Influence of Soil Moisture and Rainfall on the Efficiency of a Near-IR Hyperspectral Oil Pollution Detection Technique

Abstract

Monitoring of oil pollution of the land surface is one of the pressing ecological issues today. The work is devoted to the experimental study of hyperspectral technique for detecting oil pollution on the land surface in the near-IR. The spectral brightness coefficients of soil samples contaminated with different oil product types are experimentally measured in the 1.6–2.5 μm spectral range. The influence of soil moisture and rainfall on the reflection spectra of soils (several types of sand and soil from forest and park areas) contaminated by oil products (of Moscow and Samara oil processing plants, kerosene, gas condensate, various gasoline brands, motor oils, and diesel fuel) is studied. It is shown that spectral dips near 1.73 and 2.3 μm (typical for soils contaminated with oil products) in most cases remain in the reflectance spectra under conditions of moderately moist soil, moderate rain, and even heavy rain. A specially created neural network shows the probability of detecting oil pollution on the land surface to be more than 99% under conditions of moderately moist soil and moderate rain and more than 88% under conditions of heavy rain and moist soil for 14 spectral channels 10 nm wide in the 1.6–2.4 μm range. The results can be used in the development of pipeline leak remote monitoring systems.