Data-Efficient Hyperspectral Unmixing for Agricultural Inspection and Flood Adaptation Analysis

Abstract

Data efficiency is important in both hyperspectral agricultural inspection and flood adaptation analysis because labeled data are often limited, costly, or unevenly distributed. In food inspection, hyperspectral unmixing supports meat contamination detection by decomposing mixed spectral signatures and identifying weak contaminant signals. In flood adaptation, agent-based relocation models support analysis of household behavior, risk perception, and pre-disaster relocation under uncertainty. This topic connects these domains through data-efficient modeling, interpretable evidence extraction, and decision-support relevance. Hyperspectral benchmarks provide controlled evaluation for weak-signal detection, while flood relocation models provide structured behavioral simulation for climate risk planning. The literature direction highlights the need for models that can operate under data constraints while producing interpretable results for safety and governance decisions.