In eastern Africa, drought has occurred throughout known history with significant impacts on the agricultural sector and the economy, particularly through threats to food security. It is therefore important to examine the role of anthropogenic climate change in drought, particularly in the face of the large-scale droughts of 2010/11, 2014, and 2015 in Ethiopia and the 2016/17 drought in Somalia, Kenya, and parts of Ethiopia and surrounding countries, which have raised the spectre of climate change as a risk multiplier in the region.

Droughts are triggered and maintained by a number of factors and their interactions, including meteorological forcings and variability, soil and vegetation feedbacks, and human factors such as agricultural practices and management choices, including irrigation and grazing density . Accordingly, there are several definitions of drought in common use : meteorological drought (precipitation deficit), hydrological drought (low streamflow), agricultural drought (low soil moisture) and socioeconomic drought (including water supply and demand). This complexity of droughts poses challenges for their attribution. It is not straightforward to disentangle these interacting factors, but over long periods it may be possible to detect a climate change signal.

Previous attribution studies for eastern Africa have mainly focussed on meteorological drought drivers (precipitation deficit), with recent studies finding little or no change in the risk of low-precipitation periods due to anthropogenic climate change e.g.,. Some weather stations in eastern Africa have recorded a decrease in precipitation in recent years; however, climate models generally project an increase in mean precipitation but give conflicting results for the probability of very dry rainy seasons e.g.,. The reasons for the recent observed decrease in precipitation thus remain unclear, but the trend is within the large observed natural variability in the region, at least for the historical and current climate.

However, precipitation only covers one aspect of drought – that of the supply side of the water balance. The demand side is represented by actual evapotranspiration (ET), which is a function of moisture availability and evaporative demand. With increasing temperatures, there is an a priori assumption that rising evaporative demand will increase the demand side of the water balance and, all else equal, droughts will become more severe. However, this assumption is not based on analyses, which motivates an objective study. In this study, we aim to understand if, despite no evident trend in precipitation, increasing temperatures could be exacerbating drought.

In the current study we wish to align our drought definition as closely as possible with the major human impact of drought – the threat to food security. Across eastern Africa, the quality and quantity of food production for domestic consumption is intimately linked to agricultural conditions. We therefore use the agricultural definition of drought – low soil moisture – because soil moisture is a better indicator of crop health than precipitation alone and it embodies the net effect of the supply and demand side of the water balance in regions without irrigation. Whilst short term single-season drought episodes can be severe, we choose to analyse changes in drought on annual rather than sub-annual timescales because the worst crises in food security in this region have occurred with multiple-season droughts . We will also investigate the influence of the main meteorological drivers of soil moisture trends, i.e., precipitation and temperature. Ideally, we would study the influence of temperature on soil moisture via ET, however observational records are very limited in time and space and, as the spatial decorrelation lengths of ET are short, their informational value is limited. We therefore analyse evaporative demand (E0; sometimes also referred to as “potential evapotranspiration”, or PET, although this is strictly only one metric of E0). E0 is the amount of evaporation that would occur under prevailing meteorological conditions if an unlimited supply of water were available; in that sense, E0 measures the thirst of the atmosphere. E0 is calculable as a function of temperature, humidity, solar radiation, and wind speed. We use a variety of common parameterizations of E0 that includes both potential evapotranspiration and reference evapotranspiration and that ranges in physical representation and complexity from simple estimates based solely on temperature (the Hamon equation), through estimates that also include solar radiation as a driver (the Priestley–Taylor equation), and ultimately to fully physical estimates that further include humidity and wind speed as drivers (the Penman–Monteith equation). All necessary drivers are available for both observations and model simulations. In this manner, we bracket the complexity in E0 parameterizations in a convergence-of-evidence approach familiar to the drought-monitoring community.

We investigate E0 as a means to study the influence of temperature on soil moisture; however, for regions that are irrigated or where irrigation is being considered, E0 itself can be regarded as more relevant than soil moisture as a measure of drought tendency.

Whilst attribution studies specific to the eastern African region have not previously used soil moisture or E0 to explore drought, E0 has been used in various attribution or trend studies outside this region to explore, for example, the influence of climate change on the hydrologic cycle in China e.g.,, trends and variability at sites in western Africa , and compound events of low precipitation and high E0 in Europe .

Summarizing, the objectives of this study are first to consider the attribution question “do increasing global temperatures contribute to drier soils and thus exacerbate the risk of agricultural drought (low soil moisture) in eastern Africa?” and second to investigate if global-warming driven trends in precipitation or local temperature via E0 explain any emerging trend in agricultural drought. Our approach to attribution is comprised of the following steps: (1) definition of the study variables and explanation of the study regions; (2) description of observational data and detection of trends in observations; (3) model evaluation including description of the models; (4) attribution of trends in models; and (5) synthesis of the results. Assessments will be based on both observations and climate and hydrological model output on an annual timescale, between the years 1900 (to represent the pre-industrial era) and 2018. We will illustrate the method using examples of recent droughts in eastern Africa.

