AI Model Predicts Algal Blooms in India’s Ashtamudi Estuary from Just Two Nutrients

On the south-west coast of India, where the Kallada River meets the Arabian Sea, lies one of the country’s most ecologically treasured water bodies: the Ashtamudi Estuary. Shaped like a palm with eight branching arms, this microtidal estuary is the second largest and deepest wetland ecosystem in the region and holds international recognition as a Ramsar wetland of importance. Yet beneath its biological richness, a quieter crisis has been unfolding. Decades of untreated sewage discharge, agricultural runoff laden with fertilizers and pesticides, coconut husk retting, overfishing, dredging and land reclamation have pushed the estuary toward a state of nutrient overload that scientists now warn could fundamentally alter its ecology.

A new study published in the journal Estuarine Management and Technologies has taken a detailed statistical and machine-learning approach to understanding this problem, focusing on chlorophyll-a, the green pigment at the heart of photosynthesis in algae and phytoplankton. Chlorophyll-a is far more than a biochemical curiosity. Because its concentration rises and falls with the abundance of phytoplankton, it serves as one of the most reliable indicators of eutrophication, the process by which excess nutrients fuel runaway algal growth. Elevated chlorophyll-a signals nutrient enrichment, and with it the risk of oxygen depletion, loss of water transparency, habitat degradation and harmful algal blooms that can raise the trophic state index of a water body and threaten fisheries that local communities depend upon.

The research team, drawing on water quality data supplied by the Centre for Water Resources Development and Management in Kozhikode, Kerala, assembled an unusually rich dataset spanning 41 sampling stations distributed from the mouth of the estuary to its upstream reaches along the Kallada River. Measurements were taken monthly between 2021 and 2023, covering the monsoon, post-monsoon and pre-monsoon seasons, and included parameters such as phosphate, nitrate, salinity, sulphate, dissolved oxygen, electrical conductivity, pH, calcium, chloride, chemical oxygen demand, magnesium and temperature. Before analysis, the researchers screened the data using box plots to detect and remove outliers, trimming roughly fifteen percent of values from either end of the distributions to prevent extreme readings from distorting the statistical picture.

The seasonal story that emerged is striking. Chlorophyll-a concentrations peaked during the pre-monsoon period, followed by the post-monsoon and finally the monsoon season, which registered the lowest levels. The explanation lies in the interplay of light, temperature and hydrology. During the pre-monsoon months, abundant solar radiation and long daylight hours create ideal conditions for photosynthesis, while elevated temperatures raise metabolic rates and accelerate chlorophyll synthesis. Nutrient enrichment through stratification and tidal mixing compounds these effects. The monsoon, by contrast, floods the estuary with freshwater that dilutes chlorophyll concentrations, and the turbidity generated by sediment runoff attenuates light penetration, starving phytoplankton of the energy they need. Although runoff initially delivers a pulse of nutrients, rapid flushing quickly depletes them before phytoplankton populations can respond.

Spatial patterns proved equally revealing. Using inverse distance weighting interpolation in QGIS to generate continuous concentration surfaces from the geo-referenced sampling points, the researchers mapped chlorophyll-a across the estuary for each season. During the pre-monsoon, levels peaked along the coastal regions, driven primarily by untreated sewage from urban areas, industrial discharges and agricultural runoff. During the monsoon, concentrations shifted to the southern region, where runoff delivered nutrients but the presence of islands impeded flushing and allowed them to accumulate. Post-monsoon maps showed elevated levels in the southern and central regions, again attributed to nutrient retention facilitated by retarded flow and the estuary’s complex geomorphology. The maps make clear that both human pollution sources and the physical shape of the estuary jointly determine where algal growth concentrates.

To classify the estuary’s overall condition, the team applied Carlson’s Trophic State Index, a widely used metric that grades water bodies from oligotrophic, meaning clear and oxygen-rich, through mesotrophic and eutrophic, to hypereutrophic, where heavy algal blooms, dense macrophyte beds and even summer fish kills become routine. The observed index values for Ashtamudi ranged from 50 to 70, placing the estuary firmly in the eutrophic to hypereutrophic range. This classification is consistent with previous studies of tropical estuaries, but the temporal analysis raises an important question of whether the eutrophic condition is a persistent long-term state or an emerging trend accelerated by recent anthropogenic pressures.

