Do CFSv2 Seasonal Forecasts Help Improve the Forecast of Meteorological Drought over Mainland China?

Round 1

Reviewer 1 Report

I not find this paper to have scientific value.

Manuscript has only 5 pages of text. Authors used in the study forecast system CFSv2 but they never described that system, they only referring to some literature positions. It quite surmising  since the whole paper is so small in size. Also in my opinion conclusions are trivial and not supported by in-deep analysis. I not recommend that paper to the publication in scientific journal.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript presents an analysis of the CFSv2 ensemble’s performance in forecasting the persistence of drought according to the Standardized Precipitation index at 6-months aggregation in China (SPI6). Since the paper aims at demonstrating the seasonal predictive capability of the CFSv2 dynamical model, the structure of the paper should be revised accordingly. The study lacks some verification indicators that would contribute to a better evaluation of predictive performance. Firstly, I suggest to include some graphs and maps to contextualize the SPI6 climatology and anomalies in the study region. This would help the reader to have an idea of the scale (both in time and space) of the problem. Then, I encourage the authors to compute further verification indicators such as the RMS error skill score (RMSSS) to assess the skill of the forecast with respect to the climatological forecast and the anomaly correlation coefficient (ACC) to investigate spatial patterns. Finally, the authors should be conscious that by treating forecast in a deterministic manner, it determines a loss of information about the uncertainty in the forecast. This fact should be highlighted as a limitation of such a prediction system and discussed in the devoted section.

 

To summarize, the paper has an appropriate scientific structure and it is also well presented in the part that regards the prediction process. On the contrary, the paper has several flaws in the part regarding the verification process. Consequently, some additional work is needed to answer what stated on-premise.  In my opinion, these improvements would give merit to this paper.

 

Abstract, title and references

The aim of the work is clear; however, the findings of the study only partially respond to it. Also, the title is too ambitious because it points at “meteorological drought” but the results obtained only focus on a 6-months time scale. 

 

The references are recent and relevant. Moreover, the key studies included are appropriate although some references for seasonal verification studies are required to improve the reliability of the methodology.

 

 

Introduction

The authors present a well-structured framework of what is already known about the seasonal drought prediction. Moreover, the research question is clearly outlined. 

 

 

Data and Methods

Since most of the paper content is focused on verification of predictive performance, the concepts of forecast “accuracy” and “reliability” should be introduced and linked to the corresponding indicators. For instance, the RMS error metric is used to assess the accuracy of a forecast system while the correlation is only a measure of the association. Instead of RMS error, the RMSSS could be used to assess the skill of the forecast with

respect to baseline persistence. This score ranges from minus infinity (no skill) to 1 (perfect forecast). A score larger than 0 indicates an

improvement with respect to the climatological forecast. On the other hand, the reliability could be assessed by computing the fraction of the ensemble spread (standard deviation of the ensemble across all start dates) with the RMS error of the ensemble mean. A forecast is considered reliable if this measure is close to 1 (Slingo, J., Palmer, T., 2011. Uncertainty in weather and climate prediction. Phil. Trans. Roy. Soc. Lond. Math. Phys. Eng. Sci. 369, 4751-4767). Finally, the ability of a forecast to reproduce spatial patterns could be investigated employing the anomaly correlation coefficient (Krishnamurti, T., Rajendran, K., Vijaya Kumar, T., Lord, S., Toth, Z., Zou, X., Cocke, S., Ahlquist, J.E., Navon, I.M., 2003. Improved skill for the anomaly correlation of geopotential heights at 500 hpa. Mon. Weather Rev. 131, 1082-1102.).

 

I have also a few specific issues to be addressed for this section:

1. the software used for the computation should be mentioned;
2. details about the re-gridding process should be given;
3. the identification process of location A and location B climate zones is required (or references);
4. the range of each metric used for the assessment should be declared.

 

 

Results

 

Because of the seasonality of drought phenomena, the analysis of the aggregate predictive performance (I mean for the 12 months altogether) seems to me to be not adequate to assess the performance of a drought prediction system. Rather, I would select specific periods and concentrate the verification analysis only on those periods. In other words, periods that are climatologically dry are not intrinsically interesting for drought prediction and also tends to worsen the predictive performance because of very small precipitation. To this end, the authors have announced in the premise that such an analysis has been carried out although the results do not appear in the paper. On the contrary, Figure 3 is very effective to explain the performance at location-level; however, the correlation index is not adequate to assess forecast skill unless used in combination with other metrics.

 

I have also a few specific issues to be addressed for this section:

 

1. at line 126, the authors state that “RMS errors are mostly smaller than 10, which is reasonably small given the typical values of SPI6 are between -3 and 3.” I would conclude exactly the opposite;
2. the approach used to support the CSFv2 improvements stated at line 142 and visualized in Figure 2 should be described with more details.

 

 

 

 

Specific comments

1. Line 78, it is not clear to me how the number 24 is calculated and why November has a different number of runs; 
2. Line 81, is the observational precipitation dataset freely available? Please, specify and give credits to dataset source;
3. Figure 2, to invert the color scale of one of the two indicators, would help the reader to have the same color for good (bad) skill performance;
4. Figure 3, it should be indicated in the caption that the color scale represents the correlation index, and also the y-axis should be defined.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This is a very interesting study investigating whether seasonal forecasts from dynamical models can improve the prediction of droughts based on SPI index. The manuscript is very well written and scientifically sound. I only have some minor comments that I included for convenience in the commented pdf manuscript file.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

After reading authors careful responses to my negative review and to other reviewer I reconsidered that work as actually quite nice way of testing forecasts produced by dynamical model. It seems first quite poor but now, after the additional sections added I find it useful work and interesting to the potential readers.

 

Reviewer 2 Report

I would like to thank the authors for their careful responses, which I consider highly reasonable and acceptable.