Hydrology, Cryosphere & Earth Surface Editors' Vox

The Challenges of Drought Prediction

Advances in dynamical modeling and the use of hybrid methods have improved drought prediction, but challenges still remain to improve the accuracy of drought forecasting.


In simple terms, drought is an absence of water, but it is actually a complex phenomenon and one of the most poorly understood natural hazards due to multiple causation factors operating on different temporal and spatial scales. A review article recently published in Reviews of Geophysics focused on the challenges of drought prediction. The editors asked the authors to explain different methods of drought prediction and describe what advances have been made in improving the accuracy of forecasts.

On a global scale, how significant is drought as a climatic hazard?

Drought is among the most disastrous natural hazards and occurs in virtually all climate regimes around the globe.

Drought in East Africa
A truck delivers water to a rural area in southern Ethiopia during the severe drought of 2010-2011. Credit: Oxfam East Africa (CC BY 2.0)

Examples of major drought events during the last decade include East Africa (2011), central United States (2012) and California (2012-2015).

Effects include crop yield losses, reduced water availability, and increased wildfire risk.

While some drought events cause billions of dollars of losses, others may result in famine, mass migration and a humanitarian crisis.

What are the main causes of drought?

Drought is generally related to a precipitation deficit, known as a ‘meteorological drought’. One driver of this is anomalies in sea surface temperature which effect large scale atmospheric circulation and, in turn, influence precipitation. At the local scale, high temperatures may lead to increased evaporation and decreased soil moisture, resulting in an ‘agricultural drought’. In addition, precipitation deficit or human activities (such as groundwater extraction) can cause water supplies from rivers, lakes, or groundwater to become low, known as a ‘hydrological drought.’

How can droughts be predicted?

Statistical and dynamical methods are two commonly used types of methods for drought prediction.

The statistical method is empirically based, using historical records without consideration of physical mechanisms, and relies on the relationship between the drought indicator to be predicted and influencing factors (or predictors). It is easy to implement but falls short in modeling complicated hydroclimatic interactions that may lead to drought.

Dynamical methods are based on state-of-the-art general circulation models (GCMs) or hydrologic models and are capable of modeling physical processes of weather and climate systems. They are adaptable to predict unprecedented conditions (such as extreme drought) but require intensive investment in the model development and parameterization.

Recently the hybrid statistical-dynamical method has been developed to merge the forecast from both methods and has shown promise in improving drought prediction in certain case studies.

What are some of the challenges with drought prediction?

We need to be able to make reliable predictions with a long lead time. Currently, the ability of dynamic models to predict precipitation diminishes quickly after two weeks due to the inherent chaotic nature of the atmospheric system. This poses a challenge to the prediction of meteorological drought, as well as agricultural and hydrological drought.

Another challenge is the prediction of drought in a changing environment due to climate change and human activities. For example, hydrological drought is closely related to human activities, such as irrigation, thus accurate prediction necessitates the modeling of human activities. Current drought prediction efforts have mostly focused on natural aspects, while research in integrating human aspects is limited but growing.

What additional research, data or modeling is needed to improve drought prediction?

We need a better understanding of drought mechanisms and predictability in different regions and seasons.

Field of wilted corn during drought in Texas
Better data can improve prediction and in turn help governments and communities mitigate against the effects of drought, such as the loss of corn crop in Texas during summer 2013. Credit: USDA/Bob Nichols

Apart from efforts in generating reliable data products (for example, data assimilation), models need to be refined to incorporate key processes of drought such as land-atmosphere interaction, temperature, soil moisture, and human activities.

Based on ensemble forecasts from statistical or dynamical models, efforts are also needed for processing these forecasts, such as selection of ensemble members and combination of different forecasts, with proper quantification of uncertainties.

Can advances in drought prediction be applied to predicting other natural hazards?

The methods used for drought prediction based on certain drought indicators can be used directly to predict other hazards related to hydroclimatic variables, such as flood forecasting, although a difference may exist in the properties of variables of interest; for example, low quantile is of interest for drought prediction while high quantile is of interest for flood prediction.

Advances in drought prediction provide opportunities for other types of hazards too. For example, drought is an important contributing factor to the occurrence of wildfire, thus accurate drought prediction may provide useful information for wildfire risk mitigation.

—Zengchao Hao, College of Water Sciences, Beijing Normal University; email: [email protected]

Citation: Hao, Z. (2018), The challenges of drought prediction, Eos, 99, https://doi.org/10.1029/2018EO092715. Published on 16 February 2018.
© 2018. The authors. CC BY-NC-ND 3.0