3.1. Deterministic Approach
Figure 2 presents the scores of the eleven indicators for the 48 contiguous U.S. states. The template at the bottom describes the position of the indicators around the polar axes and their definitions. The indicators corresponding to the Exposure, Sensitivity, and Adaptive Capacity are shown by brown, blue, and green colors, respectively. This figure provides information for a detailed comparison of drought indicators between the U.S. states. For example, the drought vulnerability in the state of Mississippi is mostly related to the Adaptive Capacity indicators, especially drought plan and GDP/Capita. On the other hand, the state of Nebraska is vulnerable to drought due to three indicators of sensitivity, namely agriculture, cattle, and renewable water resources.
Figure 3 shows the relative exposure, sensitivity, and adaptive capacity of each state, the sub-indices that produce the overall vulnerability score. The relative vulnerability of the states in the contiguous U.S. is shown in
Figure 4. The single most vulnerable state is found to be Oklahoma, while Delaware is the least vulnerable. A region that stands out as being less vulnerable than the rest of the country can be found in the northeast, extending into the eastern Mississippi Valley and the southeast (
Figure 4). The humid climate in this region results in ample water resources and few droughts, leading to a generally low exposure score (
Figure 3a). Sensitivity scores for this region are also below average for the country, largely due to the lack of extensive farming in these states, while the adaptive capacity scores of these states do not vary significantly compared to the rest of the states. The exemptions in this region are New Jersey, which has a very dense population, and New York, which has a very high vulnerability due to a relatively large exposure score originating from the highest percentage of protected aquatic ecosystems per unit area. Maine’s vulnerability is also ranked higher than its surrounding states. Just like New York, Maine has plenty of protected waters, as well as recreational lakes and a high dependency on hydropower.
The Southeast U.S. is slightly more drought-prone (
Figure 3a) than its northern counterpart, which can be attributed to higher evapotranspiration rates. Georgia is the most drought-prone state in the Southeast U.S., but given its high adaptive capacity, its overall vulnerability is ranked as low, on a par with its neighboring states (
Figure 4).
Oklahoma is ranked as the most drought-vulnerable state in the country in this study. This result is mainly attributable to limited adaptive capacity, including an outdated drought plan and limited irrigation possibilities, despite significant agricultural activities, including extensive cattle ranching (Oklahoma has 28 cattle/km2, third in the country behind Kansas (29) and Nebraska (34)). Montana is ranked as the second most vulnerable state; its vulnerability is associated mainly with a limited adaptive capacity and an above-average sensitivity originating from limited renewable water resources, extensive farming, and a significant amount of electricity being produced by hydropower.
In the Pacific Northwest, a region traditionally associated with ample precipitation, the state of Oregon stands out as being more vulnerable than its northern neighbor Washington, a difference mainly attributable to more frequent droughts in Oregon than in Washington. California, located south of Oregon, is subject to re-occurring multi-year droughts, but given its aggressive adaptation measures, it is less vulnerable than many of its neighboring states, and the second most drought-resilient state in the county.
The spectrum of vulnerability among the arid states in the southwestern part of the country spans from the “Very low” (CA) to the “High” (AZ, UT, NM) category. In this region, the physical water scarcity is profound, and the drought frequency is the highest in the country. The differences in vulnerability can hence mainly be attributed to human-related sensitivity and adaptive capacity factors. Despite being the second driest state, Nevada has adapted accordingly, to minimize dependence on limited water resources. The population density is among the lowest in the country and there are naturally few protected freshwater ecosystems, which serves to lower the overall exposure index. The sensitivity index is also low due to limited agricultural activities, hydropower, and recreational lakes, and its overall vulnerability is ranked as medium. In the same geographic cluster, Arizona, New Mexico and Utah are also found. Arizona is the single driest state in the contiguous U.S., its extremely limited water resources make it vulnerable to the exceptional water shortages associated with a drought. In New Mexico’s case, the high vulnerability score can be explained in part by extensive farming (52% of the state area is used for farming, the highest percentage in the region), in combination with limited irrigation capabilities. Utah, on the other hand, is ranked in the medium vulnerability category, a score mainly related to Utah’s limited agricultural activities in comparison to its neighboring states.
The division of the vulnerability into three sub-indices allows for analysis of the component(s) that most strongly contribute(s) to a state’s vulnerability score (
Figure 3). The states with the highest exposure index are located in the Northeast (NJ and NY). Being among the least drought-prone states in the study, their high exposure scores are the fruit of their dense population and their protected aquatic ecosystems being the most extensive in the analysis. This means that droughts are rare in this region, but when they occur, a larger population of people and protected plant and animal species are at risk of being adversely affected than if the drought happened anywhere else in the contiguous U.S.
In the center of the U.S., a cluster of states emerges for which the drought sensitivity is ranked as very high (
Figure 3b). In this cluster, the high sensitivity originates from their low renewable water resources in combination with extensive agricultural activities. On the other hand, low sensitivity scores across the Southeastern states are explained by ample water resources and, in comparison with other states, limited agricultural activities, hydropower, and recreational lakes.