Section 2 of this paper presents the chosen study regions, followed by a description of the datasets used in the study. Section  describes the stepwise approach to attribution used, including assumptions and decisions made and illustrative examples. Section  synthesizes the results by region. Finally, Sects.  and present the discussion and conclusions.

In this section, we present the chosen study variables and study regions in eastern Africa and the datasets used to provide the variables to be analysed. Brief descriptions of the projects from which the datasets originate are provided in the Supplement.

In this section, we first describe the method for detection and attribution of trends in the four variables, including model validation and the synthesis of observational and model results. Section 3.2 describes the assumptions and decisions that are made concerning the data and model setup, and Sect. 3.3 provides an example of how the method is applied to real data.

In this section, we illustrate the synthesis method. Intermediate synthesis figures, which show not only the overall synthesis but also the results for individual models, are presented for the region SS for each of the four variables; the intermediate synthesis figures of all six regions can be found in the Supplement. Table  and Fig.  summarize final synthesized findings for all regions. Using both the intermediate and final synthesis results, we first draw conclusions based on different GCMs and hydrological models and then turn to conclusions for each variable.

First, we look for consistent behaviour in the trends from individual GCMs across the four variables. We note that the results from low-resolution GCMs do not consistently stand out compared to higher-resolution models and also overlap with observational uncertainty. Some general conclusions about the different GCMs are as follows: (i) for GCM-driven model runs with stronger positive trends in temperature, there is a tendency for the positive trends in E0 also to be stronger and vice versa for weaker trends; (ii) the uncertainty in precipitation trends is high compared to the trend magnitudes, which partially explains why a clear relation with soil moisture trends is not evident; and (iii) no clear relation between local temperature trends and soil moisture trends is evident.

Looking at the different hydrological models, we conclude that the trend in PCR-GLOBWB E0, which uses the Hamon E0 scheme that depends only on temperature, is generally higher than the trend in EC-Earth E0, which uses the more complex Penman–Monteith E0 scheme that additionally depends on humidity, wind speed, and solar radiation. Using this more complex scheme can influence the trend in soil moisture, especially in wetter regions.

The analyses of the individual model runs, stratifying by GCM or hydrological model, do not lead to a clear conclusion on the relation between the trends in soil moisture, precipitation, temperature, and E0. We therefore turn to the analysis of the synthesized values (see Table  and Fig.  for a summary of the outcome and Figs.  and S1 to S6 in the Supplement for synthesis diagrams). Table 5 summarizes the interpretation of the synthesized results shown in Fig. . The more the magenta bar is centred in the white box, the better the models agree with observations and the more we trust our attribution statement.

For soil moisture we find no significant synthesized trends: there is practically no change in region EE and no trend to a small, positive but non-significant trend in regions WE, NS, NK, CK and SS.

For precipitation, regions WE and NK show a positive but non-significant trend, although in region WE models and observations only partially overlap. In region NS there is a small positive trend, regions EE and CK show no trend (for EE only with partial overlap of models and observations), and region SS shows a negative, non-significant trend.

As expected from global climate change, the local annually averaged temperatures all have a significant positive trend, with best estimates between 1.0 and 1.3∘ per degree of GMST increase. Related to this, trends in E0 are also positive in four of the six regions but lower than for temperature and generally have larger confidence intervals. The regions NK and CK are the exceptions. Although weighted averages show positive trends, models show tendencies opposite to the observations. This incompatibility renders the results uncertain.

We can identify the following relationships between different variables: (i) precipitation trends have a (small) influence on soil moisture trends in regions WE, NS, and NK; (ii) in regions WE, EE, NS, NK, and CK, temperature and E0 have no discernible influence on soil moisture trends; (iii) in region SS, the non-significant negative trend in precipitation does not lead to lower soil moisture and neither do the trends in temperature or E0. While it would be desirable to link the overall findings to differences in regional climate, the differences in the synthesized results between regions are too small relative to confidence intervals to be able to say anything meaningful. It was nevertheless necessary to divide the study area into homogeneous regions, so that extremes experienced within each region are representative for that region and inhomogeneity is not influencing the location of the occurrence of extremes.

In this section, we interpret our results in the light of how our choices and assumptions may have influenced the outcomes and we compare them to previous studies.

