Testing daily 10 m WaPOR data for improved irrigation in Jendouba, Tunisia

The Food and Agriculture Organization of the United Nations (FAO), through its Water Productivity through Open-access of Remotely sensed derived data (WaPOR) project, supports countries in monitoring and improving agricultural water productivity using open-access remote sensing data. WaPOR provides consistent, spatially explicit information on key water and crop variables, including actual evapotranspiration, to inform water management, irrigation planning and policy analysis.

Within this framework, FAO initiated a pilot in Jendouba, Tunisia, to test whether improving the spatial resolution, temporal resolution and data latency of actual evapotranspiration (ETa) data would meaningfully improve irrigation advisory services. The pilot focused on the use of the Irrigation Recommendation using Earth observation data (IREY) application, co-developed by the International Water Management Institute (IWMI) and the National Field Crops Institute of Tunisia (Institut National des Grandes Cultures, INGC) in the framework of the WaPOR project.

The data at higher spatial and temporal resolution was made available in the WaPOR portal for the Jendouba area, already a WaPOR level 3 area (20 m resolution), for the period January to December 2025. Spatial resolution was improved from 20 m to 10 m, temporal resolution from decadal to daily, and data latency, from ~15 days to ~2 days. 

What was done

IWMI ran a root-zone soil moisture and water-balance simulation for the winter–spring 2025 season (January–April 2025), driving the IREY model with both the new daily 10 m ETa product and the existing decadal 20 m ETa product. 

To ensure comparability, irrigation quantities and dates were kept identical between the two runs. A four-day GFS-driven forecast issued on 25 April was used to test short-term advisory performance. Adapting the IREY model to ingest daily ETa data required substantial code changes, which were implemented by IWMI.

Key findings

The test was designed around three objectives: (1) assess whether higher spatial resolution improves parcel-level differentiation, (2) evaluate whether reduced latency enables more timely irrigation recommendations, and (3) examine the combined effect on the quality of the advisory output. The 10 m daily product showed clear improvements over the baseline.

The higher-frequency data flagged an irrigation need approximately one week earlier in late March, improving the timeliness of advisory alerts. When coupled with the short-term forecast, it also helped avoid a false alarm: the 10 m-driven forecast correctly indicated the crop would remain within the readily available water threshold over 25–29 April, whereas the 20 m-driven forecast was less reliable over the same period. 

On accuracy, the new product reduced a ~10% overestimation of ETa that had been observed in the 20 m product during the previous growing season (2023–2024). Over the season, cumulative ETa from the 10 m product was roughly 20 mm (approximately 10 percent) lower— a difference equivalent to about one additional pivot irrigation run.

The daily ETa values from the 10 m and 20 m products showed a correlation of 95 percent. However, the 10 m product yielded a more accurate crop coefficient, keeping actual ET estimates below reference ET where the 20 m product had exceeded it. 

Deep percolation totals were similar across both experiments and were driven primarily by heavy rainfall events rather than differences in ETa.

Limitations

A full seasonal comparison based on multiple forecasts  (approximately 150 forecasts) was not feasible due to the limitations related to the Global Forecast System (GFS) operated by the National Centers for Environmental Predictions (NCEP). Operational forecast data is only retained for a short period and the archived data is stored in a different format than the live feed. As a result, retrospective analysis would require significant model rework. 

Field-level validation of spatial improvements, such as parcel boundary detection, was not carried out due to the absence of suitable ground-truth parcel data for the Jendouba area.

Conclusion and recommendation

Regarding objective (1), field-level validation of spatial improvements could not be completed in this pilot due to missing parcel ground-truth data, but the finer 10 m resolution is expected to improve parcel differentiation, particularly relevant for regions with small or fragmented holdings. 

On objective (2), the pilot clearly demonstrates that the daily cadence and reduced latency of the 10 m product enable more timely irrigation alerts, identifying needs roughly a week earlier than the decadal baseline. 

As for objective (3), the overall quality of the advisory output improved: the 10 m product reduced the overestimation in ETa, produced a more accurate crop coefficient, and helped avoid a forecast-induced false alarm.

Based on these results, adoption of the 10 m daily product for operational use in the IREY application could be beneficial. Beyond Tunisia, there is emerging interest in applying the 10 m product in other contexts, including a pre-proposal for a water productivity and irrigation advisory pilot in the United Arab Emirates, which would leverage high-resolution WaPOR data for crop water requirement assessment and water accounting in an arid context. 

To further strengthen the evidence base, future work should assess the product's performance over a full growing season with successive forecasts, and explore options for archiving forecast data to enable more comprehensive retrospective evaluation.