Conagua and Google LLC signed a technical cooperation agreement on September 28, 2026, giving Mexico’s National Meteorological Service (SMN) a framework for four years of early access to AI weather tools and their evaluation under Mexican conditions. El País reported that Conagua used a recent model version to monitor hurricanes Polo and Rachel. That is evidence of operational use, not by itself a measure of improved forecast accuracy or earlier warnings. The Conagua–SMN announcement

A weather model can add another input; under this agreement, the SMN’s role includes evaluating how the tools perform in Mexican conditions.

What the Google–Conagua agreement establishes

The agreement provides for four years of early access to Google’s AI forecasting tools. The SMN is to evaluate and technically validate them against Mexico’s meteorological, hydrological and geographic conditions. The work also includes building institutional capacity to process weather and climate data through APIs, or software interfaces that let systems exchange information. The government announcement

The agreement frames AI as an additional input to the SMN’s existing monitoring, forecasting and early-warning processes. It does not assign the tools the role of issuing official warnings independently. Each party is to use its own resources and retain ownership, control and sovereignty over the information it supplies.

What the reported tracking of Polo and Rachel tells us

El País reported that Conagua used a recent Google model version to monitor the trajectories of hurricanes Polo and Rachel. That account describes an operational use; monitoring a storm’s path is not itself a measured improvement in forecast accuracy or warning lead time.

What WeatherNext 3’s specifications mean

Google describes WeatherNext 3 as producing forecasts hourly. Its stated spatial resolution varies by output: 5 km for specified surface variables such as temperature and moisture, 10 km for other surface variables, and 25 km for atmospheric variables such as wind speed. Google describes WeatherNext 2 as using a 25-km grid with forecasts in six-hour increments. These are Google’s global model specifications, not results from the SMN’s evaluation in Mexico. Google’s WeatherNext 3 announcement