the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Nov 2025
Assessing the suitability of global evapotranspiration products over irrigated areas
Abstract. Reliable estimation of evapotranspiration (ET) over irrigated croplands is crucial for agricultural water management, hydrological modeling, and monitoring of land–atmosphere exchanges. Yet the reliability of global ET datasets in these environments remains insufficiently assessed. Here, we evaluate six widely used global ET products (FLUXCOM RS, GLEAM v4.2a, PMLv2, ERA5-Land, MOD16A2, and SSEBop v6.1), covering a wide range of modeling approaches, to assess their ability to capture irrigation-related ET signals. The assessment combines spatial and seasonal evaluations across diverse agro-climatic regions, using three complementary references: a map of area equipped for irrigation, the OpenET ensemble, and eddy covariance measurements from irrigated croplands. Results reveal strong contrasts in how well the products reproduce reference patterns. PMLv2, SSEBop v6.1, and FLUXCOM RS show the highest agreement, effectively capturing irrigation-related spatial and seasonal ET variations. MOD16A2 shows moderate agreement, with consistently lower ET values than the reference datasets. ERA5-Land and GLEAM v4.2a exhibit the weakest correspondence, reflecting limitations linked to their precipitation-driven water-balance soil moisture and stress formulations. Differences among products mainly reflect how water stress is represented and whether irrigation-sensitive variables such as land surface temperature and vegetation properties are incorporated. This multi-scale evaluation provides guidance for selecting ET products in irrigated regions and highlights priorities for improving the representation of irrigation in global ET models.
- Preprint
(18093 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 10 Jan 2026)
- RC1: 'Comment on egusphere-2025-5716', Anonymous Referee #1, 23 Dec 2025 reply
-
RC2: 'Comment on egusphere-2025-5716', Anonymous Referee #2, 24 Dec 2025
reply
This study provides a very systematic and comprehensive evaluation of currently available global ET products over irrigated areas. The analyses are thorough and well structured, and the manuscript clearly demonstrates how different datasets perform in capturing irrigation-induced ET signals across multiple spatial and temporal scales. Overall, I find this to be an excellent and highly valuable contribution to the community. I believe this manuscript generally meets the publication standards of HESS, and I only have a few minor comments and suggestions for improvement.
（1）The authors focus on CONUS, Europe, and South Asia. It may be worth noting that East Asia also hosts some of the world’s most intensively irrigated agricultural regions, such as the North China Plain, where numerous studies have documented substantial irrigation-induced increases in ET. If feasible, I would recommend adding some brief discussion or analysis related to these regions to further enhance the completeness of the study.
（2）In addition, irrigation impacts on ET are influenced not only by irrigated area but also by regional climate conditions and irrigation methods (e.g., surface, sprinkler, or drip irrigation). A brief discussion of these factors could help clarify why different ET products may perform differently across various irrigated systems.
Citation: https://doi.org/10.5194/egusphere-2025-5716-RC2 -
CC1: 'Comment on egusphere-2025-5716', Nima Zafarmomen, 25 Dec 2025
reply
This study provides a highly relevant and timely evaluation of global ET datasets. While many intercomparison studies exist for natural ecosystems, the systematic assessment of how these models handle (or fail to handle) the artificial water inputs of irrigation is relatively rare. The novelty of this work lies in its multi-scale, multi-source validation framework. By integrating the Global Map of Irrigated Areas (spatial extent), the OpenET ensemble (regional seasonal dynamics), and Eddy Covariance data (local-scale magnitude), the authors provide a robust "stress test" for these products. The finding that LST-driven models (SSEBop) and biome-calibrated machine learning models (FLUXCOM, PMLv2) significantly outperform precipitation-driven water balance models (GLEAM, ERA5-Land) offers a clear roadmap for the next generation of global ET modeling.
Minor Comments
1. he authors resampled all products to a 0.1° grid (approx. 10 km) to match GLEAM’s resolution. While necessary for a direct comparison, this might disadvantage high-resolution products like SSEBop (1 km) or MOD16A2 (500 m) in regions with fragmented irrigation patterns. It would be beneficial to add a brief sentence in the Discussion acknowledging how this "coarsening" might smooth out the irrigation signal that finer-scale models are designed to capture.
2. In Section 4.2.1, the authors correctly note that "equipped" does not always mean "actually irrigated." Given this, did the authors consider using a "Monthly Irrigated Fraction" or a dynamic cropland product to see if the correlations improve during high-demand years vs. low-demand years?
3. For the Po Valley validation, the authors mention that the grid cells contain ~34% and ~46% irrigated maize. While they addressed the footprint mismatch, a comment on whether the non-irrigated portions of those pixels were "natural vegetation" or "rainfed crops" would help clarify the magnitude of the negative bias observed in models like MOD16A2.
4. Figure 7 is excellent for visualizing the 2014 California drought. However, it would be interesting to know if the "low sensitivity" of GLEAM and ERA5-Land to the 2014 restrictions is purely due to the lack of an irrigation module, or if the atmospheric forcing (VPD) was also insufficient to trigger the expected ET drop in their specific formulations.
5. I strongly recommend that the authors consider and discuss recent studies that explicitly integrate remote sensing information into irrigation-aware hydrological and land-surface modeling frameworks. In particular, the study “Assimilation of Sentinel-based Leaf Area Index for modeling surface–groundwater interactions in irrigation districts” provides relevant insights into how satellite-derived vegetation dynamics can improve the representation of irrigation effects on evapotranspiration and soil–water processes. Incorporating discussion of such work would help place the present evaluation in a broader modeling context and highlight pathways for improving global ET products in irrigated regions.
Citation: https://doi.org/10.5194/egusphere-2025-5716-CC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 236 | 152 | 19 | 407 | 22 | 22 |
- HTML: 236
- PDF: 152
- XML: 19
- Total: 407
- BibTeX: 22
- EndNote: 22
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
First of all, it is a great honor to learn from the authors' valuable research. I believe this is a relatively innovative article with certain significance for optimizing data products in agricultural irrigation areas. The article has a clear structure and detailed research content, but I have a few suggestions regarding its content.
Firstly, in the discussion, it is mentioned that models primarily driven by precipitation-influenced soil water balance mechanisms tend to underestimate soil moisture in irrigated farmlands. However, the authors do not explicitly provide the variation in precipitation across different years and months for the irrigation areas studied, only presenting the characteristic values of precipitation during the peak season in Table 2.
Secondly, Figure 7 shows the variation in ET during the irrigation season across different years. In addition to the changes in NDVI, could the variation in precipitation also be provided? I believe this would make the results clearer.
Thirdly, in the discussion, the authors focus more on the mechanisms of the models when explaining the reasons for product differences. While these details are thorough, they seem to overlook the potential impact of input data discrepancies on the results. Therefore, it is necessary to further clarify whether differences in input data could lead to variations in the results.