Sub-seasonal precipitation forecasts using preceding atmospheric intraseasonal oscillation signals in a Bayesian perspective

Li, Yuan; Wu, Zhiyong; He, Hai; Yin, Hao

doi:https://doi.org/10.5194/hess-2022-67

Preprints

https://doi.org/10.5194/hess-2022-67

Preprints

03 Mar 2022

Status: a revised version of this preprint was accepted for the journal HESS and is expected to appear here in due course.

Sub-seasonal precipitation forecasts using preceding atmospheric intraseasonal oscillation signals in a Bayesian perspective

Yuan Li, Zhiyong Wu, Hai He, and Hao Yin Yuan Li et al.

Show author details

College of Hydrology and Water Resources, Hohai University, Nanjing 210098, China

Received: 14 Feb 2022 – Discussion started: 03 Mar 2022

Abstract. Accurate and reliable sub-seasonal precipitation forecasts remain challenging. The atmospheric intraseasonal oscillation (ISO), which is one of the leading sources of sub-seasonal predictability, could be potentially used as predictors for sub-seasonal precipitation forecasts.. However, the relationships between ISO signals and sub-seasonal precipitation are of high uncertainty. In this study, we first define potential predictors by analyzing the relationship between preceding atmospheric ISO signals and precipitation for 17 hydroclimatic regions over China during the boreal summer monsoon season. The Least Absolute Shrinkage and Selection Operator (LASSO) and stepwise regression approaches are used to narrow down the number of potential predictors. A Bayesian hierarchical model is then established to predict sub-seasonal precipitation. The model performance is evaluated through a leave one-year-out cross-validation strategy for both deterministic and probabilistic forecasts. The results suggest that the statistical model we built in this study could provide skillful deterministic sub-seasonal precipitation forecasts over southeastern and southwestern hydroclimatic regions at a lead time of 20–25 days. However, the deterministic forecast skills are much lower over northeastern China, owing to the underestimation of intraseasonal variability in these regions. The probabilistic forecasts are more promising, and the results indicate that the Bayesian hierarchical model could provide skillful and reliable sub-seasonal precipitation forecasts for all hydroclimatic regions from 0-day to 25-day leads. Other sources of sub-seasonal predictability would be included in the future to further improve sub-seasonal precipitation forecast skills.

Yuan Li et al.

Status: closed

RC1:
'Comment on hess-2022-67', Anonymous Referee #1, 28 Mar 2022

The authors established a Bayesian hierarchical model (BHM) to predict the 10-60d precipitation for 17 hydroclimatic regions over China during the boreal summer monsoon season (May to October) by using the previous atmospheric intraseasonal signals. Both deterministic and probabilistic evaluations showed that the BHM provides skillful subseasonal forecasts over southeastern and southwestern hydroclimatic regions at a lead time of 20-25 days while the skills are poor over northeastern China, owing to the underestimation of intraseasonal variability.

The authors have conducted numerous calculations and employed many different statistical analysis methods. However, the explanation for their choice of the calculation and methods are deficient. Moreover, I cannot tell whether the BHM proposed in this paper show any superior skills than other statistical models or even dynamical S2S models. From this point of view, I incline to reject the manuscript, but I give an opportunity to the authors to improve the manuscript.

Major comments:

1. The intraseasonal variability and the intraseasonal oscillation are different terms. The authors focus on the prediction of intraseasonal precipitation (10-60d) over China during summer (May to October). Although the selected predictors are atmospheric intraseasonal signals, no specific BSISO or MJO pattern can be found in the previous correlation maps. The title may be more consistent with the content after removing “oscillation”.

2. The selected intraseasonal signals and the physical processes of their influencing on precipitation over China should be provided.

3. For each region and each pentad from May to October, a BHM is built to forecast precipitation at different lead time. The detail information should be shown in caption of Fig.2, Fig.3, Fig.5. Are the results in these figures for a specific pentad or the average mean from May to October? If the latter is the case, will the skill for each pentad be similar throughout the whole summer?

4. Figure 1 shows the division of the hydroclimatic regions. However, this is not a scientific way to divide China with respect to rainfall variation. Does the precipitation in each region have the coherent intraseasonal variation? If not, the correlation map is meaningless because they are calculated based on the areal-mean precipitation. Moreover, do we really need 17 regions?

5. No evidences are provided to justify the advantage of this prediction model. Does this model have better performance than the ECMWF S2S model? Or the spatial-temporal projection models (STPM)? The authors need to make some comparison.

6. From Figure 4, I can see the prediction skills mainly came from the annual cycle (which is quite stable), rather than the pentad variation (the intraseasonal component). How about the skills if only anomaly precipitation is verified? I think the skill is very limited.

