AquaMX

Inspiration

San Luis Potosí — my city — extracts 74% more water than nature can replenish. Aquifer 2411, which supplies nearly one million people, accumulates a deficit of -59.6 Mm³ every single year. Yet authorities make decisions without knowing exactly when or where the system will collapse.

The inspiration came from a personal place: watching a water crisis unfold slowly in my own city while the data to predict it already existed — scattered across government PDFs, never turned into a usable tool.

If the data exists and the computational tools exist, why isn't there a model that tells a decision-maker "your city's aquifer will reach operational crisis in year X"? AquaMX SLP was built to answer that question — first for San Luis Potosí, then for the rest of Mexico.

What we learned

• Wolfram Language is remarkably powerful for combining time series modeling, predictive analytics, and geospatial visualization in a single computational environment.
• CONAGUA's water data is public but scattered across PDFs and hard-to-query portals — there is a massive gap between available data and informed decision-making.
• An ARIMA(1,1,1) model calibrated with real CONAGUA data can generate robust projections with 95% confidence intervals, translating hydrological mathematics into actionable public policy.

How we built it

The pipeline has two layers:

Layer 1 — Python (data auxiliary) Automated download scripts from CONAGUA's SIGAGIS, REPDA, and SINA portals. Cleaning and structuring of historical time series for water table levels, extraction rates, and recharge volumes — starting with Aquifer 2411 (San Luis Potosí).

Layer 2 — Wolfram Language (the protagonist)

• TimeSeriesModelFit with ARIMA(1,1,1) to project water table levels 15 years forward
• Operational crisis year calculation based on calibrated descent rates
• GeoGraphics and GeoRegionValuePlot for interactive municipal risk maps
• Composite risk index per municipality (extraction, recharge, industrial pressure, balance)
• Interactive dashboard using Manipulate, publicly deployed on Wolfram Cloud

The architecture was designed with national scalability in mind from day one: the same model applies to any of Mexico's 653 aquifers by simply swapping the data source. San Luis Potosí is the proof of concept. Mexico is the goal.

Challenges we faced

• Fragmented data: CONAGUA does not publish historical water table series in directly downloadable format. Technical studies are locked in PDFs. We built realistic synthetic series grounded in officially published values (deficit, extraction, recharge).
• Real-time geospatial visualization: integrating GeoGraphics with dynamic dashboard data required optimization so that Manipulate responds fluidly in Wolfram Cloud.
• Translating hydrology into policy language: the hardest challenge wasn't technical — it was communication. Making "-1.77 m/yr water table descent" become "estimated operational crisis in 2036" — a number a mayor can understand and act on. That translation matters at city level today, and at national level tomorrow.

Impact & Scalability

AquaMX currently models Aquifer 2411 in San Luis Potosí, but the framework is built to scale:

Mexico has 105 over-exploited aquifers affecting millions of people. Every one of them deserves a collapse timeline. AquaMX was built to provide it.

Built With

• arima
• beautiful-soup
• clouddeploy
• conagua-sigagis
• geographics
• georegionvalueplot
• manipulate
• pandas
• python
• repda
• timeseriesmodelfit
• wolfram-cloud
• wolfram-technologies