Urban rivers provide essential benefits, from supporting biodiversity and regulating local climates to supplying water and improving residents' quality of life. Yet many are increasingly threatened by eutrophication, a process driven by excessive nutrients that can trigger algal growth, oxygen depletion, and declining water quality.
Researchers have now developed a high-resolution monitoring approach that combines drone-based multispectral imaging with an ensemble of machine learning and deep learning models. The method can rapidly estimate chlorophyll-a concentrations across urban waterways while revealing how surrounding industrial, agricultural, residential, and green areas influence water quality.
Chlorophyll-a is the main photosynthetic pigment in algae and is widely used as an indicator of algal abundance and eutrophication. Conventional monitoring generally requires collecting water samples and analyzing them in laboratories. Although accurate, this process can be costly, time-consuming, and unable to capture fine-scale variations across an entire river.
"Our approach allows us to move from a limited number of sampling points to detailed maps showing how chlorophyll-a changes throughout an urban water body," said corresponding author Bin Yang. "By combining flexible drone observations with multiple artificial intelligence models, we can obtain information that is both spatially precise and useful for environmental management."
The researchers examined urban rivers in Harbin, Qiqihar, and Suihua in Heilongjiang Province, China. They collected 57 water samples and acquired multispectral images using a camera mounted on an unmanned aerial vehicle flying at an altitude of 100 meters. The resulting images had a spatial resolution of approximately 4.5 centimeters, enabling detailed observation of localized water-quality patterns.
To interpret the imagery, the team developed an ensemble machine and deep learning, or EMD, method. It integrates support vector machine, random forest, AdaBoost, and multilayer perceptron models. Instead of assigning the same importance to each model everywhere, the system dynamically adjusts model weights for individual pixels according to local spectral characteristics.
Across repeated tests, the EMD method achieved an average coefficient of determination of 0.797 and an average root mean square error of 18.96 milligrams per cubic meter, outperforming the individual models in overall accuracy and stability. Spectral indices derived from the drone imagery further improved prediction performance.
The resulting maps also highlighted major differences among the three cities. Average chlorophyll-a concentrations were approximately 12.57 milligrams per cubic meter in Harbin and 12.00 in Qiqihar, but reached 28.43 milligrams per cubic meter in Suihua.
Land-use analysis suggested that industrial development was the dominant factor behind the elevated concentrations in Suihua. Rivers near industrial areas showed substantially higher chlorophyll-a levels than those mainly influenced by agricultural land or mixed residential and green spaces. Well-planned green areas appeared to help reduce nutrient inputs by intercepting runoff, while effective sewage treatment limited the influence of residential development.
The findings show that urban eutrophication cannot be addressed through a single management strategy. Industrial areas may require stricter wastewater controls, agricultural zones may benefit from reduced fertilizer use and ecological drainage systems, and densely populated districts may need improved sewage infrastructure and expanded green buffers.
The researchers note that seasonal differences and the relatively limited number of field samples should be considered when interpreting the results. Future studies could integrate satellite observations, additional seasonal sampling, and larger shared datasets.
===
Journal reference: He A; Yang B; Qu Q; et al. Efficient monitoring of chlorophyll-a concentration in urban water bodies based on UAV multispectral images and ensemble machine and deep learning method. AI Environ. 2026, 1(2): 93-105. DOI: 10.66178/aie-0026-0007
https://www.the-newpress.com/aie/article/doi/10.66178/aie-0026-0007
===
About the Journal:
Artificial Intelligence & Environment is an international multidisciplinary platform for communicating advances in fundamental and applied research on the intersection of environmental science and artificial intelligence (AI). It is dedicated to serving as an innovative, efficient and professional platform for researchers in the cross-discipline fields of earth and environmental sciences, big data science and AI around the world to deliver findings from this rapidly expanding field of science. It is a peer-reviewed, open-access journal that publishes critical review, original research, rapid communication, view-point, commentary and perspective papers.