Whilst it may be preferable to use soil moisture as a drought indicator, observations and simulations of precipitation are more reliable in this region . Precipitation has a large influence on agricultural droughts and is therefore appropriate to use in attribution studies in eastern Africa, supplementing the analysis of soil moisture. The outcome of previous studies that have focussed on precipitation deficits only e.g., are thus still relevant and compare well with our results, i.e., that no consistent significant trends in droughts are found. A comparison between seasonal cycles of the different variables (averaging the monthly means over recent decades) shows that the seasonal cycle of soil moisture is similar to that of precipitation in all six study regions. In contrast, the inverse seasonal cycle of temperature is not similar to that of soil moisture. Whether the E0 seasonal cycle reflects elements of the soil moisture cycle or not depends on the E0 scheme used: temperature- or radiation-based schemes show a seasonal cycle that is similar to that of temperature, whereas more advanced schemes reflect a mixture between the seasonal cycles of precipitation and temperature, as they also synthesize the seasonal cycle in humidity, which is strongly correlated to that of precipitation. We thus conclude that the influence of precipitation on soil moisture is higher than that of temperature or most E0 schemes. This is supported by the synthesized results that show negligible or no trends in soil moisture and precipitation, whereas the trends in temperature and E0 are strongly positive.

If temperature has an influence on trends in soil moisture (through E0), we expect to see that the positive trend in temperature is coupled to a drying soil moisture trend. As we average over the annual scale, we may miss parts of the season when this effect is strongest. Therefore, we selected a region and period outside the rainy season in which the seasonal peak in temperature corresponds to a dip in soil moisture (region CK, months February–March) to inspect sub-annual trends (not shown). Even then, we find that there is no negative trend in soil moisture accompanying the positive temperature trends.

We study drought trends on annual as opposed to sub-annual timescales, as long-term drought presents a greater risk for food security. On the annual timescale, we do not see strong explanatory relationships between the trends in the four studied variables (soil moisture, precipitation, temperature, and E0). To gain insight into the relationships between the variables, we additionally looked at correlations on a sub-annual timescale. Simple correlations between monthly soil moisture, precipitation, temperature, and E0 (not shown) support the conclusions of on the influence of precipitation and E0 on soil moisture at water limited sites in Europe. They found that at water-limited sites the influence of precipitation on soil moisture is much larger than that of temperature via E0. In our study, we find the same for the driest regions and the driest months in the wetter regions, and for the more temperature-based E0 schemes. This is presumably because temperature-based schemes (such as the Hamon approach) do not reflect land surface–atmosphere interactions as well as those that are also driven by humidity and wind speed (such as the Penman–Monteith approach) or, to a lesser degree, by radiation (such as the Priestley–Taylor approach).

Previous studies have shown that both the E0 scheme and their input data can have a large influence on E0 values . We confirm this using the CLM-ERA-PT (Priestley–Taylor), CLM-WFDEI-PT, and CLM-ERA-PM (Penman-Monteith) datasets (not shown). In our study regions, E0 values are consistently higher when using PM than when using PT. The differences in trends in E0 using ERA or WFDEI input or using PT or PM input are sometimes significant. However, comparing study regions, there is no consistency in the difference; in four out of the six regions the PM data shows a higher trend than the PT data, and in four out of the six regions WFDEI data shows a higher trend than the ERA data.

A study by discussed the possibility that climate model precipitation trends in East Africa are influenced by the inability of the models to reliably represent key physical processes. Rainy seasons in this region are governed by large-scale processes, such as El Niño-Southern Oscillation (ENSO) dynamics and the shifting of the Intertropical Convergence Zone (ITCZ). We view the tests we perform on seasonal cycle and frequency distributions, which provide some assurance that large-scale physical processes are reasonably well described, to be a minimum requirement for model validation. To improve the performance of models and to understand the discrepancies between models and observations, a much more thorough investigation into the models' representation of physical processes and feedbacks is required, such as demonstrated by and encouraged by the IMPALA (Improving Model Processes for African Climate) project (https://futureclimateafrica.org/project/impala/, last access: last access: 9 November 2018).

In the long term, a trend in E0 only has meaning for crop growth if there is water available for evapotranspiration. Much of eastern Africa is in a water-limited hydroclimate, requiring irrigation for crop growth. In irrigated areas within larger water-limited regions, the increased water availability shifts the local hydroclimate away from the surrounding water-limited regime towards a locally energy-limited regime. Positive trends in E0 seen in our analyses (especially if the variety of different schemes produces a robust E0 trend) could then signify a trend in actual ET and would therefore be accompanied by an increase in both irrigation water demand and, if that demand can be met, in crop growth. However, it should be noted that irrigation is not accounted for by the models or reanalysis datasets used here. Trends in E0 away from irrigated regions (i.e., in water-limited regions) will generally denote lower ET rates (through the complementary dynamics between E0 and ET that dominate in such regions), higher sensible heating of the atmosphere from a drier surface, and consequent greater drought exposure.

