This manuscript addresses a useful but challenging topic. Although there are many remotely sensed global or regional evapotranspiration (ET) datasets, their performances varied across different biomes or regions due to high uncertainty exist in ET estimates. The current manuscript provides a good try to retrieve a land ET product in 2001-2020 using a three-temperature model without resistance and parameter calibration, which is different from the available ET products generated by methods including Penman–Monteith equation-based and surface-energy-balance-residual-based methods. The validation performed at different scales sound good. The intent of the manuscript is worthy and significant, and the topic generally fits the scope of the Earth System Science Data. The manuscript is well-written, and the methods, results, and discussion are clearly presented. Seeing the potential of this, I am in general supportive of publication after minor revision.
Specific comments.
The numbers in front of the comments indicate page and line number.
It may be better to clearly state the temporal resolution and duration of the product.
Please add the data points used for validation.
“the energy balance product” is better changed to “the ET product”.
PMLv2 → PML. Missing a PT-based ET product, GLASS.
Fluxcom also has no value in some arid regions.
It is better to provide a journal article as the thesis may not be available.
“G equals to 0.315Rn” may be misleading.
The subscript l is not consist with those in equations 10 and 11.
Order of the section title was wrong as well as the following section. The authors should proof read the manuscript to avoid such mistakes.
TWSC is better replaced with ΔS. In L171, it is better to use annual TWSC.
Typos (the unit). The authors should proof read the manuscript to avoid such mistakes.
The value 133 mm/yr was from what data?
It seems the ET should be removed.
PMLv2 looks curious. “v2” may be the version number, suggesting delete it across the entire manuscript but remain a statement somewhere.
It is better to clearly state that the EC datasets are the same as those used in figure 2.
Texts in the figures are too small to read. I suggest the authors enlarge these texts to improve their quality and readability.
What does “PFT” mean? Please consider define such abbreviation.
“the whiskers indicate the extreme values” should be “the whiskers indicate the outlier values”.
We thank the reviewer for the positive evaluation and constructive suggestions and comments. We address all comments carefully point-by-point, and corresponding contexts will be incorporated into the revised manuscript. With the help of your constructive comments and suggestions, we believe that our manuscript will be improved substantially. Please see the attached file for our answers to every question/comment.
We thank the reviewers for the positive evaluation. We address all comments carefully point-by-point. With the help of your constructive comments and suggestions, we believe that our manuscript can be improved substantially. Please see the attached file for our answers to every question/comment.
We thank the reviewer for the positive evaluation. We address all comments carefully point-by-point. With the help of your constructive comments and suggestions, we believe that our manuscript will be improved substantially. Please see the attached file for our answers to every question/comment.
We thank the reviewers for the positive evaluation. We address all comments carefully point-by-point. With the help of your constructive comments and suggestions, we believe that our manuscript can be improved substantially. Please see the attached file for our answers to every question/comment.
This manuscript describes an actual evapotranspiration dataset with high spatial resolution, no parameter calibration, and good accuracy, which is particularly important in the context of increased extreme climate hazards. However, this paper needs to further verify the accuracy of its product, which will be more appropriate.
Based on the Conclusion part, it seems the authors produced ET with a resolution of 3-hour. However, the current validations against the observed ET and ET from water balance are mostly at the monthly or annual scales. For this reason, it is not clear the accuracy of the 3-hour ET data. Therefore, it is suggested that the new product have validations against the observed ET at the 3-hour scale.
In addition, the authors claim that the new ET product has a good accuracy compared with other ET products. Yet, such a conclusion is also mostly based on the validations at the monthly or annual scales. It is appropriate to compare with other ET products at the 3-hour scale, such as GLDAS and ERA5.
Therefore, it will be more rigorous to add content for validations against observation data of 3-hour and comparison with other ET products at the 3-hour scale.
Thank you for your interest in our work. In fact, validation of the 3T model was performed not only at a monthly scale but also at a daily scale in our previous manuscript. In particular, the discussion that the 3T model-based ET product could accurately capture the low ET values under extreme conditions in section 4.2 used daily ET estimates.