Minor comments:

1. Page 1 Line 9, ”..” is “.”.

2.Page 9 Line 208, the order of Fig. 2 is confusing for the reader to discern the evolution of intraseasonal atmospheric signals from Lead 25d to Lead 0d. The figure can be sliced to two figures with the first one showing the correlation between preceding U850, U200, OLR and 10-60d precipitation from Lead 25d to Lead 0d, and the second one showing the remaining H850, H500, and H200.

3. Page 15 Line 355, Fig. 3. The skill of Kling-Gupta Efficiency (KGE) in region 2, region 9 and region 12 increases with time, why? Could you please show r, β and γ before you show KGE? Because correlation coefficient and bias are the basic metric for forecast verification.

4. Page 15 Line 355, Fig. 3. The prediction skill (KGE) of region 1 is the best in 17 regions, but in Fig. 4, the BHM model shows no skills for extreme events. Please explain the reason.

5. Page 16 Line 365. What is the standard of efficient prediction in KGE and Continuous Ranked Probability Score (CRPS)? In the paper, the authors use “0.2” and “positive” as the standards, what is the reason?

6. Page 18 Line 385. The prediction skill over northeast China is relatively lower than that over southeastern and southwestern China. Although the number of samples will be induced, the results of southeastern and southwestern China can better demonstrate the skill of BHM.

7. Page 18 Line 385. There is no caption of a detail description of the size of dots.

8. Line 355, during the boreal summer monsoon season.

9. Line 55-70, So far, there are many statistical models for subseasonal prediction (some of them were already used in operational subseasonal prediction). The authors may want to read or cite the following publications, and make comparisons with their model.

Zhu Z., T. Li, P.-C. Hsu, J. He, 2015: A spatial-temporal projection model for extended-range forecast in the tropics. Clim. Dyn., 45(3), 1085-1098. doi: 10.1007/s00382-014-2353-8.

Zhu Z., T. Li, 2018: Extended-range forecasting of Chinese summer surface air temperature and heat waves. Clim. Dyn., 50(5-6), 2007-2021. doi: 10.1007/s00382-017-3733-7.

Zhu Z., T. Li, 2017: The statistical extended-range (10–30-day) forecast of summer rainfall anomalies over the entire China. Clim. Dyn., 48(1), 209-224. doi: 10.1007/s00382-016-3070-2.

Zhu Z., T. Li, 2017: Empirical prediction of the onset dates of South China Sea summer monsoon. Clim. Dyn., 48(5), 1633-1645. doi: 10.1007/s00382-016-3164-x.

Zhu Z., T. Li, 2017: Statistical extended-range forecast of winter surface air temperature and extremely cold days over China. Q. J. R. Meteor. Soc., 704(143), 1528-1538. doi: 10.1002/qj.3023.

Zhu Z., S. Chen, K. Yuan, Y. Chen, S. Gao, Z. Hua, 2017: Empirical subseasonal predicting summer rainfall anomalies over the middle and lower reaches of Yangtze River basin based on the atmospheric intraseasonal oscillation. Atmos., 8(10), 185. doi:10.3390/atmos8100185.

Zhu Z., T. Li, L. Bai, J. Gao, 2017: Extended-range forecast for the temporal distribution of clustering tropical cyclogenesis over the western North Pacific. Theor. Appl. Climatol., 130(3), 865-877. doi: 10.1007/s00704-016-1925-4.

Li W., P. Hsu, J. He, Z. Zhu, W. Zhang, 2016: Extended-range forecast of spring rainfall in southern China based on the Madden–Julian Oscillation. Meteorol. Atmos. Phys., 128(3), 331-345. doi: 10.1007/s00703-015-0418-9.

10. Line 75-80, “However, we should note that the relationships between ISO signals and precipitation are of high uncertainty for different regions at different lead times”

Yes, that is why in Zhu and Li (2017), they used REOF to divided the mainland China into 10 subregions based on the coherent nature of the 10-90 variation in each subregion. They predicted 10-30day predictand at once because considering the whole process of intraseasonal variability with the time-varying and spatial varying information. The authors may want to read the paper via the following link:

http://dqkxxb.cnjournals.org/dqkxxb/article/abstract/20200120

Citation: https://doi.org/10.5194/hess-2022-67-RC1
- AC1:
  'Reply on RC1', Zhiyong Wu, 12 Apr 2022
  
  The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-67/hess-2022-67-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/hess-2022-67-AC1
  - RC2: 'Reply on AC1', Anonymous Referee #1, 20 Apr 2022
    
    The authors have addressed well my former major concerns and comments. Therefore, I have no further comments.
    
    Citation: https://doi.org/10.5194/hess-2022-67-RC2
    
    AC3: 'Reply on RC2', Zhiyong Wu, 07 Jun 2022
    
    Thanks for the discussions, I think that they should lead to a much improved manuscript.
    
