The researchers drew on daily rainfall estimates from the CHIRPS satellite-based dataset and monthly soil moisture from the FLDAS Noah land-surface model, both averaged across the full administrative boundary of Gedaref State for the period 1982 to 2020. They supplemented these with annual sorghum and sesame records from the Gedaref State Ministry of Production and Economic Resources and the Mechanized Farming Corporation. Before anything else, the team checked how well the satellite rainfall product matched the available ground truth. Against a monthly station archive, CHIRPS showed a moderate seasonal correlation of 0.625, but agreement in absolute totals was limited, with a bias of minus 77.5 millimetres and a root mean square error of 130 millimetres. The authors attribute part of this discrepancy to the difference in spatial support: the station represents a single point in Gedaref town, while the satellite series averages across the entire state, including drier northern localities.

From the daily rainfall record, the researchers calculated four indicators designed to capture the main forms of within-season rainfall stress. The first was the April-to-November rainfall deficit, representing inadequate seasonal water supply. The second was the maximum number of consecutive dry days during the core June-to-September rainy period, a measure of the longest interruption in wetting. The third was onset delay, defined as the first date on or after 1 May when rainfall accumulated over three consecutive days reached 20 millimetres. The fourth was the share of seasonal rainfall delivered by the five wettest days, which captures how concentrated a season’s water supply is in a handful of violent storms. Across the record, these variables revealed striking variability: onset dates ranged from day 135 to day 240 of the year, maximum dry spells ranged from 2 to 13 days, and the five wettest days contributed between 10.2 and 21 percent of seasonal rainfall.

These four components were then combined into a single score, the Daily Rainfall-Structure Risk Index, or DRSRI, using principal component analysis. The analysis retained one component explaining 60.4 percent of the standardised variance, with coefficients of similar magnitude for all four variables, roughly 26, 25, 23 and 25 percent of the total weighting. Bootstrap resampling confirmed that no coefficient collapsed towards zero. Notably, an equal-weight version of the index correlated with the PCA version at 0.9997, meaning the sophisticated weighting made almost no practical difference to the annual ranking of risk. The index itself, however, turned out to be a powerful indicator of surface drying: higher DRSRI values coincided with lower modelled surface soil moisture anomalies in the top 10 centimetres of soil, with a correlation of minus 0.776 that was highly significant.

Yet the story took an unexpected turn when the researchers tested whether the index could actually predict soil moisture better than seasonal rainfall alone. Using a rolling-origin evaluation, in which models were trained on data from 1982 to 1996 and then progressively refitted as each new year arrived, the four rainfall-structure variables produced only a marginal and statistically unsupported improvement over seasonal rainfall, with root mean square errors of 0.544 versus 0.554 and a p-value of 0.739. The DRSRI alone performed worse than seasonal rainfall, with an RMSE of 0.681. The authors are candid about this result: the index is not a better predictor, but a better diagnostic. It tells you whether an unfavourable season was driven by low rainfall, late onset, a long dry spell, or concentrated storms, information that a single seasonal total cannot provide.

The agricultural analysis revealed perhaps the most consequential finding of the study. The clearest association was not with crop yields, as one might expect, but with the sorghum harvested-area ratio, the percentage of planted land reported as remaining productive through the season. This ratio was negatively correlated with DRSRI at minus 0.448 and positively correlated with seasonal rainfall at 0.440, both significant after false-discovery-rate correction. In contrast, no yield association for either sorghum or sesame remained statistically significant after multiple-testing correction. This distinction matters because yield measures output only from land recorded as productive, while the harvested-area ratio reflects how much planted land survived the season at all. In years with poor rainfall structure, farmers apparently lost a larger share of their planted area entirely.

The researchers also identified eight compound-risk years, 1984, 1990, 1998, 2001, 2004, 2009, 2013 and 2015, in which at least three of the four adverse rainfall conditions occurred simultaneously. In these years, modelled surface soil moisture anomalies averaged minus 0.970 standardised units compared with 0.250 in other years, and the sorghum harvested-area ratio anomaly averaged minus 0.956 versus 0.247. Both differences were highly significant. However, the classification proved sensitive to the thresholds used, with the exact set of flagged years changing under alternative percentile rules, even though the direction of the response remained stable. The authors frame this as an exploratory, impact-oriented classification rather than a formal definition of a multivariate climate extreme.

Two further findings complicate any simple narrative of worsening drought. First, a structural break detected after 1996 split the sorghum harvested-area record: the DRSRI association was strong before the break at minus 0.688 but weakened to minus 0.329 afterwards, while the seasonal rainfall association strengthened. Second, trend analysis after prewhitening and false-discovery-rate correction found no significant long-term change in any rainfall indicator between 1982 and 2020, yet the sorghum harvested-area ratio declined by about 9.2 percentage points per decade. This divergence suggests that the long-term deterioration in reported productive area was driven by factors other than rainfall, likely including the documented expansion of mechanised cultivation into more marginal clay plains, land degradation, and management or reporting changes.

The study is careful about its limits, and these caveats are instructive for anyone hoping to apply the approach elsewhere. The FLDAS soil moisture is modelled, not measured, and because CHIRPS rainfall contributes to the forcing of the FLDAS Noah model, the rainfall-moisture relationship represents internal consistency rather than independent validation. The Noah model also does not represent the cracking and preferential flow characteristic of the Vertisol clay soils that dominate Gedaref, adding further uncertainty. The state-wide averaging smooths local convective storms, meaning a wet day in the regional average does not guarantee rain on any particular field. The crop records, meanwhile, are annual administrative statistics lacking information on sowing dates, varieties, fertiliser use, pests, labour, or finance, all of which shape yields independently of rainfall.

For all these limitations, the work offers a practical template for drought monitoring in data-scarce drylands. The DRSRI can serve as a compact annual screening indicator, while its four components pinpoint the specific rainfall constraint operating in any given season. The authors suggest that early-season flags for delayed onset or extended dry spells could be overlaid with cropland and vegetation maps to direct field inspection where it is most needed. Future work, they argue, should move to pixel-level rainfall analysis within mapped cropland, compare satellite products against daily station observations, and incorporate independent soil moisture measurements. In a region where millions depend on rainfed sorghum and sesame, knowing not just how much rain fell but how it fell may prove the difference between anticipation and surprise.

Subject of Research: Within-season rainfall structure, modelled surface soil moisture and agricultural response in rainfed farming systems of eastern Sudan

Article Title: Rainfall structure and modelled surface soil moisture in relation to agricultural response in rainfed lands of eastern Sudan

Article References: Abaker, M., & Elfaki, J. (2026). Rainfall structure and modelled surface soil moisture in relation to agricultural response in rainfed lands of eastern Sudan. Discover Soil, 3(1), Article 181. https://doi.org/10.1007/s44378-026-00343-5

Image Credits: AI Generated

DOI: 10.1007/s44378-026-00343-5

Keywords: rainfall structure, soil moisture, rainfed agriculture, sorghum, sesame, drought, CHIRPS, FLDAS, Sudan, Gedaref, dry spells, compound climate risk