-3 imagery, the proposed XGBoost-KRR approach achieves superior accuracy (e.g., RMSE reductions and high correlations with in-situ data, of 0.924 for S2 and 0.731 for S3) and reveals complementary spectral sensitivities: Sentinel-2 provides fine-scale variability while Sentinel-3 reflects bulk water-quality signals. The resulting spatial fish catch maps align with known oceanographic processes, supporting scalable, sensor-agnostic fisheries monitoring and contributing to SDG targets on Zero Hunger and Life Below Water.

Abstract

Oceanographic factors, such as sea surface temperature and upper-ocean dynamics, have a significant impact on fish distribution. Maintaining fisheries that contribute to global food security requires quantifying these connections. This study uses multispectral images from Sentinel-2 MSI and Sentinel-3 OLCI to estimate fish catch using an Extreme Gradient Boosting (XGBoost)-kernelized Kernel Ridge Regression (KRR) technique. According to model evaluation, the XGBoost-KRR framework achieves the strongest correlation and the lowest prediction error across both sensors, suggesting improved capacity to capture nonlinear ocean-fish connections. While Sentinel-2 MSI resolves finer-scale spatial variability, emphasizing localized ecological interactions, Sentinel-3 OLCI displays smoother spectral responses associated with poorer spatial resolution. By supporting sustainable ecosystem management and strengthening satellite-based fisheries assessment, the proposed approach advances SDGs 2 (Zero Hunger) and 14 (Life Below Water).
Paper Structure (5 sections, 10 equations, 4 figures, 2 tables)

This paper contains 5 sections, 10 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Dataset
  4. Results and Discussion
  5. Conclusions

Figures (4)

  • Figure 1: Violin plots of different spectral bands of (a) Sentinel-2 and, (b) Sentinel-3 data. The pixel values of Sentinel-2 are reflectance and of Sentinel-3 are radiance values.
  • Figure 2: The distribution of fish catch dataset in kilograms used in this study. The thick dark line is the kernel density estimation of the data
  • Figure 3: Correlation plots between in-situ fish catch and predicted fish catch for (a) Sentinel-2 and (b) Sentinel-3 dataset.
  • Figure 4: Estimated fish catch towards the coast of Easternmost point of Taiwan using (a) Sentinel-2 and (b) Sentinel-3 dataset. The blue to red colormap depicts the low to high estimated fish catch.