DeepSalt: Bridging Laboratory and Satellite Spectra through Domain Adaptation and Knowledge Distillation for Large-Scale Soil Salinity Estimation

2025-10-27

AI Summary

Soil salinization severely constrains agricultural productivity and ecological security; while its spectral response enables remote sensing monitoring, laboratory hyperspectral measurements offer high accuracy but poor scalability, whereas satellite hyperspectral data provide broad coverage yet suffer from limited spatial resolution and interpretability. To bridge this gap, we propose a deep domain adaptation framework integrating a spectral adaptive unit with knowledge distillation—marking the first successful transfer of knowledge from high-fidelity laboratory spectra to operational satellite hyperspectral data. Our method jointly models spectral shape discrepancies and physically grounded constraints to enhance cross-sensor and cross-regional generalizability. Experiments demonstrate that, without requiring dense ground sampling, our approach significantly outperforms conventional remote sensing inversion methods and domain-unadapted baselines, achieving consistent improvements in estimation accuracy (R² increased by 0.12–0.21) across multiple representative saline-alkali regions. This work establishes a new paradigm for large-scale, high-accuracy, and physically interpretable soil salinity monitoring via remote sensing.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Application Domains: Internet of Things, Sensor Networks & Smart Cities

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

Soil salinization poses a significant threat to both ecosystems and agriculture because it limits plants' ability to absorb water and, in doing so, reduces crop productivity. This phenomenon alters the soil's spectral properties, creating a measurable relationship between salinity and light reflectance that enables remote monitoring. While laboratory spectroscopy provides precise measurements, its reliance on in-situ sampling limits scalability to regional or global levels. Conversely, hyperspectral satellite imagery enables wide-area observation but lacks the fine-grained interpretability of laboratory instruments. To bridge this gap, we introduce DeepSalt, a deep-learning-based spectral transfer framework that leverages knowledge distillation and a novel Spectral Adaptation Unit to transfer high-resolution spectral insights from laboratory-based spectroscopy to satellite-based hyperspectral sensing. Our approach eliminates the need for extensive ground sampling while enabling accurate, large-scale salinity estimation, as demonstrated through comprehensive empirical benchmarks. DeepSalt achieves significant performance gains over methods without explicit domain adaptation, underscoring the impact of the proposed Spectral Adaptation Unit and the knowledge distillation strategy. The model also effectively generalized to unseen geographic regions, explaining a substantial portion of the salinity variance.

Problem

• Bridging laboratory and satellite spectral data for soil salinity estimation
• Overcoming limited scalability of laboratory spectroscopy for large-scale monitoring
• Transferring high-resolution spectral insights to satellite hyperspectral sensing

Innovation

• Knowledge distillation transfers lab spectra to satellites
• Spectral Adaptation Unit enables cross-domain feature alignment
• Deep learning framework eliminates need for ground sampling