the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Mar 2023
Analysis of seasonal climate and streamflow forecasts performance for Mainland Southeast Asia
Abstract. Seasonal forecast is an early warning system that contributes to anticipatory management by providing spatial and temporal information of the near future. This study first examined the skill of ECMWF system 5 (SEAS5) sub–seasonal–to–seasonal (S2S) forecasts over Mainland Southeast Asia (MSEA). We evaluated the SEAS5 skill of temperature and precipitation for 30 years (1985–2014) against two reference model datasets, WFDE5 and APHRODITE, using probabilistic forecast verification skill metrics at grid cells for each month. Then, the SEAS5 data was used to force the Variable Infiltration Capacity (VIC) hydrological model to predict runoff and streamflow. These hydrological results were compared against the WFDE5-driven streamflow reanalysis and observed station data, using the same probabilistic skill statistics. The results show a prediction potential for temperature beyond two months in advance. The skill of precipitation and streamflow forecasting is limited to the first month. Strong seasonal and regional dependence occurs. The model shows high forecast skills during the pre-monsoon (April–May) and post-monsoon (October–November), arguably the period when its usefulness is potentially highest. Conversely, poor skill is observed during the rainy monsoon season (June–August). In eastern and southern MSEA, i.e. in eastern Thailand, Cambodia, Vietnam and Malaysia, considerable skill levels are found. Year–to–year precipitation tercile plots highlight skill in predicting the anomalous seasonal conditions caused by the ENSO. Overall, SEAS5 and derived hydrological forecasts show useful skill that can potentially be used for hydrological and agricultural anticipatory management in this region.
- Preprint
(3074 KB) - Metadata XML
-
Supplement
(2324 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2023-56', Anonymous Referee #1, 29 Apr 2023
General comments
The main goal of the manuscript is to evaluate the ability of the forecasting model ECMWF SEAS5 to simulate the climatology of precipitation and temperature over Malaysia and the accuracy of streamflow predictions, at various lead times, forced by the seasonal atmospheric model for different seasons and sub-regions.
Because of the relevance for water managers, as reinforced in the introduction sections it is clear that the manuscript is of general interest for HESS´s readers.
There are, however, several points in the manuscript that I believe that need to be considered by the authors in order to improve the paper interest.
In my view the manuscript needs to be reduced and focus in analyzing what is relevant in a tropical country, which is rainfall rather than discharge. Then, discussion can concentrate on what matters, which is river discharges. The current version is too long, which makes difficult to read.
Considering the current limitations of climate model for accurately predicting many month in advance, I wander why the statistics chosen are mainly numerical (thus is, correlation) rather than using categorical indexes of performance.
There are several specific comments below.
Specific comments
Lines 30-35. While I understand the advantage for water managers of using probabilistic forecast at intra-seasonal scales, this paragraph regarding climate change might be out of context because the adverse impacts of global climate change increase along decadal time-scales, while the focus of the study are the forecast for lead times of ~ 30 days.
Data description item: the manuscript has so many acronyms that makes it hard to remember while reading. My suggestion is to include a table in the data description section indicating time and space resolution of each data set.
- 90-95. I am not very familiar with the WFDE5 dataset which was used as a reference data. In my experience, what is relevant in tropical areas for hydrological forecast is to remove the errors in rainfall since precipitation has the greatest impact on discharges. How the WFDE5 dataset compares with rainfall estimations derived, for instance form satellite products such as IMERG, etc. I´ve seen validations were performed against the APHRODITE database, which is based in station data, but apparently not for the whole hindcast period.
- 130-135. If I understood correctly, the experiments were performed only for ENSO years. How many events were considered? It is not clear for me whether a single experiment (thus is, a single forecast) for each season was carried out; or several runs for each season were for different initial conditions throughout each season was used to calculate the statistics of performance.
