Abstract: The increasing availability of web mapping tools creates new opportunities to bridge decision-makers’ climate information needs with technical capabilities. These new tools, however, raise familiar, unresolved issues related to cartographic representation. Using an on-line drought mapping tool, this study seeks to understand which spatial unit best meets the desire drought managers have for “local” information, their comprehension of uncertainties introduced in mapping information at local scales, and their willingness to trade off accuracy for information at a desired unit. We found that the most useful local map information includes regional context and boundaries which present their local area of interest. Even among this experienced, well-educated, professional group, those who had not taken a GIS or cartography class did not fully recognize the role of interpolation in creating and introducing uncertainty to some drought maps. Those who did recognize the uncertainty introduced by interpolation still strongly favored maps that provided estimated values for all areas vs. station point accuracy. Mapping poses a unique set of challenges to communicating risk and uncertainty. As more decision-support efforts incorporate web mapping, greater attention is needed to assure that users understand the tradeoffs between accuracy and precision in creating local information, the imprecision of boundaries, as well as the limits of forecasts. Clearly conveying spatial accuracy and uncertainty is a challenge that merits greater attention in using maps to communicate drought and other environmental risk information.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Greater web access and new software technologies have created opportunities to provide innovative decision support for water resource managers. The use of on-line, dynamic mapping is a clear example of a format that better meets user needs and preferences. Information once provided in graphs and tables can now be integrated with spatial data and represented in maps of watersheds, counties, climate divisions, and other spatial units. Further, these tools allow basic information to be recalculated and customized to meet decision-makers’ preferences.
As new capabilities and increased availability of technology create greater demand for cartographic decision-support tools, there is a corresponding need to increase our understanding of the types of spatial information users most need (Petersen, 2003). Making the most effective use of these visually oriented tools involves some understanding of both how the information is produced and how it is consumed. As is the case with many other types of information today, more is not necessarily always better. Some maps may offer greater utility than others but may require supplemental information to meet user needs. There is also a need to address cartographic literacy and communication to ensure that information is conveyed clearly. Lack of understanding can result in miscommunication as the images may provide an impression of certainty that is not always present in the underlying data (Monmonier, 2006). In addition, manipulation of data to meet user preferences for spatial scale introduces uncertainty that may be significant but not well understood. These issues can erode the effectiveness of decision-support tools, and as such, need to further consideration.
This paper reports on an effort to assess user preferences and understanding of the Dynamic Drought Index Tool (DDIT) for North and South Carolina, an online drought mapping tool. Building on findings from other efforts to provide climate information for decision support, we surveyed members of the South Carolina and Catawba-Wateree Drought Management Groups. Respondents were asked to consider a number of maps derived from the DDIT in order to refine our understanding of: what units best serve the desire for “local” information, users’ understanding of uncertainties introduced in mapping information at local scales, and their willingness to trade off accuracy for a desired scale. In the following sections, we first review research on decision makers’ needs for climate information and related work on cartographic communication. Then, we present our research methods followed by a discussion of results, findings, and implications.
Decision makers in climate sensitive sectors typically rely on information from a variety of sources (e.g., Pagano et al., 2001). Greater forecasting ability and awareness of climate impacts have fostered an increasing emphasis on bringing climate research findings into decision processes through the development of decision-support tools (e.g., CCSP, 2003; Cash et al., 2006).
The increased availability and interest in new information technologies, such as web communication and the development of dynamic modeling and mapping tools, is creating novel approaches with which to bridge decision-makers’ needs and abilities with scientific capabilities. Many decision-support tools have been developed by researchers to address the information “needs” of decision makers with the belief that they will contribute to better-informed decisions, but a lack of fit with the decision context and lack of communication with decision makers have meant that many such tools have not been broadly adopted (NRC, 1999).
The effectiveness of decision-support tools depends heavily on matching decision-makers’ needs, preferences, and abilities with scientific insights into climate processes (Jacobs and Pulwarty, 2003). Researchers working on decision making among water resource managers suggest that the full utility of such tools might not be realized without including decision makers throughout the development and implementation (Jacobs et al., 2005; Rayner et al., 2005). This “end-to-end” approach to improve decision support is being developed within several applied climate research organizations (e.g., Agrawala et al., 2001; Lemos and Morehouse, 2005).