    Citation: https://doi.org/10.5194/hess-2022-67-AC3
RC3:
'Comment on hess-2022-67', Anonymous Referee #2, 11 May 2022

See attached file

Citation: https://doi.org/10.5194/hess-2022-67-RC3
- AC2: 'Reply on RC3', Zhiyong Wu, 02 Jun 2022
  
  The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-67/hess-2022-67-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/hess-2022-67-AC2

Status: closed

RC1:
'Comment on hess-2022-67', Anonymous Referee #1, 28 Mar 2022

The authors established a Bayesian hierarchical model (BHM) to predict the 10-60d precipitation for 17 hydroclimatic regions over China during the boreal summer monsoon season (May to October) by using the previous atmospheric intraseasonal signals. Both deterministic and probabilistic evaluations showed that the BHM provides skillful subseasonal forecasts over southeastern and southwestern hydroclimatic regions at a lead time of 20-25 days while the skills are poor over northeastern China, owing to the underestimation of intraseasonal variability.

The authors have conducted numerous calculations and employed many different statistical analysis methods. However, the explanation for their choice of the calculation and methods are deficient. Moreover, I cannot tell whether the BHM proposed in this paper show any superior skills than other statistical models or even dynamical S2S models. From this point of view, I incline to reject the manuscript, but I give an opportunity to the authors to improve the manuscript.

Major comments:

1. The intraseasonal variability and the intraseasonal oscillation are different terms. The authors focus on the prediction of intraseasonal precipitation (10-60d) over China during summer (May to October). Although the selected predictors are atmospheric intraseasonal signals, no specific BSISO or MJO pattern can be found in the previous correlation maps. The title may be more consistent with the content after removing “oscillation”.

2. The selected intraseasonal signals and the physical processes of their influencing on precipitation over China should be provided.

3. For each region and each pentad from May to October, a BHM is built to forecast precipitation at different lead time. The detail information should be shown in caption of Fig.2, Fig.3, Fig.5. Are the results in these figures for a specific pentad or the average mean from May to October? If the latter is the case, will the skill for each pentad be similar throughout the whole summer?

4. Figure 1 shows the division of the hydroclimatic regions. However, this is not a scientific way to divide China with respect to rainfall variation. Does the precipitation in each region have the coherent intraseasonal variation? If not, the correlation map is meaningless because they are calculated based on the areal-mean precipitation. Moreover, do we really need 17 regions?

5. No evidences are provided to justify the advantage of this prediction model. Does this model have better performance than the ECMWF S2S model? Or the spatial-temporal projection models (STPM)? The authors need to make some comparison.

6. From Figure 4, I can see the prediction skills mainly came from the annual cycle (which is quite stable), rather than the pentad variation (the intraseasonal component). How about the skills if only anomaly precipitation is verified? I think the skill is very limited.

Minor comments:

1. Page 1 Line 9, ”..” is “.”.

2.Page 9 Line 208, the order of Fig. 2 is confusing for the reader to discern the evolution of intraseasonal atmospheric signals from Lead 25d to Lead 0d. The figure can be sliced to two figures with the first one showing the correlation between preceding U850, U200, OLR and 10-60d precipitation from Lead 25d to Lead 0d, and the second one showing the remaining H850, H500, and H200.

3. Page 15 Line 355, Fig. 3. The skill of Kling-Gupta Efficiency (KGE) in region 2, region 9 and region 12 increases with time, why? Could you please show r, β and γ before you show KGE? Because correlation coefficient and bias are the basic metric for forecast verification.

4. Page 15 Line 355, Fig. 3. The prediction skill (KGE) of region 1 is the best in 17 regions, but in Fig. 4, the BHM model shows no skills for extreme events. Please explain the reason.

5. Page 16 Line 365. What is the standard of efficient prediction in KGE and Continuous Ranked Probability Score (CRPS)? In the paper, the authors use “0.2” and “positive” as the standards, what is the reason?

6. Page 18 Line 385. The prediction skill over northeast China is relatively lower than that over southeastern and southwestern China. Although the number of samples will be induced, the results of southeastern and southwestern China can better demonstrate the skill of BHM.

7. Page 18 Line 385. There is no caption of a detail description of the size of dots.

8. Line 355, during the boreal summer monsoon season.

9. Line 55-70, So far, there are many statistical models for subseasonal prediction (some of them were already used in operational subseasonal prediction). The authors may want to read or cite the following publications, and make comparisons with their model.

Zhu Z., T. Li, P.-C. Hsu, J. He, 2015: A spatial-temporal projection model for extended-range forecast in the tropics. Clim. Dyn., 45(3), 1085-1098. doi: 10.1007/s00382-014-2353-8.

Zhu Z., T. Li, 2018: Extended-range forecasting of Chinese summer surface air temperature and heat waves. Clim. Dyn., 50(5-6), 2007-2021. doi: 10.1007/s00382-017-3733-7.