- 145, item 4.2.1 Near surface temperature. The authors need to justify all the statistics involving temperature. While I do understand the hydrological implications of the accuracy of temperature forecasts in a temperate country because it has to do with melting, in my view it is of low interest in a tropical country. Are there any hydrological implications, I mean on discharge values whether the predicted temperature was 28 ⁰C while the observations was 27⁰ C?
- 160 The fact that the observation driven data set, APHRODITE, “shows 160 a higher skill magnitude compared with the evaluation against WFDE5, especially during the rainy season” might be due to biases in the modeling driven WFDE5 (?).
Figure 7 it appears to me that runoff prediction is better than discharge but the statement of line 217 concluded the opposite.
- The problem with this analysis against discharge observation is related to the fact that it includes the uncertainties of the VIC model. Besides all the errors due to the forecasts of the mode, the WFDE5 initialization, etc, how much of the variability can be attributed to the hydrological model itself?
Citation: https://doi.org/10.5194/hess-2023-56-RC1 - AC1: 'Reply on RC1', Ubolya Wanthanaporn, 17 May 2023
-
RC2: 'Comment on hess-2023-56', Anonymous Referee #2, 04 May 2023
General comments
The authors firstly evaluate the performance of ECMWF system 5 (SEAS5) forecasts in Mainland Southeast Asia (MSEA). The SEAS5 forecasts are used as inputs to VIC hydrological model and to predict runoff and streamflow. The results show that high forecast skills exist during the pre-monsoon (April–May) and post-monsoon (October–November), but limited skill is observed during rainy monsoon season. The results are meaningful for peoples who are interested in seasonal hydrogeological forecasts in the research region.
Specific comments:
- There are several parameters that may need to be calibrated in VIC, but I didn’t found descriptions of parameter calibration for the VIC hydrological model. Did the authors just use the default parameters in VIC model?
- The authors sometimes use the term “subseasonal” to describe SEAS5 forecasts, e.g. “ECMWF system 5 (SEAS5) sub–seasonal–to–seasonal (S2S) forecasts”. However, from my point of view, SEAS5 mainly provides monthly forecasts and is a seasonal forecast product. Subseasonal forecasts usually refer to daily weather forecasts with lead times between 15-60 days. I suggest the authors to check the use of term “subseasonal” in the manuscript.
Language and technical issues:
- There are too many lines in Fig. 3 and Fig. 5, which makes the figure difficult to understand. The authors may consider improve those figures.
- The language needs to be thoroughly improved. For example, Line 205, “to forecasts” should be “forecast”; Line 227, “skilful” should be “skill” or “skills” in my opinion.
Citation: https://doi.org/10.5194/hess-2023-56-RC2 - AC2: 'Reply on RC2', Ubolya Wanthanaporn, 17 May 2023
-
RC3: 'Comment on hess-2023-56', Anonymous Referee #3, 22 May 2023
This manuscript described an interesting study on evaluation of the forecast skill of SEAS5 in MSEA. Authors concluded that SEAS5 has high forecast skills during the pre-monsoon (April–May) and post-monsoon (October–November), while poor skill is observed during the rainy monsoon season. The paper was written in good style and logical lines. Please see my following comments:
Detailed information of ten streamflow gauging stations should be listed (Lines 103-105). Which basin or sub-basin are these stations located, what are their relationships in terms of upstream and downstream? Figure1, A undelay basin map may be better compared to the country map, same as following figures.
Please add data availability and data source contents (link for SEAS5, WFDE5, APHRODITE).
As pointed out by other reviewers, the paper is unnecessarily too long because of displaying too many figures and results that can be moved to the supplementary materials. Please consider shorten the paper.
Fig. 3, 5, Each colored line follows the skill of a single forecast. Then which color represent which single forecast? The meaning of legend “Lead m 0/1/2” should be explained.
Figure 4,6,8,9 Please consider transfer the figures to seasonal scale (MAM, JJA, SON), which is consistent with later description. There are too many figures in the main text which lead readers confusing. Fig. 13-15 can also be concise, use supplementary to display the repetitive and similar information.
