1. Project Overview
In April 2024, Ramoton Company utilized a drone-mounted multispectral remote sensing system to acquire multispectral data from a river in Qingdao, while simultaneously collecting in-situ water sample data. Through preprocessing of multispectral images, pollution degree analysis, and water sample data testing, a qualitative assessment of water environment pollution and quantitative inversion of water quality parameters were conducted, providing relevant evidence for monitoring the water environment of the river.
2. Data Collection
Drone Multispectral Data Collection: This phase employed the M300 RTK + RT600 Pro multispectral remote sensing system for data collection, which has a flight time of up to 35 minutes and can cover approximately 5 km² in a single flight.
Flight Parameters
| Indicator | Value |
|---|---|
| Flight Altitude | 120 m |
| Flight Speed | 7 m/s |
| Longitudinal Overlap | 80% |
| Lateral Overlap | 70% |
Water Sample Data Collection: Water samples from the river were collected simultaneously with drone data using a submerged sampler, with each sample taken at a depth of 0.5 m below the water surface.
In this multispectral water quality detection, nine water quality indicators were measured: chlorophyll a, total phosphorus, total nitrogen, permanganate index, transparency, dissolved oxygen, chemical oxygen demand, turbidity, and total suspended solids. Dissolved oxygen and transparency were measured on-site, while the remaining water quality parameters were tested in the laboratory, following national or industry standards such as “spectrophotometry,” “probe method,” and “gravimetric method.”
3. Data Preprocessing
Initially, the raw multispectral data was preprocessed using specialized software, undergoing three steps: band registration, stitching, and reflectance calibration, ultimately generating a multispectral orthophoto (DOM) for the surveyed area.
4. Eutrophication and Black Odor Water Pollution Analysis
Using the self-developed industry application software’s “River Ecology – Eutrophication Analysis Module” and “Black Odor Water Pollution Analysis Module,” the preprocessed TIFF images were analyzed, allowing for the one-click export of graded raster files and thematic maps for “Eutrophication/Black Odor Water.”
4.1 Eutrophication Analysis
Eutrophication refers to the excessive accumulation of nutrients (such as nitrogen and phosphorus) in water bodies, leading to the overgrowth of algae and aquatic plants. This phenomenon can result in excessive algal growth, depleting oxygen in the water and causing hypoxia, disrupting the balance of aquatic ecosystems, and even leading to water body death.
According to the black odor water pollution identification results, the river predominantly exhibited a non-black odor state. Combined with ground measurement data, the ammonia nitrogen and dissolved oxygen indicators were far from reaching black odor levels. Comprehensive analysis indicates that the algorithm for identifying black odor water bodies performed well.
5. Quantitative Inversion of Water Quality Parameters
Based on the collected DJI Matrice 350 drone multispectral images and water quality data, data analysis was conducted to establish a quantitative relationship model between water quality parameters and multispectral remote sensing data, achieving quantitative inversion of various water quality parameters for the target water body. The distribution of sampling points is shown in the figure below.
The nine measured water quality indicators include chlorophyll a, total phosphorus, total nitrogen, permanganate index, transparency, dissolved oxygen, chemical oxygen demand, turbidity, and total suspended solids, with the quantitative inversion results shown in the figure below.
The accuracy of the inversion results was evaluated using Mean Relative Error (MRE) and Root Mean Square Error (RMSE), with the accuracy assessment results for each indicator presented in Table 2.
| Water Quality Parameter | Average Relative Error | RMSE |
|---|---|---|
| Chlorophyll a | 7.87% | 1.25 |
| Total Phosphorus | 10.75% | 0.05 |
| Total Nitrogen | 44.54% | 0.83 |
| Permanganate Index | 19.68% | 0.43 |
| Transparency | 12.54% | 0.03 |
| Dissolved Oxygen | 12.71% | 0.66 |
| Chemical Oxygen Demand | 27.96% | 1.32 |
| Turbidity | 8.25% | 0.02 |
| Total Suspended Solids | 8.15% | 0.02 |
From the table, it can be seen that, with the exception of total nitrogen, the average relative error of each water quality parameter is below 30%, meeting the needs for routine and high-frequency global water quality monitoring.
Although the average relative error for total nitrogen is relatively large, the RMSE magnitude is relatively small. Considering the measured total nitrogen content, it is evident that the distribution gradient of total nitrogen values in this river is uneven, thus affecting the sensitivity of total nitrogen inversion. Future algorithm optimizations will better balance data distribution and handle data variability to ensure the inversion accuracy of all water quality parameters.
6. Project Summary
The DJI M350 drone multispectral remote sensing technology can acquire rich spectral information about ground objects, offering unique advantages and broad application prospects in river water quality environment monitoring. Ongoing research on drone multispectral remote sensing and application algorithms continues to deepen. Compared to satellite remote sensing images, drone multispectral data provides more precise, flexible, and higher temporal and spatial resolution data acquisition, suitable for remote monitoring tasks targeting rivers and shorelines. Compared to traditional manual sampling methods, multispectral remote sensing technology can reduce the costs of obtaining river water quality parameters and provide a global display of water quality conditions in the target area, avoiding the errors associated with “point-based assessments” of water environments.