Zhu Z., T. Li, 2017: The statistical extended-range (10–30-day) forecast of summer rainfall anomalies over the entire China. Clim. Dyn., 48(1), 209-224. doi: 10.1007/s00382-016-3070-2.

Zhu Z., T. Li, 2017: Empirical prediction of the onset dates of South China Sea summer monsoon. Clim. Dyn., 48(5), 1633-1645. doi: 10.1007/s00382-016-3164-x.

Zhu Z., T. Li, 2017: Statistical extended-range forecast of winter surface air temperature and extremely cold days over China. Q. J. R. Meteor. Soc., 704(143), 1528-1538. doi: 10.1002/qj.3023.

Zhu Z., S. Chen, K. Yuan, Y. Chen, S. Gao, Z. Hua, 2017: Empirical subseasonal predicting summer rainfall anomalies over the middle and lower reaches of Yangtze River basin based on the atmospheric intraseasonal oscillation. Atmos., 8(10), 185. doi:10.3390/atmos8100185.

Zhu Z., T. Li, L. Bai, J. Gao, 2017: Extended-range forecast for the temporal distribution of clustering tropical cyclogenesis over the western North Pacific. Theor. Appl. Climatol., 130(3), 865-877. doi: 10.1007/s00704-016-1925-4.

Li W., P. Hsu, J. He, Z. Zhu, W. Zhang, 2016: Extended-range forecast of spring rainfall in southern China based on the Madden–Julian Oscillation. Meteorol. Atmos. Phys., 128(3), 331-345. doi: 10.1007/s00703-015-0418-9.

10. Line 75-80, “However, we should note that the relationships between ISO signals and precipitation are of high uncertainty for different regions at different lead times”

Yes, that is why in Zhu and Li (2017), they used REOF to divided the mainland China into 10 subregions based on the coherent nature of the 10-90 variation in each subregion. They predicted 10-30day predictand at once because considering the whole process of intraseasonal variability with the time-varying and spatial varying information. The authors may want to read the paper via the following link:

http://dqkxxb.cnjournals.org/dqkxxb/article/abstract/20200120

Citation: https://doi.org/10.5194/hess-2022-67-RC1
- AC1:
  'Reply on RC1', Zhiyong Wu, 12 Apr 2022
  
  The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-67/hess-2022-67-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/hess-2022-67-AC1
  - RC2: 'Reply on AC1', Anonymous Referee #1, 20 Apr 2022
    
    The authors have addressed well my former major concerns and comments. Therefore, I have no further comments.
    
    Citation: https://doi.org/10.5194/hess-2022-67-RC2
    
    AC3: 'Reply on RC2', Zhiyong Wu, 07 Jun 2022
    
    Thanks for the discussions, I think that they should lead to a much improved manuscript.
    
    Citation: https://doi.org/10.5194/hess-2022-67-AC3
RC3:
'Comment on hess-2022-67', Anonymous Referee #2, 11 May 2022

See attached file

Citation: https://doi.org/10.5194/hess-2022-67-RC3
- AC2: 'Reply on RC3', Zhiyong Wu, 02 Jun 2022
  
  The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-67/hess-2022-67-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/hess-2022-67-AC2

Yuan Li et al.

Viewed

Total article views: 592 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
461	110	21	592	6	7

HTML: 461
PDF: 110
XML: 21
Total: 592
BibTeX: 6
EndNote: 7

Views and downloads (calculated since 03 Mar 2022)

Month	HTML	PDF	XML	Total
Mar 2022	216	35	6	257
Apr 2022	65	21	5	91
May 2022	77	18	4	99
Jun 2022	51	14	6	71
Jul 2022	27	8	0	35
Aug 2022	19	12	0	31
Sep 2022	6	2	0	8

Cumulative views and downloads (calculated since 03 Mar 2022)

Month	HTML	PDF	XML	Total
Mar 2022	216	35	6	257
Apr 2022	65	21	5	91
May 2022	77	18	4	99
Jun 2022	51	14	6	71
Jul 2022	27	8	0	35
Aug 2022	19	12	0	31
Sep 2022	6	2	0	8

Viewed (geographical distribution)

Total article views: 562 (including HTML, PDF, and XML) Thereof 562 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 16 Sep 2022

Short summary

The atmospheric intraseasonal oscillation (ISO) could be potentially used as predictors for sub-seasonal precipitation forecasts. In this study, we define potential predictors by analyzing the relationship between preceding atmospheric ISO signals and precipitation. A Bayesian hierarchical model (BHM) is then established to generate probabilistic forecasts. The results suggest that the model we built in this study is skillful and reliable for sub-seasonal precipitation forecasts.


Total:	0
HTML:	0
PDF:	0
XML:	0