The precision of satellite-based irrigation quantification in the Indus and Ganges basins
Abstract. Even though irrigation is the largest direct anthropogenic interference with the terrestrial water cycle, limited knowledge on the amount of water applied for irrigation exist. Quantification of irrigation via evapotranspiration (ET) or soil moisture residuals between remote sensing models and hydrological models, with the latter acting as baselines of natural conditions without the influence of irrigation, have successfully been applied in various regions. Here, we implement an novel ensemble methodology to estimate the precision of ET-based net irrigation quantification by combining different ET and precipitation products in the Indus and Ganges basins. A multi-model calibration of 15 models independently calibrated to simulate natural rainfed ET was conducted prior to the irrigation quantification. Based on the ensemble average, the 2003–2013 net irrigation amounts to 246 mm/year (78 km3/year) and 115 mm/year (76 km3/year) in Indus and Ganges basin, respectively. Net irrigation in Indus basin is evenly split between dry and wet period, whereas 73 % of net irrigation occurs during the dry period in Ganges basin. We found that although annual ET from remote sensing models varied by 91.5 mm/year, net irrigation precision was within 25 mm/season during the dry period, which emphasizes the robustness the applied multi-model calibration approach. Net irrigation variance was found to decrease as ET uncertainty decreased, which related to the climatic conditions, i.e. high uncertainty under arid conditions. A variance decomposition analysis showed that ET uncertainty accounted for 81 % of the overall net irrigation variance and that the influence of precipitation uncertainty was seasonally dependent, i.e. with an increase during the monsoon season. The results underline the robustness of the framework to support large scale sustainable water resource management of irrigated land.
Søren Julsgaard Kragh et al.
Søren Julsgaard Kragh et al.
Søren Julsgaard Kragh et al.
Viewed (geographical distribution)
The manuscript provides a novel and useful evaluation of a technique to estimate net irrigation over a global irrigation hotspot (the Indus and Ganges basins). Overall, the results are impressive, with quantification of net irrigation from an ensemble of unique realization having strong agreement in most cases. I recommend publishing the manuscript after addressing the below comments and expect this will be a widely used study by the remote sensing, hydrology, and agricultural science communities.
Is the transferability of calibrated parameters evaluated in space and time (i.e., to non-calibrated areas and times?). It seems the accuracy of the methodology relies on these calibrated parameters being transferable from non-irrigated places or times to irrigated areas. It would be valuable to estimate irrigation from non-irrigated areas following the same method used to estimate irrigation over the irrigated areas to provide an estimation of the method’s potential systematic biases. Similarly, it would be valuable to show an ET bias time series for the calibrated model over non-irrigated times and irrigated times, respectively (which could be a supplemental figure). It would be helpful to include more detail about the calibration procedure as well. Namely, for irrigated pixels, is the model calibrated during non-irrigated periods or are calibrations from rainfed areas transferred to irrigated areas? In either case, there needs to be an evaluation the parameter transferability.
Please include how data and code can be accessed.
Please provide a thorough grammatical edit of the manuscript before re-submission.
Line 12: “an novel” should be “a novel”
Line 45: “led” should be “lead”
Line 46: Should this be Koch et al. (2020) instead of 2000?
Line 64: I believe 1960 should be “1960s”.
Lines 101-114: What are the sources of uncertainty in the calculations of PET & ET from this model?
Line 101: Please include some more information on the ET calculation (e.g., an equation or equations to show the ET calculation). Is there a soil water module involved in the hydrologic model?
Line 110: Please explain the choice of why FAO-56 PM was used to compute PET.
Line 119: in the supplementary information, can you please provide the list of parameters that were calibrated?
Eq. (3): This definition of net irrigation assumes that 100% of applied irrigation is returned to the atmosphere via ET within the month it is applied. In some instances, this may not be the case. I suggest changing the title to “The precision of satellite-based net irrigation quantification in the Indus and Ganges basins”
Fig. 2b & 2d: please constrain the axes limits to the temporal domain of the study, which looks to be constrained by the ET data.
Fig. 4: Please clean up this figure. Some labels on the left are cut off and the dots are awkwardly overlapping. The color scheme could also benefit from a change.
Line 265: To support that the baseline models can accurately simulate reference ET products, please include a time series comparison over the calibration period. A low MAE can result from large random biases, rather than accurate estimation of ET, thus the MAE is limited in information.
Figure 5: Why is the month of February chosen here? The spatial maps only show a single snap shot (in Feb.) of an instance that helps illustrate the points made in lines 272-285, but presentation of more data is needed to know that this snapshot is not a special case. Perhaps creating a boxplot for differences in ET (ref-baseline) of irrigated lands (across all periods of analysis) for all ET products would be beneficial here to show these results are generalizable, and maintain this snap shot to help illustrate the point through this single example.
Figure 6: does this ensemble include ERA5-Land. I expect it does not because the prior plot just illustrated that irrigation is not present in this data set. Either way, please clarify this in the manuscript text. (Similar comment also applies to Figure 7). If ERA5-Land is included, shouldn’t the estimated irrigation for that ensemble member be close to 0 resulting in a large reduction of the precision?
Figure 7 caption: misspelling of “charts” and “decomposition”
Line 290: Please note the ratio of the standard deviation to the mean irrigation to give a quantification of how large the ensemble uncertainty is relative to the magnitude of irrigation.
Line 309: Why is precision much lower in arid regions?
Lines 322-325: This statement claims lower precision is expected during the wet period because irrigation is lower during this time of year. However, Figure 7 estimates a peak in irrigation occurs during the wet part of the year in the arid and semi-arid regions. Please reconcile this.
Line 329: “reliably” instead of “reliable”