For hydrological simulation by VIC, the parameter calibration and model validation processes should be clarified. The influencing factors on stream flow should be discussed, like land use change, dam construction, et al. How these factors influencing the forecast skills of SEAS5, can be discussed. These required a basin-to-basin analysis in MSEA, please authors consider compare the basin variation characteristics, rather than the sub-region analysis in current version.
Citation: https://doi.org/10.5194/hess-2023-56-RC3 - AC3: 'Reply on RC3', Ubolya Wanthanaporn, 02 Jun 2023
Status: closed
-
RC1: 'Comment on hess-2023-56', Anonymous Referee #1, 29 Apr 2023
General comments
The main goal of the manuscript is to evaluate the ability of the forecasting model ECMWF SEAS5 to simulate the climatology of precipitation and temperature over Malaysia and the accuracy of streamflow predictions, at various lead times, forced by the seasonal atmospheric model for different seasons and sub-regions.
Because of the relevance for water managers, as reinforced in the introduction sections it is clear that the manuscript is of general interest for HESS´s readers.
There are, however, several points in the manuscript that I believe that need to be considered by the authors in order to improve the paper interest.
In my view the manuscript needs to be reduced and focus in analyzing what is relevant in a tropical country, which is rainfall rather than discharge. Then, discussion can concentrate on what matters, which is river discharges. The current version is too long, which makes difficult to read.
Considering the current limitations of climate model for accurately predicting many month in advance, I wander why the statistics chosen are mainly numerical (thus is, correlation) rather than using categorical indexes of performance.
There are several specific comments below.
Specific comments
Lines 30-35. While I understand the advantage for water managers of using probabilistic forecast at intra-seasonal scales, this paragraph regarding climate change might be out of context because the adverse impacts of global climate change increase along decadal time-scales, while the focus of the study are the forecast for lead times of ~ 30 days.
Data description item: the manuscript has so many acronyms that makes it hard to remember while reading. My suggestion is to include a table in the data description section indicating time and space resolution of each data set.
- 90-95. I am not very familiar with the WFDE5 dataset which was used as a reference data. In my experience, what is relevant in tropical areas for hydrological forecast is to remove the errors in rainfall since precipitation has the greatest impact on discharges. How the WFDE5 dataset compares with rainfall estimations derived, for instance form satellite products such as IMERG, etc. I´ve seen validations were performed against the APHRODITE database, which is based in station data, but apparently not for the whole hindcast period.
- 130-135. If I understood correctly, the experiments were performed only for ENSO years. How many events were considered? It is not clear for me whether a single experiment (thus is, a single forecast) for each season was carried out; or several runs for each season were for different initial conditions throughout each season was used to calculate the statistics of performance.
- 145, item 4.2.1 Near surface temperature. The authors need to justify all the statistics involving temperature. While I do understand the hydrological implications of the accuracy of temperature forecasts in a temperate country because it has to do with melting, in my view it is of low interest in a tropical country. Are there any hydrological implications, I mean on discharge values whether the predicted temperature was 28 ⁰C while the observations was 27⁰ C?
- 160 The fact that the observation driven data set, APHRODITE, “shows 160 a higher skill magnitude compared with the evaluation against WFDE5, especially during the rainy season” might be due to biases in the modeling driven WFDE5 (?).
Figure 7 it appears to me that runoff prediction is better than discharge but the statement of line 217 concluded the opposite.
- The problem with this analysis against discharge observation is related to the fact that it includes the uncertainties of the VIC model. Besides all the errors due to the forecasts of the mode, the WFDE5 initialization, etc, how much of the variability can be attributed to the hydrological model itself?