Major factors influencing the use of climate information and forecasts include relevance of information to the decision context. Jacobs (2002, p. 9) highlighted the importance of “getting the scale and timing” as a key step in connecting science, policy, and decision-making. Decision-makers’ preferred scale is often assumed to be of greater local resolution. But, while that is true in some situations, its universality has not been established. Rather, understanding what may be an appropriate scale for various decisions requires understanding the context, relevance, and preference for drought information within differing decision contexts. The units and boundaries selected for a map can be significant considerations because they represent processes, jurisdictions, and authorities relevant to management responsibilities (Jacobs and Pulwarty, 2003). Additional factors affecting the use include the perceived risk; how information and uncertainty are communicated; real and perceived accuracy (skill) of information and forecasts; access to and awareness of information and resources; ability of users to accurately interpret information; and users’ level of trust in the information (Pulwarty and Redmond, 1997; NRC, 1999, 2006; Agrawala et al., 2001; Pagano et al., 2001; Hartmann et al., 2002; Changnon, 2004; Jacobs et al., 2005;O’Connor et al., 2005; Steinemann, 2006).
Within the broader realm of information, literacy challenges created by a growing information universe, geographic information literacy requires a distinct set of skills, such as evaluation of “accuracy” and understanding characteristics of spatial data and representations (Krygier and Peoples, 2003). Cartography provides a long tradition of research into map reading and the effectiveness of map symbolization and visualization (e.g., Dent, 1972; Robinson and Petchenik, 1975; MacEacheron, 1995; Lloyd, 1997). Previous research has demonstrated that how cartographic display can aid in decision-making by allowing users to visualize complex problems, reducing the time it takes to interpret data, and providing spatial context to information (Cooke, 1992; Crossland et al., 1995;Mowrer and Congalton, 2000). However, some cartographers are concerned that readers do not give maps the same critical scrutiny given to written text (e.g., Monmonier, 1996). While trust is an important part of making climate forecasts useful, Monmonier (1996, p. 1) stressed,
Not only is it easy to lie with maps, it’s essential. To portray meaningful relationships for a complex, three-dimensional world on flat sheet of paper or a video screen, a map must distort reality…. There’s no escape from the cartographic paradox: to present a useful and truthful picture, an accurate map must tell white lies.
For example, drought conditions vary with soil type and localized rainfall, but national drought-monitor maps generalize conditions into broad categories for regions rather than attempt to present all of those fine variations (NDMC, 2007). Presenting general patterns necessarily obscures the complexity of spatial variation, as well as limits imposed by data-sparse regions, missing data, or uncertainty in observations (Bayarjargal et al., 2006).
Effective cartographic communication requires understanding the needs and abilities of the map-reader by the mapmaker (Dent, 1972). Symbology and scale pose challenges, particularly when they are used to convey probabilistic information, such as weather forecasts (Jacobs and Pulwarty, 2003; NRC, 2006). Hartmann et al. (2002) reported that map contour lines used in seasonal climate forecasts maps led users to expect much more extreme values within bull’s eye areas. Map interpretation capacities can be limited, even among those with advanced training in environmental issues (e.g., Hartmann et al., 2002; Ishikawa et al., 2005; Broad et al., 2007). For instance, environmental science and policy masters students were not able to provide consistently correct interpretations of the probability forecasts in maps distributed internationally to nonspecialist and specialist audiences (Ishikawa et al., 2005). Broad et al. (2007) found that the public and media had multiple types of misinterpretations of the hurricane “Cone of Uncertainty” map, a widely distributed, educational product meant to inform public safety. Providing relevant information also requires consideration of the appropriate scale for mapping. In communicating information about environmental quality and risks to policy makers, Bartels and van Beurden (1998) called for greater attention to the fundamentals of cartographic development and design starting with modeling input and addressing map symbology, classes and class breaks, colors, scale and projection, and basic map features. They argue that symbology, classes, class breaks, and colors are particularly important in conveying environmental risk. Lindley and Crabbe (2004) focused on how air quality data are mapped, emphasizing the importance of classification systems and commenting briefly on the significance of interpolation strategies.