Your comments are similar to those of the second reviewer. We further tested the performance of the 3T model at the daily scale with all EC observations (because the results in Section 4.2 only contain extreme conditions) as well as at the 3-hour temporal scale. The results indicate that the 3T model is robust at different temporal scales. Please see our reply to reviewer 2 for details.
We thank the reviewer for the positive evaluation. We address all comments carefully point-by-point. With the help of your constructive comments and suggestions, we believe that our manuscript will be improved substantially. Please see the attached file for our answers to every question/comment.
We thank the reviewers for the positive evaluation. We address all comments carefully point-by-point. With the help of your constructive comments and suggestions, we believe that our manuscript can be improved substantially. Please see the attached file for our answers to every question/comment.
This manuscript uses the classic 3T model to generate global terrestrial evapotranspiration (ET) product at 0.25 degree resolution and 3hour temporal resolution. I am glad to see the 3T model has been expended to a global scale. The authors also tested their estimates against flux observations at a monthly scale and water balance ET estimates at an annual scale. Overall, the paper is well written, and the key message is get crossed. Having said that, I have some major concerns on the method applied to the global scale, and validation processes. Follows are the key concerns:
The author grouped the climate regimes into 5 groups according to Koppen-geiger climate classification, then assumes the means of reference net radiation values of the soil and vegetation components are similar within the same group of samples. I am afraid that the classification is too coarse, which will result in large uncertainty in estimating the two important parameters. I think the more details should be exhibited and the uncertainty should be quantified.
The validation of the 3T products is a bit weird. The benefit to use the 3-T model is to detect ET variation in a short period of time. Considering that the 3-T products has been run at 3-hour temporal scale, its robustness should be demonstrated at 3-hour or daily scale at least. Currently, the authors focused their validations against flux measurements at a monthly scale, which is too coarse to be acceptable.
Calculation of water balance ET can be improved. The authors used GLDAS forcing data to drive the 3T model. However, the calculation of water balance ET relies on another precipitation product GPCC. I think the authors need to test the consistency between GLDAS precipitation and GPCC precipitation. In addition, I encourage the authors not only test its performance at a mean annual scale, but also need to test its interannual variability. This will demonstrate its full strength.
In conclusion, the authors had a good attempt to generate global ET product based on the 3T model. However, its robustness has not been fully demonstrated. There are numerous ET products available across the globe. The readers will wonder why this one should be deserved to be published in ESSD. To demonstrate its strength, I suggest the authors put more efforts for validations in arid semi-arid regions and in short period of time. They can show some particular case under extreme drought conditions, i.e. comparing its performance with others in the extreme climatic conditions.
We thank the reviewer for the positive evaluation and consideration that generating a global ET product is a much-needed idea. Although there are numerous ET products that have been rigorously evaluated, notable disagreement exists among these products, indicating that high uncertainties remain in ET estimates and products. One highlight of our study is that we retrieve a land ET product using a three-temperature model without resistance and parameter calibration, which is different from the available ET products generated by methods including Penman–Monteith equation-based and surface-energy-balance-residual-based methods. Our validation results indicated that the proposed ET product has reasonable accuracy. However, some misunderstandings occurred due to our unclear description; for example, validation of the ET product was verified at the daily scale under extreme conditions. Nonetheless, we further evaluated the ET product at 3-hour and daily scales in this revision as suggested. Please see the following point-by-point replies for details.
Responses to Comments 1:
The Köppen-Geiger climate regimes consist of 5 groups (equatorial, arid, warm temperate, snow, and polar), but they were subdivided into 31 climatic regions according to different conditions of precipitation and temperature (Kottek et al., 2006).