Citation: https://doi.org/10.5194/hess-2023-56-RC1 - AC1: 'Reply on RC1', Ubolya Wanthanaporn, 17 May 2023
-
RC2: 'Comment on hess-2023-56', Anonymous Referee #2, 04 May 2023
General comments
The authors firstly evaluate the performance of ECMWF system 5 (SEAS5) forecasts in Mainland Southeast Asia (MSEA). The SEAS5 forecasts are used as inputs to VIC hydrological model and to predict runoff and streamflow. The results show that high forecast skills exist during the pre-monsoon (April–May) and post-monsoon (October–November), but limited skill is observed during rainy monsoon season. The results are meaningful for peoples who are interested in seasonal hydrogeological forecasts in the research region.
Specific comments:
- There are several parameters that may need to be calibrated in VIC, but I didn’t found descriptions of parameter calibration for the VIC hydrological model. Did the authors just use the default parameters in VIC model?
- The authors sometimes use the term “subseasonal” to describe SEAS5 forecasts, e.g. “ECMWF system 5 (SEAS5) sub–seasonal–to–seasonal (S2S) forecasts”. However, from my point of view, SEAS5 mainly provides monthly forecasts and is a seasonal forecast product. Subseasonal forecasts usually refer to daily weather forecasts with lead times between 15-60 days. I suggest the authors to check the use of term “subseasonal” in the manuscript.
Language and technical issues:
- There are too many lines in Fig. 3 and Fig. 5, which makes the figure difficult to understand. The authors may consider improve those figures.
- The language needs to be thoroughly improved. For example, Line 205, “to forecasts” should be “forecast”; Line 227, “skilful” should be “skill” or “skills” in my opinion.
Citation: https://doi.org/10.5194/hess-2023-56-RC2 - AC2: 'Reply on RC2', Ubolya Wanthanaporn, 17 May 2023
-
RC3: 'Comment on hess-2023-56', Anonymous Referee #3, 22 May 2023
This manuscript described an interesting study on evaluation of the forecast skill of SEAS5 in MSEA. Authors concluded that SEAS5 has high forecast skills during the pre-monsoon (April–May) and post-monsoon (October–November), while poor skill is observed during the rainy monsoon season. The paper was written in good style and logical lines. Please see my following comments:
Detailed information of ten streamflow gauging stations should be listed (Lines 103-105). Which basin or sub-basin are these stations located, what are their relationships in terms of upstream and downstream? Figure1, A undelay basin map may be better compared to the country map, same as following figures.
Please add data availability and data source contents (link for SEAS5, WFDE5, APHRODITE).
As pointed out by other reviewers, the paper is unnecessarily too long because of displaying too many figures and results that can be moved to the supplementary materials. Please consider shorten the paper.
Fig. 3, 5, Each colored line follows the skill of a single forecast. Then which color represent which single forecast? The meaning of legend “Lead m 0/1/2” should be explained.
Figure 4,6,8,9 Please consider transfer the figures to seasonal scale (MAM, JJA, SON), which is consistent with later description. There are too many figures in the main text which lead readers confusing. Fig. 13-15 can also be concise, use supplementary to display the repetitive and similar information.
For hydrological simulation by VIC, the parameter calibration and model validation processes should be clarified. The influencing factors on stream flow should be discussed, like land use change, dam construction, et al. How these factors influencing the forecast skills of SEAS5, can be discussed. These required a basin-to-basin analysis in MSEA, please authors consider compare the basin variation characteristics, rather than the sub-region analysis in current version.
Citation: https://doi.org/10.5194/hess-2023-56-RC3 - AC3: 'Reply on RC3', Ubolya Wanthanaporn, 02 Jun 2023
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 657 | 241 | 50 | 948 | 140 | 38 | 47 |
- HTML: 657
- PDF: 241
- XML: 50
- Total: 948
- Supplement: 140
- BibTeX: 38
- EndNote: 47
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 304 | 32 |
| China | 2 | 107 | 11 |
| Netherlands | 3 | 96 | 10 |
| Germany | 4 | 88 | 9 |
| France | 5 | 24 | 2 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 304