The related issue of introducing uncertainty through areal interpolation, Monmonier’s (1996, p. 1)“cartographic paradox,” is fundamental to the process of mapping climate information at the finer scales that decision makers desire. The degree of uncertainty varies with a number of factors including the quality of the monitoring network, the characteristics of the phenomena (e.g., convective rainfall or temperature), and surface variations (e.g., soil type and land cover). Most producers of climate information recognize this uncertainty (Guttman and Quayle, 1996; Rossel and Garbrecht, 1999), and some have quantified it (e.g., Rossel and Garbrecht, 2001). However, because dynamic mapping technologies have evolved rapidly (Abrams and Hall, 2006), and cartographic standards for conveying data uncertainty have been pushed into new technological realms, it is not clear that decision makers are aware of the uncertainties introduced by data aggregation and interpolation.
To the extent that climate information often provides probabilities of drought or other hazards, broad insights from the risk communication field are also relevant to mapping environmental data (NRC, 1996; NRC, 2006). The National Research Council report, “Improving Risk Communication” (NRC, 1989, p. 26) states that “risk communication is successful to the extent that it raises the level of understanding of relevant issues of actions and satisfies those involved that they are adequately informed within the limits of available knowledge” (emphasis in original). Related efforts in environmental risk communication further emphasize the complexity in the step of providing information and the variety of ways uncertainty influences risk perceptions (Chess et al., 2005; Tuler, 2006). These issues of cartographic and risk communication intersect with those of communicating climate information at several levels from the design and relevance of the overall message to the ease of interpretation.
Jacobs (2002, p. 10) asks, “If decision makers understood that there is a tradeoff between accuracy (how close to the truth you are) and precision (whether the information is specific to the area of interest), which would they choose?” For those who are aware of this tradeoff, what is the significance of scale vs. uncertainty in their decision-making? Greater understanding of the relationship between preferences for accuracy and desire for relevant information is important to furthering communication efforts, such as the development of the DDIT described below.
Developing a Dynamic Drought Index
The Carolina Integrated Sciences and Assessments (CISA) is one of eight Regional Integrated Science and Assessment (RISA) teams around the country that focus on addressing “complex climate-sensitive issues” within the context of fisheries, water resources, wildfire, public health, coastal, and agriculture sectors (NOAA, 2007). The CISA team investigates how decision makers in the Carolinas use climate information to manage water, and how such use can be expanded most beneficially to foster well-informed decisions. Through collaboration and dialog, a need for hydrologic and climatic information was identified in the Carolinas in response to implications from the drought of 1998-2002 (Carbone and Dow, 2005; O’Connor et al., 2005), the South Carolina Drought Act of 2000, information needs of water resource managers, and the Federal Energy Regulatory Commission (FERC) relicensing of dams. CISA responded to these needs by collaborating with the North and South Carolina State Climatologists and Duke Power, an applicant for FERC relicensing, to develop the DDIT for basins in North and South Carolina (Carbone et al., 2008).
The DDIT is a stakeholder-driven tool that allows users to visualize hydrologic and climatic information. The indices include PDSI; PHDI; Z-index; the modified PDSI (PMDI); the crop moisture index (CMI); the standardized precipitation index (SPI) for 1, 3, 6, 9, 12, and 24-months; and the Keetch-Byram Drought Index (KBDI). Also available are raw streamflow data aggregated at 7 and 14 days, and at 1, 3, 6, 12, and 24 months. In addition to raw data values, percentiles for each of these variables were computed from empirical cumulative distribution functions. This option mirrors that of NOAA’s “long- and short-term blends” and allows users the ability to blend drought indices using their own weighting scheme to match their particular sensitivities (Svoboda et al., 2002; Hayes et al., 2005; Steinemann et al., 2005). The tool first calculates indices and percentiles at individual stations, and then interpolates these values to a 4-km grid. Spatial averaging from the grid allows the drought measures to be displayed at different spatial units including the USGS 2-, 4-, 6-, and 8-digit hydrologic unit codes (HUCs), climate divisions, drought management areas, counties, or other relevant regions. These units differ in scale and management function. The data can be visualized in the form of a map, bar graph, or table. The DDIT is internet-based and dynamic, and provides great flexibility in allowing the user to select which indices, spatial units, and time frames they wish to display.
