the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Extent of gross underestimation of precipitation in India
Abstract. Underestimation of precipitation (UoP) in the hilly and mountainous parts of South Asia is estimated by some studies to be as large as the observed precipitation (P). For instance, correction factors (CFs) developed by the recent PBCOR dataset have values exceeding 2.0 across the wettest regions of India, some of which have experienced catastrophic flooding in the recent past. However, UoP has been analyzed only to a limited extent across India. Towards bridging this gap, this study analyzes watershed-scale UoP using various P datasets within a water imbalance analysis. Among these P datasets, the often-used Indian Meteorological Department (IMD) dataset is of primary interest.
Gross UoP was identified by analyzing the extent of imbalance in the annual water budget of watersheds corresponding to 242 river gauging stations where quality controlled data on catchment boundaries and streamflow is available. Water year (WY) based volume of observed annual P was compared against observed annual streamflow (R) and satellite-based actual evapotranspiration (ET). Across many watersheds of both Northern and Peninsular India, the spurious water imbalance scenarios of P ≤ R, or P << R + ET, were realized. It is shown that management of water, such as groundwater extraction, reservoir storage and water diversion (imports or exports), is generally minimal compared to annual P in such watersheds. It is also shown that annual changes in terrestrial water storage are also minimal compared to annual P in such watersheds. Assuming data on R (and ET to a lesser extent) to be reliable, it is concluded that UoP is very likely the cause of such imbalance.
All 12 of the P datasets analyzed here suffer from UoP, but the extent of UoP varies by dataset and region. The reanalysis-based datasets ERA5-Land and IMDAA are less affected by UoP than IMD, and the spatial patterns of estimated CFs based on these two datasets are also consistent with those made independently by the PBCOR dataset. Based on the 30-year period of WY 1985–2014, P for the whole of India could be up to 19 % (ERA5-Land) to 37 % (IMDAA) higher than IMD, with substantial variability within years and river basins. For instance, P for the Indian portion of the Ganga River Basin, for the same 30-year period, could be up to 36 % (ERA5-Land) to 54 % (IMDAA) higher than IMD. The actual magnitude of UoP is speculated to be even greater. Moreover, trends in IMD's P are not always present in ERA5-Land and IMDAA. Studies using IMD should exercise caution since UoP could lead to misrepresentation of water budgets and long-term trends.
The empirical approach of identifying watersheds affected by UoP using a water imbalance approach is contingent on data availability. It is speculated that if additional data on R becomes available, particularly in Northern India, many other watersheds affected by UoP would be identified. While the scientific community is striving to continually improve P products, India's water agencies can help the community better quantify UoP by making observed hydrometeorological data more widely available. Limitations of this study are discussed.
- Preprint
(5051 KB) - Metadata XML
-
Supplement
(18688 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2024-18', Anonymous Referee #1, 07 Mar 2024
The manuscript entitled "Extent of gross underestimation of precipitation in India" by Goteti and Famiglietti, analyzes watershed-scale underestimation of precipitation over India using various precipitation datasets within a water imbalance analysis. The authors have designed the study with relevant methodology through the knowledge attained from associated literature. There are some corrections required in the manuscript, although the authors have explained the results and the methodology in detail, there are a few disconnections between sentences and minor grammar corrections required. I have raised a few questions that require clarification, also suggested some necessary modifications and additions to the manuscript.
GENERAL COMMENTS:
I would suggest the author to include the analysis for the summer monsoon season (i.e., June-July-August-September), as the summer monsoon accounts for a major portion of the annual precipitation over India.
OTHER COMMENTS:
1. I would recommend the authors to use ‘precipitation’ instead of ‘P’, particularly in the introduction.
2. Line 34: “Raw data from P gauges” should be revised to “Raw data from rain gauges.”
3. Line 43: “However, gauge-based gridded datasets can be far from ideal” Do you have any evidence to support this statement for IMD data (concerned about word ‘far’)?"
4. Line 44: This paper, King et al. (2013), did not utilize the IMD data
5. Line 46: What do you mean by 'other errors'? demonstrate them?"
6. Fig. 1: I would suggest plotting the rainfall estimates from the actual rain gauge stations along with the IMD gridded precipitation data. It could be possible (potentially) that the interpolation method used in constructing the IMD data introduces some biases.