3.2. Probabilistic Approach
Figure 5 shows example PDFs of DVIs generated for the states of Delaware, North Dakota, Illinois, and Oklahoma. The blue point refers to the deterministic DVI, and grey bars refer to 1000 DVIs calculated from a weighted average of 11 vulnerability indicators and 1000 samples of randomly uniform weights. The shape of PDFs provides additional information about the uncertainty of DVI in each state. For example, North Dakota shows the most uncertain results despite its lower DVIs compared to Oklahoma. The highly uncertain DVI for a given state shows the high sensitivity of that state to the weights of vulnerability indicators. These states should be judiciously managed because relying on their deterministic DVI can be misleading. On the other hand, states like Delaware show very limited uncertainty and relative vulnerability.
To investigate the relative vulnerability and uncertainty of the DVI components (namely Exposure, Sensitivity, and Adaptive Capacity), boxplots of distributions per category are generated (
Figure 6). The boxplots show that Adaptive Capacity is the most uncertain component of vulnerability. This uncertainty is mainly attributable to the largely binary adaptive capacity indicator “Drought Plan” (
Figure 1), where all but six states have the extreme values of zero (current drought plan exists) or one (no drought plan on record). Similar-looking box plots of high uncertainty also appear in the other categories, but at state level, and typically occur when a state has an extreme value for one of their indicators, the weight of which significantly changes the resulting DVI.
Figure 6 also eases the state-wise comparison of DVI per category and their uncertainties. For example, a comparison between New Jersey and New York shows that both states have high exposure with large uncertainty. This can be related to their geographical location, where both states are subject to the same climatic conditions. The uncertainty of sensitivity at both states is also a small value, while the value of sensitivity is higher in New York. However, the adaptive capacity of these two states is completely different. New Jersey shows high value with a small uncertainty, but New York has low value with large uncertainty. The uncertainty of categories also provides an additional message about the relative contribution of categories compared to
Figure 4. For instance, both
Figure 3 and
Figure 6 confirm that adaptive capacity is the main indicator explaining vulnerability in the state of Mississippi. However, looking closely,
Figure 6 reveals that adaptive capacity is a highly uncertain indicator in this state, and there are several occasions in which adaptive capacity can be a small value.
Figure 7 illustrates the probabilistic VCs for the U.S. states. The main advantage of this figure is that while it simplifies the interpretation of the vulnerability index by dividing it into five classes, it still accounts for the uncertainty of weights. The figure distinguishes states with clear VCs from those with highly uncertain VCs. For example, the states of Delaware, Massachusetts, Montana, and Oklahoma show very high certainty in their classifications. Some states, however, are more sensitive to the weights of indicators, and assigning a single VC to these states should be done with caution.
When comparing the maps in
Figure 3a–c, a few geographic patterns emerge. Typically the states with medium to very high exposure scores have low, or even very low sensitivity scores. This is probably due to long-term adjustment strategies, as historically, empirical data has tended to influence regional economic development and steer it away from sectors that are not profitable in the geographical (and climatological) setting. Industrialization and modern technology have somewhat eased these natural constraints, but their historical impact is still visible.
The extent of modern-day preparedness for drought as a hazard is reflected in
Figure 3c, showing adaptive capacity. This map mirrors some of the same patterns: states with medium to very high exposure scores typically have a competitive adaptive capacity score. The opposite is also true; it is among the states with the lowest exposure scores that the adaptive capacity also is lacking. This means that these states are very unlikely to be subject to a severe drought, but when they are, they have limited ability to respond to and recover from the hazard, making them vulnerable. The equal weighting of the indicators and sub-indices in this study fails to capture this vulnerability. An alternative weighting scheme could consider this and alter the results, but would introduce a different set of limitations, as the source of vulnerability is different in every state. Rather, policymakers and planners are encouraged to not limit their attention to their relative vulnerability ranking (
Figure 4) but also to consider their sensitivity and adaptive capacity score (
Figure 3) to find the source of their vulnerability and work on improving those areas. What is also important to consider is that the relative impact of drought varies by state. For example, a drought impacting Nebraska with its vast agricultural lands (92% of state area) could be argued to be more severe than a drought impacting a state like Alabama, which uses less than 26% of its land for agriculture; however, Alabama’s ability to cope and recover from drought is limited due to economic and policy restraints, and hence the impact on the agriculture, farmers, and the state as a whole might be more severe in Alabama.
Going beyond the individual state borders, the areal percentage of contiguous U.S. found in each drought vulnerability category is shown in
Figure 8. 25% of the area is in the low or very low category, 42% in the medium category, and 33% in the high or very high categories. It should be noted that the high and very high categories only contain seven and three states respectively, but these states constitute a significant portion of the contiguous U.S. As a comparison, the very low category, which covers 6% of the area, comparable to that of the highest vulnerability category, contains four states.
The study at hand gives a snapshot of current vulnerability to drought in the U.S. The indicators used show limited changes over time, but it is expected that the drought frequency and hence exposure to drought will change as a consequence of climate change. An analysis of future drought vulnerability would, apart from considering a changed exposure, also need to consider updated adaption strategies; however, such an analysis is beyond the scope of this research.