Linking reported drought impacts with drought indices, water scarcity, and aridity: the case of Kenya
Abstract. The relation between drought severity, as expressed through widely used drought indices, and drought impacts is complex. In particular in water-limited regions where water scarcity is prevalent, the attribution of drought impacts is difficult. This study assesses the relation between reported drought impacts, drought indices, water scarcity, and aridity across several counties in Kenya. The monthly bulletins of the National Drought Management Authority in Kenya have been used to gather drought impact data. A Random Forest (RF) model was used to explore which set of drought indices best explains drought impacts on: pasture, livestock deaths, milk production, crop losses, food insecurity, trekking distance for water, and malnutrition. The findings of this study suggest a relation between drought severity and the frequency of drought impacts, whereby the latter also showed a relation with aridity, whilst water scarcity did not. The results of the RF model reveal that drought impacts can be explained by a range of drought indices across regions with different aridity. While the findings strongly depend on the availability of drought impact data and the socio-economic circumstances within a region, this study highlights the potential of linking drought indices with text-based impact reports. In doing so, however, spatial differences in aridity and water scarcity conditions have to be taken into account.
Marleen R. Lam et al.
Status: final response (author comments only)
CC1: 'Comment on egusphere-2022-458', Paolo Tasseron, 17 Aug 2022
- AC4: 'Reply on CC1', Marleen Lam, 05 Dec 2022
RC1: 'Review of Lam et al. (2022): Linking reported drought impacts with drought indices, water scarcity, and aridity: the case of Kenya', Anonymous Referee #1, 25 Nov 2022
- AC1: 'Reply on RC1', Marleen Lam, 05 Dec 2022
- AC5: 'Reply on RC1', Marleen Lam, 05 Dec 2022
RC2: 'Comment on egusphere-2022-458', Anonymous Referee #2, 28 Nov 2022
- AC2: 'Reply on RC2', Marleen Lam, 05 Dec 2022
- AC3: 'Reply on RC2', Marleen Lam, 05 Dec 2022
- AC6: 'Reply on RC2', Marleen Lam, 05 Dec 2022
Marleen R. Lam et al.
Marleen R. Lam et al.
Viewed (geographical distribution)
This study discusses the possible linking of drought indices with text-based documents by a drought management authority in Kenya. An assessment is made how drought impacts relate with drought indices and water scarcity under various circumstances, by using a Random Forest Model. In doing so, the authors aim to contribute to ongoing debates about operational needs for drought monitoring. The importance of improving early warning systems to mitigate adverse consequences of drought is corroborated by the study and its results.
In general, I believe the authors present a very well-written manuscript, with clear and complete (sub)sections. The quantity and quality of references to relevant state-of-the-art studies is spot on. I can only suggest some minor and a few moderate revisions require attention prior to publication. These are summarised below. Well done!
Minor and moderate suggestions for revision.
Consider changing ‘society’ to ‘societies’
I think the terms meteorological and hydrological drought do not need such an elaborate explanation. If you think the readership does need this explanation, then consider adding a short definition of ‘soil moisture/agricultural drought’ as well.
Consider including the Köppen classification to describe climatological setting in Kenya, to highlight the diversity between the relatively wet southwest and dry/arid/desert north and east.
How was the difference in spatial resolution (0.25 degree) with the other datasets (0.1 degree) dealt with in by the authors?
Consider changing ‘for each person’ to ‘per capita’
Replace ‘have been’ by ‘was’
How are the grid cells spread over the different counties? Did you only compute the means of grid cells completely within a county border? Please specify this, also the resolution that was used for the grids.
So the indices were calculated for 1980-2020, but only used for 2014-2020. What is the added value of calculating 1980-2014, if it is not being used?
‘have’ = ‘has’
does this aggregation have any drawbacks?
Could you indicate p-values in this table? i.e. p<0.05 = *, p<0.01 = **, p<0.001 = ***
It is a bit unclear what the authors mean by 2, 4, 2 months out of 10/4,0,4 months out of 12. Consider rewriting this sentence.
The AUC abbreviation was not previously introduced and should be written out fully.
Figure 4 & 5
Font size should be increased. In addition, the range of the X-axis is different for each figure. If relevant for comparison between the different categories (which I think is the case), consider using the same range for each figure. You might even want to plot the points of multiple categories in one figure with different colours, instead of having 14 individual subplots.
Please avoid using the term ‘reliable’ when talking about an objective appraisal, as this is a subjective judgement without qualitative or quantitative data to support the statement. It would improve the statement to shortly explain how the iterative processes and the focus on abnormal conditions took place, as I do not recall reading about this in the methods section. I believe this is important, because the bulletins form an integral part of your analyses.
Great to see the authors mention multiple outlets to complement NDMA bulletins!
‘Good’: please quantify this by for instance mentioning the average AUC value, or repeat the AUC value of the top 3 categories.
‘a kind of lag’ – please specify in more detail.
‘low population density does not imply low water stress’ – this calls for a reference.
‘should have suffered from water scarcity during periods of drought due to the high population density’ – this reads a bit dark. Consider rephrasing the sentence so it does not read as if you want these counties to suffer.
this was already mentioned earlier in the manuscript and does not need repetition here.
The authors mention comparisons with these studies are difficult due to different socio-economic and climatic circumstances. That’s a fair point, but how do they compare in terms of resolution? Is it practically possible to compare these quantitatively, or does this also not work?
Could you please include a reference to this existing database (or name it)?
Is there any indication of the spatial resolution required to capture the regional differences?
This last paragraph of the conclusion reads well and contains and sensible information, but I think it undermines the results of the study. It reads as if the work explored in this study is disqualified a bit, since focus is put on the need for finer resolutions to contribute to the development of early warning systems. I recommend the authors to ‘praise’ their own work a bit more in this last paragraph, instead of talking it down.