7. Fig. 1(b): I am concerned about the PBCOR data that authors have used to show the ratio of bias correction. The number of rain gauge stations used in this data (Fig. 2; Beck et al. 2020), particularly over India, is far fewer than the IMD rain gauge stations (Fig. 1; Prakash et al. 2015). So, it is possible that the observed largest ratios could be attributed to PBCOR datasets rather than IMD. I would suggest the authors to compare the PBCOR data with the IMD for both mean and extreme (e.g.; 99th percentile) cases before using it for bias correction ratio calculations.
Prakash, S., Mitra, A. K., Momin, I. M., Pai, D. S., Rajagopal, E. N., & Basu, S. (2015). Comparison of TMPA-3B42 versions 6 and 7 precipitation products with gauge-based data over India for the southwest monsoon period. Journal of Hydrometeorology, 16(1), 346-362.
8. Line 61: “If estimates from PBCOR are reasonable”, follow comment 8
9. Lines 93-95: “using procedures… watersheds of Northern India”, How did you compile? Please explain in detail.
10. Lines 97-100: Include discussion about uncertainties associated with the streamflow data.
Citation: https://doi.org/10.5194/hess-2024-18-RC1 - AC1: 'Reply on RC1', Gopi Goteti, 01 May 2024
-
RC2: 'Comment on hess-2024-18', Anonymous Referee #2, 03 Apr 2024
Comments to the Editor
Dear Editor,
Thank you for considering me as a reviewer of the manuscript. I found the manuscript to be interesting and relevant to the existing literature on precipitation underestimation. The authors have used the watershed water balance method to analyse precipitation with some interesting insights into hydroclimatology. The major issues with the manuscript are:
- The study fails to account for the fundamental assumptions of the catchment water balance method. The primary assumptions of catchment/watershed water balance are of a closed system and over lapping physical and hydrological boundaries, i.e. the watershed boundaries overlap with their respective aquifer boundaries. Both rarely occur in nature. Most aquifers are leaky, and multiple watersheds may fall in a single aquifer, or adjacent watersheds may share a common aquifer. Thus catchments with leaky aquifers (https://wires.onlinelibrary.wiley.com/doi/abs/10.1002/wat2.1386) or non-overlapping aquifer-watershed boundaries with significant inter-basin groundwater transfer (https://www.sciencedirect.com/science/article/abs/pii/S0022169420300433), can have significant water imbalance, apart from the limitations mentioned in Section 5 of the manuscript. It may also explain why most off-balance catchments were seen in mountains, as they have significantly high geological connectivity.
- The assumption that mountain and/or forested areas have minimal watershed management is not true. Many of the mountain regions and/or forested areas are included in watershed management through extensive government and civil society efforts. The activities include soil and water conservation efforts, which can change local and landscape-level hydrology. I would recommend the authors to have a relook at the assumption and its impact on the results. https://www.cabidigitallibrary.org/doi/full/10.5555/20123144409
- The majority of the watersheds are in central and southern India, which reduces the studies’ representation.
- I was excited to see a study on precipitation underestimation/overestimation. Few studies have approached doing hydrology backwards (https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2008WR006912), and rarely any from India. However, the study loses a chance to dwell on the processes behind UoP, which could have been explored by going deeper in a selected watershed like Nethravati. The study shows an overall UoP scenario in central and southern India but doesn’t inform us enough on why it might be happening, apart from data issues.
I recommend the manuscript for a major revision to address the abovementioned and the detailed comments given to the authors.
Detailed comments
L100 Could you please add a table with salient features of the 242 watersheds. Size distribution, Elevation bands, rainfall characteristics, % forests, %ET, etc.
There seems to be a lack of watersheds from Himalaya, Western Ghats, Western and Central India. Can the study claim to represent India in the title when most of the watersheds are in central and southern India?
L118-119: What is the strength of interpolation between APHRODITE and IMD. Would recommend adding any results to supplementary table.
L149-150: The CGWB groundwater dataset is known to be quite limited by the number of wells and their representativeness. How does that effect the manuscript?
L162-165: Are there any studies on the validation of TWS from GRACE against ground data? Please do add a section on the limitations of GRACE, especially in complex mountains or geologically diverse systems.