In fact, in our study, we further classified these 31 climatic regions of the Köppen-Geiger climate regimes into more detailed subregions via principal component analysis (PCA) and K-means clustering methods. Specifically, PCA was used to select major variables to describe regional characteristics from meteorological factors (i.e., net radiation, air temperature, humidity, wind speed, precipitation, and air pressure) and land surface conditions (i.e., albedo, land surface temperature, NDVI, soil moisture, and soil temperature). Thereafter, these variables were used to classify the 31 climate regions through the K-means clustering method. Because of meteorological and land surface variations, the subregions varied from 90 to 110 in different months. As such, each classified subregion has an area of approximately 1.35×106 km2 (considering 100 subregions) and may be good enough for the reference assumption of the 3T model at the global scale.
To test the impact of region size on the 3T model-based ET estimation, we performed a comparison under two different classification methods with different subregions and sizes. In the first method, Köppen-Geiger climate regimes with 31 subregions were used, whereas detailed subregions with numbers of 90-110 were adopted as the second method. The daily ET estimates in 2011 were used as examples. In general, the two groups of daily ET estimates showed little difference, with mean values of 47 and 42 W m-2, respectively, and were close to the EC observation, with RMSE values of 32 and 33 W m-2 (see Figures R1a and R1b in the attachment). At the yearly scale, however, the 3T model-based ET estimates from 90-110 subregions were much closer to the water balance ET than those estimates from 31 subregions (see Figures R1c and R1d in the attachment). The results indicate that the smaller the region where the reference parameters were obtained, the more accurate the 3T model, which is consistent with our previous findings (Xiong et al., 2015, 2019). The uncertainty of the 3T model at the global scale caused by region size will be incorporated into the revised manuscript.
Reference:
Kottek, M., Grieser, J., Beck, C., Rudolf, B., and Rubel, F.: World map of the Köppen-Geiger climate classification updated, Meteorol. Z., 15, 259-263, https://doi.org/10.1127/0941-2948/2006/0130, 2006.
Xiong, Y. J., Zhao, S. H., Tian, F., and Qiu, G. Y.: An evapotranspiration product for arid regions based on the three-temperature model and thermal remote sensing, J. Hydrol., 530, 392-404, https://doi.org/10.1016/j.jhydrol.2015.09.050, 2015.
Xiong, Y. J., Zhao, W. L., Wang, P., Paw U, K. T., and Qiu, G. Y.: Simple and applicable method for estimating evapotranspiration and its components in arid regions, J. Geophys. Res.: Atmos., 124, 9963-9982, https://doi.org/10.1029/2019JD030774, 2019.
Responses to Comments 2:
In fact, validation of the 3T model is performed not only at a monthly scale but also at a daily scale. In particular, the discussion that the 3T model-based ET product could accurately capture the low ET values under extreme conditions in section 4.2 used daily ET estimates. We apologize for our unclear description.
We further tested the performance of the 3T model with all daily EC observations (because the results in section 4.2 only contain extreme conditions). The results showed that the 3T model-based ET estimates agreed well with the observations (N=294058), with an RMSE of 33 W m-2 (or 1.1 mm day-1) (see Figure R2a in the attachment), which was also comparable to other ET products, such as GLDAS (RMSE: 32 W m-2 or 1.1 mm day-1) (this study), PML (RMSE: 0.7 mm day-1) (Zhang et al., 2021), and SEBS (RMSE: 1.6 mm day-1) (Chen et al., 2021).
At a 3-hour temporal scale, the data are too large to perform an entire comparison at a global scale. Moreover, as you may know, high temporal data even at the daily scale, especially remote sensing data, may encounter missing values for several reasons, such as clouds and precipitation (the MODIS land surface temperature product is a good example), which complicates the comparison and contains more uncertainty. In fact, the performance of the 3T model over a short period of time (30 minutes to 1 hour) has been tested at the point scale, and the results from both our studies (Qiu and Zhao, 2010; Tian et al., 2014; Qiu et al., 2015; Xiong et al., 2019) and others (Zhou et al., 2014; Zhang et al., 2020) show that the 3T model generally performed well. Nonetheless, we further tested the performance of the 3T model at the 3-hour scale across the world using EC observations in 2011 (6278 data points) (Figure R3a). Data were selected from the 15th day of each month in 2011. Although the RMSE (74 W m-2) was slightly greater than that at the daily scale compared to the EC observations (see Figure R3a in the attachment), the 3T model-based ET estimates at the 3-hour scale agreed well with the GLDAS ET, with an r of 0.89 and RMSE of 21 W m-2 (see Figure R3b in the attachment).