Within the context of the DDIT, users can specify that climate and hydrologic point data be aggregated into differing spatial units, such as counties, climate divisions, or hydrologic units, using the inverse distance weighted (IDW) method of areal interpolation and aggregation. All interpolation processes introduce uncertainty into mapped data that depend on the density and distance between data points. Several interpolation methods were examined in the development of this tool and IDW was selected because it performed as well or better than other techniques and its lower computational demands allowed for faster processing of queries (Rhee, 2007; Carbone et al., 2008). Interpolation can introduce errors in drought severity estimates. However, IDW if used for areal interpolation and aggregation produces “more consistent frequency distribution of drought severity categories between different spatial scales” than simple averaging and is thus the method utilized (Rhee, 2007; Rhee et al., 2008).
Prior to this study, during presentations of the tool to drought committees, some members indicated that the complexity and number of options in the tool met their requests, but were also daunting (Hope Mizzell, SC State Climatologist, August 30, 2007, personal communication). This analysis was designed to help the CISA team better understand user preferences for information from the DDIT and inform future web site design that accommodated interests in engaging at various levels of complexity. Given that uncertainty is a central issue in risk communication, all interpolation processes introduce uncertainty, and web mapping technology is relatively new, we also wanted to verify that the maps were understood as intended. These issues frame the analysis. Below we explore in more detail the useful scale of presentation, awareness of the cartographic tradeoff required to accommodate scale preferences, and how well the tool communicates uncertainty. We also assess the utility and desirability of other functions. For instance, the tool was updated to include the weekly drought indices, information that over 50% of users reported they would find more helpful than monthly indices.
We conducted a web-based survey of members of the South Carolina Drought Management Committee (n = 29) and of the Catawba-Wateree Drought Management Committee (n = 79). The survey included 32 questions about the usefulness of different forms of presentation formats (e.g., tables, graphs, box plots, and maps) as well as basic interpretation issues and demographic information. Here, we report on what scale and units best meet the desire for “local” information, users’ understanding of uncertainties introduced in mapping, and their willingness to trade off accuracy for a desired scale. The exact wording of the questions is provided in the text accompanying the tables. When the one overlapping member, six problem email addresses, and one member involved in pretesting were omitted, the total sample included 99 individuals. These individuals represent community water systems, industry, agriculture, conservationists, state environmental management agencies, and municipalities. The Catawba-Wateree committee includes representatives from both North and South Carolina. We focused on these groups because members had been given presentations on the DDIT in their committee meetings and because their professional and committee responsibilities place them within key user groups.
Following an initial letter from the Chair of each committee encouraging members to participate, we sent all members a cover letter with a link to the survey. A week later, one follow-up email was sent to encourage a higher response. We did not follow up further because we did not want to strain the good will in our long-term working relationship with these groups. We received a 34% response rate.
Background questions included in the survey showed respondents to be primarily in managerial positions (83%); experienced, with over 70% having 11 or more years in their current position; and well educated, 52% having college degrees and another 35% having graduate degrees as well. The respondent pool included representatives from water systems (32%), city or county government (26%), natural resource management agencies (19%), energy and industry (16%), and agriculture and forestry (6%).
Presenting Drought Data Through the DDIT
Working with decision makers to understand their preferences about the DDIT provided us insight into more general questions about using maps to provide relevant climate information and communicate uncertainty. As discussed above, what scale best serves interests in “local” information has not been explored deeply. We took first steps in this direction by seeking to clarify if decision makers prefer a time series of information for one place over a spatial statewide perspective and by investigating which areal units, smaller than a climate division, are preferred by decision makers.
