See what's in the water beneath your land.
AquaPredict is a groundwater quality prediction service. Data-driven neural network models forecast key quality indicators from routine water tests — helping farmers and water resources departments decide how groundwater can safely be used.
The world's hidden water supply.
Groundwater sits out of sight beneath our feet — yet billions of people and much of the world's farmland depend on it. Its quality decides whether it can be drunk, used on crops, or needs treatment first.
of all liquid freshwater on Earth is groundwater.
of the water withdrawn for domestic use worldwide comes from groundwater.
of all water used for irrigation is supplied by groundwater.
Source: UN World Water Development Report 2022, "Groundwater: Making the invisible visible"
Developed for a region that lives on groundwater.
The prediction models behind AquaPredict were developed and tested on groundwater from the Alnekheeb basin in Iraq — an area that relies heavily on groundwater, where knowing its quality is essential.
Testing water in a lab for every indicator takes time and money. Prediction models turn the measurements you already have into the quality indicators you need.
Alnekheeb basin, Iraq
A groundwater-dependent region used to assess and model groundwater quality.
Three indicators that decide how water can be used.
Hardness
The amount of dissolved calcium and magnesium. Hard water leaves scale in pipes, pumps and irrigation equipment and affects taste.
MATTERS FOR: drinking, equipment, pipesSodium Adsorption Ratio
Compares sodium with calcium and magnesium. High SAR water can damage soil structure, making it harder for water to soak in.
MATTERS FOR: irrigation, soil healthSalinity
The total level of dissolved salts. Salty water stresses crops, lowers yields and builds up salt in the soil over time.
MATTERS FOR: crops, drinking, livestockEight standard water-quality measurements.
The models learn from common chemical parameters measured in groundwater samples — the ions that determine its hardness, sodium hazard and salinity.
Two neural networks, put head to head.
The research compared two artificial intelligence models, assessing them with statistical indicators and graphical analysis.
RBF-NN
Radial Basis Function Neural Network
A neural network that maps inputs through radial "distance-based" functions, well suited to smooth, non-linear relationships.
PNN
Probabilistic Neural Network
A neural network built on probability theory. In the research, the PNN outperformed the RBF-NN in predicting groundwater quality — and it powers AquaPredict.
The research also studied how the choice of input parameters affects accuracy, finding that adding variables can sometimes improve predictions. AquaPredict uses this to pick the best combination of measurements for each indicator.
Source: Allawi et al., Engineering Applications of Computational Fluid Mechanics 18(1), 2364749 (2024). doi:10.1080/19942060.2024.2364749
From a water sample to a clear decision.
- 01
Share your data
Provide routine water-test results from a well or groundwater source.
- 02
Model predicts
The neural network predicts hardness, sodium adsorption ratio and salinity.
- 03
Get your report
A plain-language summary of groundwater quality for each source.
- 04
Decide with confidence
Plan irrigation, treatment or water allocation based on the results.
For the people who manage the water.
Farmers
Know whether well water suits your crops and soil before you irrigate.
- Avoid damaging soil with high-sodium water
- Match crops to the water's salinity
- Protect pumps and drip lines from scale
Water resources departments
Monitor groundwater quality across a region faster and at lower cost.
- Screen many wells with existing test data
- Spot areas where quality is declining
- Support planning and water-allocation decisions
Civil engineers working on water resources.
Faculty of Engineering
AquaPredict was founded in 2024 by a civil engineering research team specializing in water resources, combining hydrogeology and water-quality knowledge with data-driven modelling.
Frequently asked questions
What does AquaPredict predict?
Three groundwater quality indicators: hardness, sodium adsorption ratio (SAR) and salinity.
What data does it need?
Standard water-quality measurements: sulphate (SO4), chloride (Cl), nitrate (NO3), calcium (Ca), magnesium (Mg), sodium (Na), bicarbonate (HCO3) and carbonate (CO3).
Which models are used?
Two neural networks were compared: a Radial Basis Function Neural Network (RBF-NN) and a Probabilistic Neural Network (PNN). The PNN performed better.
Where were the models tested?
On groundwater from the Alnekheeb basin in Iraq, a region that depends heavily on groundwater.
Why does SAR matter to farmers?
Water with a high sodium adsorption ratio can damage soil structure, reducing how well water soaks into the soil and harming crop growth over time.
Built on open, peer-reviewed research.
Read the open-access study behind AquaPredict's groundwater quality prediction models.
Allawi, M. F., Al-Ani, Y., Jalal, A. D., Ismael, Z. M., Sherif, M. M., & El-Shafie, A. (2024). Groundwater quality parameters prediction based on data-driven models. Engineering Applications of Computational Fluid Mechanics, 18(1), 2364749.
https://doi.org/10.1080/19942060.2024.2364749