Special Issue on Data-Driven Approaches to Droughts

Since prehistoric times, water scarcity and droughts have posed serious challenges to civilizations. Drought events are regional, border-crossing phenomena that affect the management of transnational water resources and have a high potential for causing damage in terms of economic losses, ecological harm, and loss of life in affected areas. Droughts intensify environmental impairments. If droughts persist over long durations with sufficient severity, some consequences of droughts can be irreversible.

Recent droughts have been exceptional with increased severities and durations. Low precipitation has been exacerbated by high temperatures, increased evapotranspiration, and decreased runoff. Although these extremes resonate with projected impacts of greenhouse warming, it is unclear whether droughts can be attributed to increased greenhouse gases, natural climatic variability, or other confounding factors. Studies show that the recent warming is consistent with the Intergovernmental Panel on Climate Change Assessment Report 4 (IPCC AR4) projections of anthropogenic climate change. Scientists argue that the great North American droughts of the past 200 years were caused by sea surface temperature (SST) anomalies. However, the causes for such SST anomalies remain poorly understood, as do the potential impacts of increasing greenhouse gases on Pacific SSTs.

Climate models suggest that the current century will face more severe and prolonged droughts, with related ramifications such as increases in forest fires. Meanwhile, land-use changes and issues related to food security imply that water demands will remain constant or go up according to projections under even the mildest scenarios. Increasing population demands already raise very serious questions about future water sustainability. Concerns about economic viability, food safety, groundwater contamination, and health and safety have highlighted the need for better management systems. Simulations suggest that many parts of the United States will face sustainability challenges, with or without climate change. Although there remain relatively large uncertainties in projections of precipitation and temperature, the consensus is that global climate change due to increased greenhouse gases will exacerbate water deficiencies.

The Hydroclimate and Surface Water Technical Committees, within the Watershed Council of Environmental and Water Resources Institute of American Society of Civil Engineers, established a task committee with the goal of furthering our knowledge and understanding of the complex issues surrounding droughts. All drought-related studies—be they for characterization, analysis, or for forecasting purposes—rely on data. The data are often very heterogeneous and sparse in space, and usual record lengths are less than desirable for meaningful analyses. Different methods of reconstructions and interpolation techniques are used to generate data sets for analysis. This special issue, through a combination of forum articles, technical papers, case studies, and technical notes, responds to this challenge in an attempt to address drought-related issues.

A review of droughts was provided by Mishra and Singh (2010). Given the timing and severity of the recent 2012 drought that wreaked havoc in the Midwest, the article by Mallya et al. sets the stage by examining this drought through existing and new methods. The papers in this special issue can be broadly classified under the following themes.

Among the various techniques used for characterizing droughts, copulas are one of the emerging and popular methods for modeling the joint distribution of drought attributes. As noted by many researchers (e.g., Kao and Govindaraju 2010), copulas suffer from the curse of dimensionality. Special forms of copula families, such as Archimedean and metaelliptical ones, do offer attractive properties that have been used by numerous papers in this special issue. Notably, Madadgar and Moradkhani utilize trivariate copulas to quantify the dependence between drought attributes such as duration, severity, and intensity, in Oregon’s Upper Klamath River basin. The impact of climate change on future droughts was evaluated using five general circulation models (GCMs) under one emission scenario. The authors concluded that this region will likely see less intense droughts affected by climate change.

Even though some broad conclusions remain consistent, the role of large uncertainty among GCMs continues to be a challenge. Chen et al. utilize monthly standardized precipitation index (SPI) time series to jointly describe drought attributes of duration, severity, interval time, and minimum SPI values using simulated daily rainfall data for a 500 year period over the Upper Han River basin in China. The four-dimensional dependence in each drought state was used to develop drought probabilities and return periods. Khedun et al. also use copulas to study joint distribution of drought attributes by using the Noah land surface model-derived runoff values over Rio Grande/Río Bravo del Norte basin. Because observed streamflows may be regulated by anthropogenic activities, the authors contend that such data may not always be useful for climate change studies. Several climatic zones were identified, and a separate copula model was employed for each zone to relate, water deficit duration and severity and drought inter-arrival time from a 3 month standardized runoff index series. Univariate and conditional return periods were studied for their potential in long-term management of water resources for the region.

Maity et al. use the Plackett family of copulas (Kao and Govindaraju 2008) to determine the joint dependence between resilience, reliability, and vulnerability indexes defined by specifying thresholds on modeled soil moisture data over the Malaprabha River basin in India. Reliability and resilience were found to be highly correlated, so that the problem was reduced to the determination of the bivariate dependence between resilience and vulnerability. In contrast to the copula approach, Hao and Singh utilize entropy theory to explain the joint dependence between drought durations and severities. The theory relies on the use of constraints on the moments of the marginal and joint distribution of these attributes. The authors applied this method to monthly streamflow data from Brazos River at Waco, Texas, for hydrologic droughts, and compared the performance of the entropy-based method to the results obtained using copulas.