In the second part of this section, we address the questions of awareness and uncertainties involved in producing maps of local information. We conclude with findings on the revealed preferences for accuracy vs. a desired scale. Transforming drought data from the station points where it is collected to map it at various local and other areal units places a new light on a familiar issue in communicating climate information – how to best present uncertainty. Mapping drought also requires that users of the information trade off accuracy for utility. Reluctance to use probabilistic data has been a barrier to forecast use, but awareness of uncertainties introduced through mapping observations has not been addressed.
Useful Scale for Local Information
To understand the relationship between users’ desire for local information and scale of presentations, we presented three forms of DDIT output (Figure 1) with the following introduction:
“The DDIT can display many different drought indices in three forms:
1In a map form at adjustable scales showing distribution and category of drought, but only for one point in time.
2In table form over a specified period of time that shows more precise monthly data, but only for one place.
3In bar graph form that shows the trend over time, but only for one place.” (emphasis in original)
Almost all users (91%) found the map, which provided a snapshot of county data and the regional context, to be a helpful display method. The bar graph and table, both of which provided a time series of drought severity for a place, were not as broadly endorsed, but were also helpful to the majority of respondents (77 and 57%, respectively).
When asked to choose which of the display methods was most helpful, 71% of respondents selected the map. An open-ended question revealed two major explanations for selecting a display method as most useful. Of those who selected the map (n = 24), 25% indicated that they preferred the map because the visualization of data was quick and easy to interpret. For instance, a respondent wrote, “the map is instantly recognizable. The graph or table takes a little longer to figure out where you are.” The remaining 75%, however, responded that the map was most helpful because it provided a regional context. Statements included comments such as, “it provides regional context instead of a single county’s status” and “I like to be able to see my particular county and the counties above me to evaluate drought conditions.” For the majority of these water managers, the regional context for local information provided by the map makes this form of display more helpful than the collection of local data presented in either form of time series.
The map presented in the first question used county boundaries, but in mapping applications, regional and local areas can be delimited by a variety of boundaries, ranging from city limits to counties, regulatory districts, or watersheds at multiple scales. For droughts, the point data collected at National Weather Service (NWS) stations is also important local information because it is at these points that the assessment of drought is most accurate, although still subject to measurement error. Each of these spatial units might be called “local,” but they do not serve all decision makers equally well. For instance, a watershed may include parts of several counties, and while drought data at the county level matches some agricultural and loss claims practices, it will not offer much insight into the condition of the watershed.
To understand which definition of “local” is more useful and to distinguish between the importance of geographic and operational scales, we asked water managers to consider the usefulness of four areal units. These are the climate division boundaries used by the National Climate Data Center in their climate forecast products and in the National Drought Monitor; county boundaries, which capture the operational scale of drought and disaster declarations; the eight-digit HUC representing watershed processes, and the NWS station locations, where the data are collected (Figure 2). The climate division, HUC, and NWS station maps included county boundaries in the background to provide a common and familiar frame of reference.
Over 60% identified the county unit aggregation as the most helpful; followed by 31% for the eight-digit HUC units. Six percent found the climate divisions the most helpful; none found the point values most helpful. In response to an open-ended question about the rationale for their choice, respondents who selected the county map commented on the ease of interpretation, familiarity with the county scale, and the desire for county level information. According to respondents, the county map was “easier to read,” for both respondents and the public, provided “recognizable jurisdiction, recognizable level of drought” and, these traits together, allowed users to “get more county specific information at a glance.” Those who selected the eight-digit HUC code map consistently emphasized the importance of basins, rivers, streams, and watershed processes to their management. For instance, one respondent stated, “It relates conditions to the more appropriate hydrological boundaries for water management.”
The multiple interests of most managers are apparent in their evaluation of the individual maps. Table 1 summarizes the water managers’ responses about the usefulness of each of the four types of maps. While the county map was more often selected as extremely useful, over a third of respondents identified the eight-digit HUC code and, to a lesser extent, the broader perspective offered by climate division maps as “very useful.” Managers indicated that mapping added value to the underlying NWS station data.
