Monitoring Soil Copper in Urban Land Using Visibale and Near-Infrared Spectroscopy with Spatially Nearby Samples

Abstract

Soil heavy metal contamination in urban land can affect biodiversity, ecosystem functions, and the health of city residents. Visible and near-infrared (Vis-NIR) spectroscopy is fast, inexpensive, non-destructive, and environmentally friendly compared to traditional methods of monitoring soil Cu, a common heavy metal found in urban soils. However, there has been limited research on using spatially nearby samples to build the Cu estimation model. Our study aims to investigate how spatially nearby samples influence the Cu estimation model. In our study, we collected 250 topsoil samples (0–20 cm) from China’s third-largest city and analyzed their spectra (350–2500 nm). For each unknown validation sample, we selected its spatially nearby samples to construct the Cu estimation model. The results showed that compared to the traditional method (Rp2 = 0.75, RMSEP = 8.56, RPD = 1.73), incorporating nearby samples greatly improved the model (Rp2 = 0.93, RMSEP = 4.02, RPD = 3.89). As the number of nearby samples increased, the performance of the Cu estimation model followed an inverted U-shaped curve—initially increasing and then declining. The optimal number of nearby samples is 125 (62.5% of the total), and the mean distance between validation and calibration samples is 17 km. Therefore, we conclude that using nearby samples significantly enhances the Cu estimation model. The optimal number of nearby samples should strike a balance, covering a moderate area without there being too few or too many.

Keywords

• soil spectroscopy
• proximal sensing
• heavy metal
• spatial analysis