Is the increasing frequency and severity of droughts liable to pose more significant challenges to socioeconomic and environmental sustainability as we move further into the 21st century? The conclusions from several papers in this issue would suggest that the answer is a resounding yes. A popular area of study has been the examination of droughts in future climate projections using general circulation models (GCMs). However, researchers acknowledge the large variability between outputs of relevant hydrologic variables from different GCMs and suggest ways to deal with this inter-GCM variability to be adopted in drought studies. Using multiple low-flow (drought) indexes, the paper by Maldonado and Moglen evaluates the relative impacts of climate change and land-use change for the Occoquan Reservoir system in northern Virginia over the 21st century. Following statistical downscaling using quantile mapping, the authors used ensemble averaging over multiple GCMs, and find that land-use change (primarily in terms of percentage impervious areas) tends to dominate climate change impacts, until saturation conditions are achieved for land use. While climate change is expected to affect the timing and intensity of rainfall events, their influence is compensated by increased use of reclaimed water flows that builds in more resilience and decreases vulnerability to droughts.

Ojha et al. advocate a nested bias correction approach that allows outputs from individual GCMs to conform to observations in terms of selected moments at various temporal scales. The authors examine the performance of various bias-corrected GCMs, rank them based on different indexes, and use the bias-corrected future GCM outputs to study drought characteristics over various regions in India. Fakhri et al. highlight the sources of inter-GCM uncertainty on dry and wet spells of precipitation events in Shahrekord, Iran, with special emphasis on the role of downscaling in widening the uncertainty bands. Liu et al. employ SPI and PDSI to examine past (1900–2009) and future droughts in the Arkansas Red River basin. Climate projections from 16 GCMs, after bias correction and statistical downscaling, were employed in the study. Results from both indexes for A2 and A1B greenhouse emission scenarios suggest that droughts will be more frequent and severe in the second half of the 21st century, highlighting the need for new strategies to cope with droughts. Di Matteo et al. examine the impact of geologic settings on climate change influences on ground water. Using data from two springs, they conclude that springs connected to base flow may be more vulnerable to climatic changes.

While future projections of climate change have to rely on modeled values, it is not uncommon to use model results for characterizing droughts, especially in instances where long and continuous observations of necessary hydrologic variables are not available to conduct a robust statistical analysis. The versatile nature of the entropy approach was exploited by Rajsekhar et al. to conduct regionalization using simulated streamflow values from the variable infiltration capacity model. In this regard, the authors utilize drought severity and duration as attributes. They found that Texas could be partitioned into eight homogeneous regions for drought severity, and nine regions when considering drought duration. Once the regions were identified, the authors found that parts of Texas are more vulnerable to severe droughts, while the longest droughts seemed to have occurred in southern part of the state. Bellamy et al. augment streamflow records using tree ring data and examine the differences in long term drought characterizations resulting from use of short-term human induced flows versus long-term natural flows. They developed the magnitude-duration-frequency curves for the Upper Green River Basin, and evaluate the role of different kinds of data in determining the drought characteristics of droughts of this basin.

Several studies have utilized or designed new indexes for addressing particular drought-related questions. The case study by Karamouz et al. describes an integrated drought management system linked to a decision support system for the Aharchay River basin in Iran. The system integrates data collection and validation, rainfall-runoff prediction, reservoir operation, and drought analysis and monitoring. Hedging rules that conserve water and adjust reservoir operations during dry periods were devised. Socioeconomic aspects were incorporated in drought management utilizing an optimization model for land use planning, and water pricing was used to measure drought impacts. The authors demonstrate how the DSS could be used to plan for climate change affects.

Maity et al. also develop a drought management index (DMI) to study the relative propensity of different areas to droughts. The authors suggest that over five years of continuous data are needed to obtain robust parameters for constructing the DMI. Mallya et al. utilize hidden Markov models (HMMs) to achieve probabilistic classification of drought states using precipitation and streamflow data within Indiana. The authors propose an HMM-based drought index where the there are no fixed thresholds, and the data decide on the shape and location parameters of the emission distributions. The use of drought states as latent variables allows for new ways of interpreting and modeling droughts. Sharda et al. developed a community water deficit index (CDWI) for small- to mid-sized communities that are particularly vulnerable to climatic changes. The index relies on estimates of differences between water supply and demand, and reveals how proper management may reduce the severity of a drought. Applications to two mid-sized communities show the ability of this index to forecast droughts based on El Niño Southern Oscillation signals.

It is hoped that articles in this special issue will prove useful to the hydrologic community and aid further drought research.