Table 1. Usefulness of Map Presentations.*
Notes: Percentages may not equal 100 due to rounding error. n varies because some respondents skipped the question.
*Question wording: “How useful would the level of detail found in map xx be for your water management decisions?”
NWS stations (n = 32***)
8-digit HUC (n = 30)
Counties (n = 32)
Climate Divisions (n = 32)
The respondents’ choices and explanations of what makes a map useful highlight the importance of boundaries in determining the usefulness of a map. Some boundaries are a better match to the needs of particular decisions or decision-making strategies, and institutional contexts, such as those that require public involvement or established reporting mechanisms.
This set of questions provides several insights into the attributes of helpful local information. A scale that allows local information to be framed in a regional context is a highly valued aspect of presenting information through maps. The way in which local information is demarcated also influences its usefulness. The basin mapping appeals to those who find information on hydrologic processes most useful. A larger percentage of respondents prefer county maps because they are easy to read, in part because the boundaries are familiar. The match of the boundaries with a governmental authority was also a factor. The preference for the county boundaries is consistent with findings in the cartographic literature (e.g., Lloyd, 1997). The concern for jurisdiction also reflects on the significance of governmental authorities and responsibilities. The most accurate data, displayed as station points, was not widely considered to be very useful. The following section pursues the issue of accuracy and precision in greater depth and addresses respondents’ awareness of uncertainties introduced through the interpolation process.
Awareness of Mapping Uncertainties
The issue of introducing uncertainty through the process of areal interpolation is unavoidable in the process of mapping and providing information on the local scale that decision makers’ desire. But, because GIS technologies are relatively new and there is not a common cartographic standard for conveying data uncertainty, it is not clear that decision makers have been introduced to or are aware of the uncertainties introduced in creating areal coverages. To the extent that they are aware of uncertainties, greater understanding of the relationship between preferences for accuracy and desire for precision in the form of local information is important to furthering communication efforts.
In the survey, we approached the understanding of areal interpolation in two ways. We showed respondents’ the same four maps presented earlier (Figure 2). All of these maps are based on the raw data presented in map A. To get a sense of respondents’ awareness of how interpolation might change the appearance of a map, we asked “Which map(s) do you think are based on the data presented in map A? (select all that apply).” A third of the respondents recognized the interpolation in all of the maps (Table 2). For the remaining two-thirds, that interpolation relationship was not consistently obvious.
Table 2. Percentage of Respondents Indicating That NWS Station Data Are the Basis for Other Maps.
Maps Based on NWS Station Data*
(n = 30)
*Question wording: “Which map(s) do you think are based on the data presented in map A?” (Select all that apply.)
Map B and map C (HUC and county)
Only map B (HUC)
Only map C (county)
Only map D (climate division)
Recognition of different types of interpolation algorithms may account for some responses; however, all of the respondents who selected the best answer had taken GIS or other classes that addressed principles of mapmaking. Only two people who had taken GIS classes selected an answer other than “all maps.”
To cross check our findings, we approached the issue of interpolation through a second question asking which of the four maps (Figure 2) presents the more accurate data. As our respondents are serving on drought management committees, receiving briefings and interpreting drought stages, we believe it reasonable to assume some familiarity with drought monitoring methods.
Nearly 40% correctly identified the more accurate representation of data presented in map A, the station point map. Eight of the 12 who responded correctly had taken GIS or related classes. Approximately 40% identified one of the other maps as most accurate, while nearly 20% believed that the accuracy is the same for all maps.
Taken in combination, responses to these two questions indicate that even experienced drought managers who have not taken GIS or cartography classes are less likely to recognize interpolation processes or how those transformations influence accuracy.
Whether an individual’s assessment of accuracy is correct or not, the question Jacobs (2002, p. 10) raised – how would decision makers trade off between perceived accuracy and precision – remains. Examining responses to the question about usefulness against responses to assessments of accuracy provides some insight into the tradeoff. Despite the better accuracy, only one of the 12 respondents who identified the NWS station data as the most accurate reported that they found it very or extremely useful. For the majority of this small subset, the greater spatial precision provided by interpolation outweighed the diminished accuracy. Maps that provided information for all areas, even more uncertain information, were valued more highly than maps that presented more accurate information and left large areas empty and forced users to consider themselves how station data might be relevant to their areas of interest.