These results indicate that the 3T model is robust at different temporal scales. These results will be incorporated into the revised manuscript.
Reference:
Chen, X., Su, Z., Ma, Y., Trigo, I., and Gentine, P.: Remote Sensing of Global Daily Evapotranspiration based on a Surface Energy Balance Method and Reanalysis Data, J. Geophys. Res.: Atmos., 126, e2020JD032873, https://doi.org/10.1029/2020JD032873, 2021.
Qiu, G. Y. and Zhao, M.: Remotely monitoring evaporation rate and soil water status using thermal imaging and “three-temperatures model (3T Model)” under field-scale conditions, J. Environ. Monit., 12, 716-723, https://doi.org/10.1039/B919887C, 2010.
Qiu, G. Y., Li, C., and Yan, C.: Characteristics of soil evaporation, plant transpiration and water budget of Nitraria dune in the arid Northwest China, Agric. For. Meteorol., 203, 107-117, https://doi.org/10.1016/j.agrformet.2015.01.006, 2015.
Tian, F., Qiu, G. Y., Lü, Y., Yang, Y., and Xiong, Y.: Use of high-resolution thermal infrared remote sensing and “three-temperature model” for transpiration monitoring in arid inland river catchment, J. Hydrol., 515, 307-315, https://doi.org/10.1016/j.jhydrol.2014.04.056, 2014.
Xiong, Y. J., Zhao, W. L., Wang, P., Paw U, K. T., and Qiu, G. Y.: Simple and applicable method for estimating evapotranspiration and its components in arid regions, J. Geophys. Res.: Atmos., 124, 9963-9982, https://doi.org/10.1029/2019JD030774, 2019.
Zhang, Y., Kong, D., Gan, R., Chiew, F. H., McVicar, T. R., Zhang, Q., and Yang, Y.: Coupled estimation of 500 m and 8-day resolution global evapotranspiration and gross primary production in 2002–2017, Remote Sens. Environ., 222, 165-182, https://doi.org/10.1016/j.rse.2018.12.031, 2019.
Zhang, Y., Qin, H., Zhang, J., and Hu, Y.: An in-situ measurement method of evapotranspiration from typical LID facilities based on the three-temperature model, J. Hydrol., 588, 125105, https://doi.org/10.1016/j.jhydrol.2020.125105, 2020.
Zhou, X., Bi, S., Yang, Y., Tian, F., and Ren, D.: Comparison of ET estimations by the three-temperature model, SEBAL model and eddy covariance observations, J. Hydrol., 519, 769-776, https://doi.org/10.1016/j.jhydrol.2014.08.004, 2014.
Responses to Comments 3:
In fact, the water balance ET was independent of the estimates from the 3T model because inputs of the 3T model only consist of net radiation (Rn), air temperature (Ta), and land surface temperature (LST) from GLDAS forcing data. Precipitation is not required in the inputs of the 3T model. Therefore, consistency between GLDAS precipitation and GPCC precipitation is unnecessary.
We agree with the comment that the performance of the 3T model should be tested at the interannual scale to enhance its robustness. We selected 10 EC sites covering different biomes across the world to test the interannual variability of the 3T model-based ET estimates at a daily scale. The variations in the 3T product generally fit the observation (see Figure R2 in the attachment). We also compared multiyear (2003–2013) mean monthly ET values between several ET products, and the results further indicate that the interannual variability of the 3T model was similar to the other ET products (see Figure R4 in the attachment). These results will be incorporated into the revised manuscript.
We thank the reviewers for the positive evaluation. We address all comments carefully point-by-point. With the help of your constructive comments and suggestions, we believe that our manuscript can be improved substantially. Please see the attached file for our answers to every question/comment.