A Lightweight UAV-based Spectroscopic System for Water Quality Monitoring in Mining-Impacted Environments: Setup, Data Processing, and Validation

Monitoring water quality in areas affected by mining activities requires high-spatial and temporal resolution data, which remains a challenge for traditional satellite and ground-based methods. We present a novel, cost-effective instrumental setup for water surface reflectance measurements using a miniature light-weight Ocean Optics STS-VIS microspectrometer (40 x 42 x 24 mm; 337–823 nm; 1.2 nm spectral resolution; 1024 bands). The sensor was mounted on the DJI Phantom 3 Advanced UAV using a custom-developed, 3D-printed holder to ensure stability and precise nadir orientation. With a field of view (FOV) of 25° and an operational flight altitude of 3 m, the system achieved a spatial resolution (ground footprint) of 1.2 m per measurement point, allowing for precise targeting of narrow water bodies. The system was used across three diverse mining regions: the Chalkidiki Peninsula and Kirki (Greece), and Andalusia (Spain).

To derive accurate reflectance from raw intensity data, a standardized calibration protocol was established, involving dark spectrum subtraction and reference measurements using a Spectralon panel (Spectral Evolution; 100% reflectance). Flights were conducted manually to minimize propeller propeller-induced surface turbulence, following standardized patterns (longitudinal and diagonal for streams; sun-relative for water bodies).

The processing workflow addresses the high volume of raw data (up to 400 spectra per site). We implemented a smoothing pipeline and filtration:

(1) application of Savitzky-Golay filters (SGF) with a 2nd-degree polynomial and varying window sizes (66, 99, 132),

(2) statistical outlier removal based on +/-1.5 standard deviations,

(3) visual inspection eliminating interference from bank vegetation or rocks above the water.

Based on our analysis, the SGF window size of 99 was selected as optimal. While the window size 66 left significant residual noise and the window size of 132 caused the loss of critical spectral absorption features, the window size of 99 provided sufficient noise reduction while preserving the integrity of the spectral signal.

The final averaged spectra were correlated with water sample laboratory analyses using Partial Least Squares Regression (PLSR). Our results identified key wavelengths sensitive to specific mining-related water quality parameters. This study demonstrates that the proposed UAV-spectrometer integration provides a robust, flexible, and high-precision alternative for monitoring contaminated aquatic systems in logistically challenging environments.

The presented analysis was conducted under the support of the EC through the MultiMiner project, funded under the European Union’s Horizon Europe research and innovation programme (Grant Agreement No. 10109137474), and under the support of the MINEYE project, funded under the European Union’s Horizon Europe research and innovation programme (Grant Agreement No. 101138456).