Providing effective decision support for climate variability and change requires making scientific understanding available in a timely, relevant, credible, and comprehensible manner. Achieving these goals requires working with decision makers to understand the context and method for decision making. Drawing on a specialized, rather than a large respondent pool, this study contributes to understanding issues that need to be addressed in greater detail in those decision-maker scientist dialogs by providing further insight into (1) what units are useful to local drought information and (2) users’ awareness and acceptance of the uncertainty and tradeoffs involved in mapping climate information to a “local” scale.
Providing relevant local information is not a matter of scale alone. For many of these drought managers, scale which presents their local area of interest within a regional context is an important characteristic of a useful map. While maps were the most helpful form of presentation, there was a high level of support for the greater local information provided by time series bar graph or table. In addition, the boundaries and units used to guide local aggregation are important to a map’s usefulness. The responses indicated several different reasons for finding counties most useful, although a substantial proportion favored HUC units and found climate divisions moderately to very useful. The ability to accommodate these preferences is clearly an advantage of an on-line, dynamic mapping tool.
As more decision-support tools incorporate dynamic mapping, greater attention is needed to assure that cartographic processes and communication are comprehensible. Even among this experienced, well-educated, professional group, those who had not taken a GIS or cartography class did not fully recognize the role of interpolation in creating different maps and necessarily introducing uncertainty. The group without cartographic training was also much less likely to correctly identify the most accurate map. But, with one exception, that most accurate map was not seen as very or extremely useful by the group that correctly identified its accuracy (nor any of those without cartographic training). Precision, in the form of information specific to an area of interest, is clearly more important than accuracy for a single station for that subset of users. This finding underscores the need to measure the differences between point-derived vs. regionally derived values (Rossel and Garbrecht, 2001).
These findings have broad implications for future research and applications in climate-related decision support, and for other decision support efforts using mapping. While there has not been a great deal of cross fertilization among the risk communication, cartographic, and climate communities, some research on risk communication and mapping has highlighted the importance of symbology and classification systems (Bartels and van Beurden, 1998; Lindley and Crabbe, 2004). Understanding preferred scales for water resources and drought decisions will be significant in advancing the use of GIS and dynamic mapping for decision-making and guiding efforts to downscale model output of climate change scenarios.
In discussing the misunderstanding of the widely used hurricane “cone of uncertainty” graphic, Broad et al. (2007) called for increased use of social science understanding of risk communication to inform the development of graphic communication products. Greater use of dynamic, on-line mapping capabilities to provide the more contextualized types of local information that decision makers find most useful also requires more attention to the process of risk communication – entering into a two-way dialog to better understand user needs both in terms of type of information, assure that it is comprehensible, and communicate probabilistic information and associated uncertainties in a way that maintains credibility.
Mapping poses a unique set of challenges to communicating risk and uncertainty. The cartographic paradox – the need to tell “white lies” to present information clearly – presents a major challenge to assuring that relevant information remains credible and comprehensible. Assuring that users understand the tradeoff between accuracy and precision in creating local information, the imprecision of boundaries, as well as the limits of forecasts becomes more important as technical capabilities expand. This tension between desired presentation and credible presentation merits greater attention in efforts to communicate drought and other environmental risk information.
Pursuing this research in an end-to-end approach will involve working with decision makers to develop strategies to better explain and represent uncertainty. Efforts will consider the variety of initiatives within geospatial sciences on improving the communication of uncertainty. These include using color hue to indicate confidence, rollover information boxes, and flickering or toggling between maps (Monmonier, 2006). Another effort in decision support for climate information has created a required registration and tutorial process for users interested in use of a Forecast Evaluation Tool (CLIMAS, 2007). A different approach to matching the information presented to the audience is through nested design of web sites, somewhat like a search engine and an advanced search engine, which allows audiences to drill down to access greater levels of specificity, flexibility, and complexity desired in a decision-support tool.