There are some factors influencing droughts and attribution results that are beyond the scope of this paper. For example, there is some evidence that warm spells are increasing in length, particularly in Ethiopia and the northern Somalia and Somaliland region , as is the number of consecutive dry days in some parts of eastern Africa, which may have an impact on drought length and increase the rapidity of onset and the intensity of drought .

Furthermore, it is likely that increasing temperatures have a negative impact on food security during droughts through, e.g., decreased immunity of livestock, or increased water demand for cooling and water supply and references therein. In addition, in regions suffering from recent meteorological drought, non-meteorological factors such as increasing population and land-use changes also play a role in worsening the declining vegetation conditions, even after precipitation returns to normal .

It is also still unknown how vegetation will respond to substantial increases in CO2 concentration. Two counteracting effects – physiological (restriction of stomatal openings leading to decreased evapotranspiration) and structural (increased leaf area leading to more stomata and increased evapotranspiration) responses – are expected, but their net effect is unknown e.g.,. There are indications that “dynamic vegetation models” that include these CO2 effects show a weaker response of drought to climate change . One of the hydrological models used in this study (LPJmL) uses dynamic vegetation modelling but there were no notable effects.

In this first multi-model, multi-method attribution study using several drought estimates in eastern Africa, we address the recurring question on whether increasing global temperatures exacerbate drought. Previous attribution studies for the eastern Africa region have examined drought from a meteorological perspective (precipitation deficit) and have found no clear trends above the noise of natural variability. In this study, we examined trends in eastern African drought from an agricultural perspective (soil moisture) as well as the meteorological perspective (precipitation, temperature, and E0 for six regions in eastern Africa. We also investigate whether global-warming-driven trends in these meteorological variables can be seen to contribute to trends towards drier soils. In this section, we draw conclusions for each variable in turn and make recommendations.

Of the four studied variables, soil moisture is most closely related to food security via crop health. In standardized soil moisture data, we found no discernible trends. The uncertainties in trends from model runs were found to be large, and there are no long observational runs available. This emphasizes that the use of an ensemble of models is imperative. Due to the large uncertainties in both soil moisture observations and simulations, we find no trend emerging from natural variability.

Precipitation was found to have a stronger influence than temperature or E0 on soil moisture variability, especially in the drier study regions (the significant positive trend in temperature is not reflected by a decrease in soil moisture). However, the confidence intervals on precipitation trend estimations are large and no clear trend is evident.

As expected from the increase in global temperatures, we find significant positive trends in local temperatures in all six regions. The synthesized trend is between 1.0 and 1.3 times the trend in GMST, which corresponds to a local temperature rise of 1.1 to 1.4∘ from pre-industrial times to 2018. However, the influence of this warming on annual soil moisture trends appears limited.

Soil moisture is more directly linked to E0 (via ET) than it is to temperature. Trends in E0 are predominantly positive, although in the regions NK and CK the uncertainty in this trend is large. This generally agrees with the positive trends in temperature. Similar to the results for temperature, we do not find strong relations between trends in E0 and soil moisture. Nevertheless, the results can still be of interest, both for irrigated regions where crop growth is limited only by meteorological conditions and for water-limited regions where the availability of water to evaporate greatly constrains forage growth. Due to large differences in results from different hydrological model runs, we recommend that E0 attribution analyses be carried out using an ensemble of hydrological models. These should use various (observational) input datasets and driving GCMs, although the decision to cover various E0 schemes is a trade-off between the desire to be representative of the uncertainty surrounding all approaches currently in use to not bias results towards a particular method (which is what we leant towards here by including, for example, the temperature-based Hamon approach) and the need to adhere to physical rigour in using the complete suite of drivers and an E0 parameterization that reflects all relevant dynamics (e.g., in the Penman–Monteith approach).

We conclude that although soil moisture is the preferred indicator of agricultural drought, we recommend that any soil moisture analysis be supplemented with precipitation analysis due to the superior reliability of precipitation measurements and the large influence of precipitation on drought in this region. Besides, soil moisture also has a physical lower limit: once the soil is dry it will remain dry. In water-limited regions an analysis of precipitation is thus a helpful addition.

Finally, communication of the uncertainties in the analyses of soil moisture, precipitation, temperature, and E0 (and any drought indicators) to policy makers, the media, and other stakeholders is crucial. Decision-makers need to properly weight and synthesize streams of potentially competing information from the variety of models, but without insight into the uncertainties in trends in the different drought indicators, they are missing this crucial information. They need to know how much the scientists trust their own conclusions, lest results be misinterpreted and conclusions become meaningless.