Detection of agricultural flash droughts using impact-informed thresholds

Ubeda-Trujillo, Ingrid; Werner, Micha; Bertini, Claudia; Coenders-Gerrits, Miriam; Jewitt, Graham

doi:10.5194/egusphere-2026-1010

Preprints

https://doi.org/10.5194/egusphere-2026-1010

Preprints

10 Mar 2026

| 10 Mar 2026

Status: this preprint is open for discussion and under review for Natural Hazards and Earth System Sciences (NHESS).

Detection of agricultural flash droughts using impact-informed thresholds

Ingrid Ubeda-Trujillo, Micha Werner, Claudia Bertini, Miriam Coenders-Gerrits, and Graham Jewitt

Abstract. A rapid and sustained evaporative stress causes soils to lose moisture within days to weeks during sensitive crop phases. This rapid soil depletion can severely affect agricultural systems and manifest as events commonly referred to as agricultural flash droughts (AFDs). AFDs are often detected using literature-based thresholds that assume uniformity across time and space. This may, however, poorly represent the occurrence and impacts of AFDs in local contexts and limit usefulness to decision-makers who require context-specific information. This study aims to overcomes these limitations by introducing an impact-informed calibration of the Standardized Evaporative Stress Ratio (SESR) by linking SESR thresholds to observed crop losses. Thresholds are derived temporally across crop management periods and spatially across soil textures. The proposed calibration approach is tested in Nicaragua’s Dry Corridor of Central America, where rainfed agriculture predominates. Results suggest that although SESR exhibits coherent evaporative stress signals during years with drought-induced crop losses, literature-based thresholds showed limited skill in translating these signals into agriculturally relevant flash drought detection across different crop management periods and soil textures. Threshold performance varies across crop management periods and soil textures. Reliable thresholds could not be detected in sandy soils, whereas clay-rich soils exhibited stable and skilful thresholds. In loamy–silty soils, skilful thresholds were mainly observed during March and the pre-sowing period, with limited skill thereafter. Validation results also showed that calibrated thresholds, within the crop periods-soil combinations identified as skilful, successfully detected flash drought events associated with crop losses more frequently than those reported in the literature. Together, these findings indicate that monitoring AFD using impact-informed thresholds has an operational value for drought management and can support better decision-making.

Received: 21 Feb 2026 – Discussion started: 10 Mar 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2910 KB)

Supplement (1817 KB)

Download & links

Ingrid Ubeda-Trujillo, Micha Werner, Claudia Bertini, Miriam Coenders-Gerrits, and Graham Jewitt

Status: open (until 27 Apr 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-1010', Anonymous Referee #1, 23 Mar 2026 reply
Review for the paper ‘Detection of agricultural flash droughts using impact-informed thresholds’

Overall Summary and Recommendation
This manuscript explores an improvement to the existing approach of detecting AFDs by calibrating SESR at different soil texture and stages within cropping season as well as integrating information regarding verified drought events that causes crop loss. While the study is useful in drought monitoring, the manuscript needs major revision before recommending for acceptance.
The manuscript presented the introduction, discussion and conclusion very well with few minor comments but needs improvement in the method section including the results. Alignment of the discussion is need and some clarification to improve the coherency of discussion of the manuscript.
Specifically, I recommend (1) language editing to improve overall grammar and correct typological errors for better readability (for few identified sentences and paragraphs), (2) structuring the manuscript according to the template of the journal as well the structure of the contents, (3) presenting detailed information specially in methodology as well as in results, and (4) improving the presentation of figures and its captions in the methodology, results including the supplemental materials. For now, here are the major comments that to be needs clarified and addressed:
Major Comments
I felt the manuscript needs a language editing to improve the overall readability for some sentences and paragraphs. There were typological errors and some sentences that needs grammatical correction for clarity and coherency. Consider also the language preference for the journal (English guidelines) as well as the overall structuring of the manuscript (Submission).

In the abstract, it introduces the problem, aim, and results. This can be further improved by highlighting the main results in line to the aim. In the current text, the results mainly presented the limitation of the literature-based approach instead of how well the proposed approach improve the detection of AFD as compared with literature-based approach. Give an emphasis on the skill of the new approach in detecting AFD over literature-based approach.

Method section can be further improved by structuring a coherent flow of the processes in conjunction to the objectives presented. A confusion on the dataset described in section 2.2 where MODIS MYD16A3GF having annual temporal resolution was later describe in succeeding text of using MODIS data with an 8-day resolution specifically on calculating SESR. Moreover, a need for a more detailed description of the specification of the datasets being used.

Methods sections 2.2, 2.3, if possible, provide flowcharts for ease visualization and reduce the text intensive description of the procedure or processes done.