The statistical core of the study came from Pearson correlation analysis, applied after testing the data for normality and selecting only parameters significant at the 95 percent confidence level. One result stood out above all others: phosphate exhibited the strongest positive correlation with chlorophyll-a, with a coefficient of 0.893, significant at the 0.01 level. Nitrate followed with a correlation of 0.586, while salinity, sulphate, electrical conductivity and dissolved oxygen showed moderate associations. Dissolved oxygen, notably, correlated negatively, which the authors caution cannot be interpreted directly because oxygen levels fluctuate with organic matter decomposition and water turbulence independently of algal biomass. Electrical conductivity and salinity move in lockstep, so the team settled on phosphates, nitrates, salinity and sulphate as the significant drivers of chlorophyll-a in the estuary.

The dominance of phosphate carries real ecological weight. In eutrophic systems, phosphorus frequently becomes the limiting nutrient, the resource whose scarcity constrains further algal growth, while nitrogen tends to act as a co-limiting factor under hypereutrophic conditions. The spatial analysis reinforced this interpretation: variations in chlorophyll-a across the estuary tracked phosphate concentrations almost identically, whereas the chlorophyll-nitrate relationship was far looser. The sources of this phosphate are well documented in earlier work on the lake, including urban sewage discharge, coconut husk retting activities in the estuarine region, rainfall-driven land runoff and the release of phosphorus from silty, organic-rich sediments at the lake bottom. Nitrate, meanwhile, arrives through nitrification in estuarine sediments, oxidation of ammonia, domestic practices and agricultural fertilizer use.

With the key drivers identified, the researchers turned to artificial neural networks, computational models inspired by the human brain that excel at capturing the nonlinear relationships typical of estuarine systems, where tidal dynamics, freshwater inflow, nutrient loading and seasonal variability interact in complicated ways. Using MATLAB’s Neural Network Toolbox with the Levenberg-Marquardt feedforward backpropagation algorithm, the team trained models on 70 percent of the data, validated on 15 percent and tested on the remaining 15 percent. Five scenarios were tested, ranging from a five-input model using phosphates, nitrates, sulphate, salinity and dissolved oxygen down to a phosphate-only model. The first four scenarios all achieved coefficients of determination around 0.99, with root mean square errors below 0.05, and the performance barely changed as parameters were removed one by one. Only when nitrates were excluded, leaving phosphate alone, did accuracy drop, with the regression value falling to roughly 0.91. The conclusion is remarkable: phosphates and nitrates together are sufficient to predict chlorophyll-a with near-perfect accuracy, and they exert far stronger control than salinity, dissolved oxygen or sulphate.

Perhaps the most practically valuable result came from a parallel set of models built for rapid, on-site assessment. Laboratory analysis of nutrient concentrations takes time and money, so the team explored whether chlorophyll-a could be predicted from parameters that field workers can measure instantly: temperature, dissolved oxygen, pH and salinity. Six scenarios were tested, and the winner was the simplest of all. A model using only pH and salinity predicted chlorophyll-a with an accuracy of 88.7 percent, outperforming every combination that included temperature or dissolved oxygen. Because salinity and pH can be read with handheld instruments in seconds, monitoring agencies can now make a rapid estimate of algal biomass at any station and concentrate their sampling and laboratory effort only where predicted chlorophyll-a levels warrant closer scrutiny. The authors acknowledge limitations, including gaps in monthly coverage and the absence of long-term data, and they call for higher-resolution datasets collected over longer periods and broader areas, along with work to pinpoint the specific sources of chlorophyll-a in the estuary. Still, the message is clear: in a Ramsar-listed estuary under mounting pressure, two cheap field measurements and a trained neural network may be enough to keep watch over one of India’s most fragile coastal ecosystems.

Subject of Research: Machine-learning prediction of chlorophyll-a and eutrophication drivers in the Ashtamudi Estuary, India

Article Title: Analyzing the factors influencing chlorophyll-a in the Ashtamudi estuary on the south-west coast of India

Article References: Priya, K. L., Raj, M. R., Nahas, H., Krishnapriya, K., Remesh, K., Renjith, K. R., Haddout, S., & Casila, J. C. (2026). Analyzing the factors influencing chlorophyll-a in the Ashtamudi estuary on the south-west coast of India. Estuarine Management and Technologies, 3, 169-185. https://doi.org/10.3897/emt.3.183902

Image Credits: AI Generated

DOI: 10.3897/emt.3.183902

Keywords: chlorophyll-a, eutrophication, Ashtamudi Estuary, artificial neural networks, phosphate, nitrate, trophic state index, water quality monitoring, Ramsar wetland, Kerala, salinity, algal blooms