L175 There seems to be an objective-method mismatch. The methods section describes the theoretical and per-watershed aspects well, but a flowchart mimicking the methods followed to achieve the three objectives would help us understand the study.
L175-180 What about leaky aquifers and inter-connected basins with significant groundwater transfer?
L200-205 Could you please define the exports and imports as per CGWB 2019 and how they are distinct from ∆GW natrual, ∆GW human and ∆Reservoir, with examples if possible?
L210-211 The assumption is usually appropriate in catchments with high infiltration and transmissivity, eg Karst aquifers. Do any of the watersheds have such physical conditions?
L240 – 245 This is a big assumption to make. Many of the mountain regions and/or forested areas are included in watershed management through extensive government and civil society efforts. The activities include soil and water conservation efforts which can change local and landscape level hydrology. Please have a relook. https://www.cabidigitallibrary.org/doi/full/10.5555/20123144409
L246 Is off-balance the same as imbalance here? Please maintain consistency if they are the same or clarify if not.
L258-260 Please provide details of the spatial syncing methods and any sensitivity analysis to check if there was any loss of information.
More specific comments are given as pop-up notes in the manuscript.
Best wishes
- AC2: 'Reply on RC2', Gopi Goteti, 01 May 2024
Status: closed
-
RC1: 'Comment on hess-2024-18', Anonymous Referee #1, 07 Mar 2024
The manuscript entitled "Extent of gross underestimation of precipitation in India" by Goteti and Famiglietti, analyzes watershed-scale underestimation of precipitation over India using various precipitation datasets within a water imbalance analysis. The authors have designed the study with relevant methodology through the knowledge attained from associated literature. There are some corrections required in the manuscript, although the authors have explained the results and the methodology in detail, there are a few disconnections between sentences and minor grammar corrections required. I have raised a few questions that require clarification, also suggested some necessary modifications and additions to the manuscript.
GENERAL COMMENTS:
I would suggest the author to include the analysis for the summer monsoon season (i.e., June-July-August-September), as the summer monsoon accounts for a major portion of the annual precipitation over India.
OTHER COMMENTS:
1. I would recommend the authors to use ‘precipitation’ instead of ‘P’, particularly in the introduction.
2. Line 34: “Raw data from P gauges” should be revised to “Raw data from rain gauges.”
3. Line 43: “However, gauge-based gridded datasets can be far from ideal” Do you have any evidence to support this statement for IMD data (concerned about word ‘far’)?"
4. Line 44: This paper, King et al. (2013), did not utilize the IMD data
5. Line 46: What do you mean by 'other errors'? demonstrate them?"
6. Fig. 1: I would suggest plotting the rainfall estimates from the actual rain gauge stations along with the IMD gridded precipitation data. It could be possible (potentially) that the interpolation method used in constructing the IMD data introduces some biases.
7. Fig. 1(b): I am concerned about the PBCOR data that authors have used to show the ratio of bias correction. The number of rain gauge stations used in this data (Fig. 2; Beck et al. 2020), particularly over India, is far fewer than the IMD rain gauge stations (Fig. 1; Prakash et al. 2015). So, it is possible that the observed largest ratios could be attributed to PBCOR datasets rather than IMD. I would suggest the authors to compare the PBCOR data with the IMD for both mean and extreme (e.g.; 99th percentile) cases before using it for bias correction ratio calculations.
Prakash, S., Mitra, A. K., Momin, I. M., Pai, D. S., Rajagopal, E. N., & Basu, S. (2015). Comparison of TMPA-3B42 versions 6 and 7 precipitation products with gauge-based data over India for the southwest monsoon period. Journal of Hydrometeorology, 16(1), 346-362.