In section 2.3.2, what do you mean by the 3km sampling interval? It is not clear to me how the sampling method is done. Does it mean you selected a 3km x 3km grid for each soil texture? the 388 timeseries is also not clear which time scale it is referring to? I am confused with the data being processed because the MODIS MYD16A3GF has annual temporal resolution and here you are mentioning an 8-day SESR values? In addition, why do you need to do sampling? is it not possible to process all the pixels covered by each soil texture? was the data too large that takes too long to perform the computations?

In section 2.3.2, what is the relevance on why you consider the criteria of having VHI below 0.35 at the end of cropping season as indication of drought years with crop losses? why not use the VHI on the first month of the growing season, or the VHI at the 3 month of the growing season which coincides with the flowering and yield formation stage for maize and the most sensitive stages to drought which could cause yield reduction or crop damage?

In section 2.3.3, is the first paragraph describing the literature-based AFD detection? If so, I think the title needs to add a term like 'literature-based AFD detection' just to showcase the method. It is also better if a more detailed description on how this approach (literature-based) is applied to detect an AFD for a reader who is not familiar with the method. At first reading, I could not understand how the method works. Moreover, a separate section to discuss the proposed enhancement to the existing method is suggested. In this way, as a reader I can separately compare the two approaches.

IN SECTION 2.3.6, how do you evaluate the performance of the calibrated threshold and the performance of the literature-based approach? what are the metrics you measured to determine whether an improvement is achieved by the calibrated SESR Threshold? what does it mean by V1 =1 or V1 = 0, does it mean a hit or miss? For example, for year 2014, if it is V1=1 then it is a hit meaning it predicted or detected the occurrence of AFD? what is crop loss condition = 1 means?

In section 3.1, for the results, since literature-based approach have already used the SESR as the parameter in detecting AFD, why do you need to correlate SESR with VHI. A need to expound the relevance of this results in connection to the overall aim of the study. In addition, I also don't see results summarizing the identified years having drought with crop losses and without crop losses used in the calibration and validation. For transparency of the data used, please include the results maybe as supplemental tables.

In section 3.2, Performance was presented however, it is also important to showcase the calibrated threshold as it was presented as a supplemental material Fig. S3. A section should be allotted for the presentation of the variation of the calibrated optimal thresholds prior to section 3.2. In addition, the context where SESR threshold being calibrated for the different 'crop management period' was not implemented, instead the results showed a calibrated SESR thresholds at monthly basis and not on the time frame for each crop management period. Furthermore, the calibrated SESR threshold was not clearly describe how it was selected among the incremental runs of maximizing . It was also confusing when I relate it in section 3.3 where it presents plots of optimal SESR thresholds. There were several optimal SESR derived but how did you select the best SESR presented in Fig. 6 among these SESR?

In section 3.3, Is this the criteria (i.e., precision ≥ 0.70, FAR ≤ 0.30, ROC-AUC ≥ 0.60, and CI95% ≤ 0.40) used in selecting the best SESR Threshold? It was not described in the methodology. Please clarify that on top of the method presented of establishing the SESR threshold by maximizing , you have these additional criteria. It is confusing also that several optimal SESR threholds was generated as shown in your Fig. S3 and in this section you retain a certain SESR threshold based on these additional criteria. Clarify and make the method of selecting the optimal SESR threshold coherent with the results presented. Please improve the methodology part to better understand the results being presented.

In section 3.4, It is noteworthy to include the supplemental figures 4 and 5 as main results and not as a supplemental material.

Minor Comments:
Please refer to the attached manuscript where minor comments are made (including the major comments).

Reply
Citation: https://doi.org/10.5194/egusphere-2026-1010-RC1

Ingrid Ubeda-Trujillo, Micha Werner, Claudia Bertini, Miriam Coenders-Gerrits, and Graham Jewitt

Supplement

https://doi.org/10.5194/egusphere-2026-1010-supplement

Ingrid Ubeda-Trujillo, Micha Werner, Claudia Bertini, Miriam Coenders-Gerrits, and Graham Jewitt

Viewed

Total article views: 192 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
116	60	16	192	22	10	14

HTML: 116
PDF: 60
XML: 16
Total: 192
Supplement: 22
BibTeX: 10
EndNote: 14

Views and downloads (calculated since 10 Mar 2026)

Month	HTML	PDF	XML	Total
Mar 2026	113	58	16	187
Apr 2026	3	2	0	5

Cumulative views and downloads (calculated since 10 Mar 2026)

Month	HTML	PDF	XML	Total
Mar 2026	113	58	16	187
Apr 2026	3	2	0	5

Viewed (geographical distribution)

Total article views: 186 (including HTML, PDF, and XML) Thereof 186 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 04 Apr 2026

Short summary

Some droughts develop so quickly that soils dry out, and crops can be damaged before existing methods notice them. These effects are especially harmful in dryland regions, where crops rely on rainfall and farmers often have limited resources. This study defines drought alert points that indicate when soils dry quickly enough to affect crops. By tailoring these alerts to each soil type and crop stage, they become more reliable, helping farmers and planners act in time to reduce crop damage.


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