8. Line 61: “If estimates from PBCOR are reasonable”, follow comment 8
9. Lines 93-95: “using procedures… watersheds of Northern India”, How did you compile? Please explain in detail.
10. Lines 97-100: Include discussion about uncertainties associated with the streamflow data.
Citation: https://doi.org/10.5194/hess-2024-18-RC1 - AC1: 'Reply on RC1', Gopi Goteti, 01 May 2024
-
RC2: 'Comment on hess-2024-18', Anonymous Referee #2, 03 Apr 2024
Comments to the Editor
Dear Editor,
Thank you for considering me as a reviewer of the manuscript. I found the manuscript to be interesting and relevant to the existing literature on precipitation underestimation. The authors have used the watershed water balance method to analyse precipitation with some interesting insights into hydroclimatology. The major issues with the manuscript are:
- The study fails to account for the fundamental assumptions of the catchment water balance method. The primary assumptions of catchment/watershed water balance are of a closed system and over lapping physical and hydrological boundaries, i.e. the watershed boundaries overlap with their respective aquifer boundaries. Both rarely occur in nature. Most aquifers are leaky, and multiple watersheds may fall in a single aquifer, or adjacent watersheds may share a common aquifer. Thus catchments with leaky aquifers (https://wires.onlinelibrary.wiley.com/doi/abs/10.1002/wat2.1386) or non-overlapping aquifer-watershed boundaries with significant inter-basin groundwater transfer (https://www.sciencedirect.com/science/article/abs/pii/S0022169420300433), can have significant water imbalance, apart from the limitations mentioned in Section 5 of the manuscript. It may also explain why most off-balance catchments were seen in mountains, as they have significantly high geological connectivity.
- The assumption that mountain and/or forested areas have minimal watershed management is not true. Many of the mountain regions and/or forested areas are included in watershed management through extensive government and civil society efforts. The activities include soil and water conservation efforts, which can change local and landscape-level hydrology. I would recommend the authors to have a relook at the assumption and its impact on the results. https://www.cabidigitallibrary.org/doi/full/10.5555/20123144409
- The majority of the watersheds are in central and southern India, which reduces the studies’ representation.
- I was excited to see a study on precipitation underestimation/overestimation. Few studies have approached doing hydrology backwards (https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2008WR006912), and rarely any from India. However, the study loses a chance to dwell on the processes behind UoP, which could have been explored by going deeper in a selected watershed like Nethravati. The study shows an overall UoP scenario in central and southern India but doesn’t inform us enough on why it might be happening, apart from data issues.
I recommend the manuscript for a major revision to address the abovementioned and the detailed comments given to the authors.
Detailed comments
L100 Could you please add a table with salient features of the 242 watersheds. Size distribution, Elevation bands, rainfall characteristics, % forests, %ET, etc.
There seems to be a lack of watersheds from Himalaya, Western Ghats, Western and Central India. Can the study claim to represent India in the title when most of the watersheds are in central and southern India?
L118-119: What is the strength of interpolation between APHRODITE and IMD. Would recommend adding any results to supplementary table.
L149-150: The CGWB groundwater dataset is known to be quite limited by the number of wells and their representativeness. How does that effect the manuscript?
L162-165: Are there any studies on the validation of TWS from GRACE against ground data? Please do add a section on the limitations of GRACE, especially in complex mountains or geologically diverse systems.
L175 There seems to be an objective-method mismatch. The methods section describes the theoretical and per-watershed aspects well, but a flowchart mimicking the methods followed to achieve the three objectives would help us understand the study.
L175-180 What about leaky aquifers and inter-connected basins with significant groundwater transfer?
L200-205 Could you please define the exports and imports as per CGWB 2019 and how they are distinct from ∆GW natrual, ∆GW human and ∆Reservoir, with examples if possible?
L210-211 The assumption is usually appropriate in catchments with high infiltration and transmissivity, eg Karst aquifers. Do any of the watersheds have such physical conditions?
L240 – 245 This is a big assumption to make. Many of the mountain regions and/or forested areas are included in watershed management through extensive government and civil society efforts. The activities include soil and water conservation efforts which can change local and landscape level hydrology. Please have a relook. https://www.cabidigitallibrary.org/doi/full/10.5555/20123144409
L246 Is off-balance the same as imbalance here? Please maintain consistency if they are the same or clarify if not.
L258-260 Please provide details of the spatial syncing methods and any sensitivity analysis to check if there was any loss of information.
More specific comments are given as pop-up notes in the manuscript.
Best wishes
- AC2: 'Reply on RC2', Gopi Goteti, 01 May 2024
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
446 | 127 | 27 | 600 | 20 | 18 | 17 |
- HTML: 446
- PDF: 127
- XML: 27
- Total: 600
- Supplement: 20
- BibTeX: 18
- EndNote: 17
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1