Drought data for analysis

In order to construct reliable multivariate statistical models of joint drought deficit distributions, large amounts of sufficiently long historic observations are desirable. For instance, a 50-year minimum recording length is adopted by National Weather Service (NWS) in performing at-site rainfall frequency analysis (Bonnin et al., 2004). This 50-year standard was also chosen in this study as a minimum requirement. Considering the nature of droughts, stations were considered acceptable if monthly precipitation data had an 80-year minimum recording length, and monthly streamflow data contained 50-years minimum recording length.

Precipitation records were obtained from the daily surface dataset (TD 3200) of cooperative stations (COOP) from National Climate Data Center (NCDC). After data processing (such as combination of nearby stations), a total number of 73 stations with record lengths greater than 80 years were obtained. Locations of the rainfall gauging stations are shown below. Monthly precipitation was computed based on aggregated daily values. In cases where data was missing for the entire month, it was replaced by the historic mean of that specific month (i.e., assuming moisture status of that unknown month to be neither wet nor dry).

For streamflow, the United States Geological Survey (USGS) daily streamflow dataset was utilized in the present study. Unlike precipitation, streamflow data are subjected to human interference, and therefore data contain both natural and regulated flows. This practice does not cause serious problems in flood frequency analysis (flows are much larger), but it is expected to result in more errors for low flow (drought) conditions. Therefore, only unregulated stations were included to ensure an unbiased analysis. After imposing the 50-year record length requirement, a total of 36 unregulated stations were available for the study area (shown below). Daily mean flow data were collected and processed to form monthly mean discharges.

The two tools that display these historic data in a user friendly way are - JDI Viewer and Water Deficit Viewer. Following are a few words about these two tools. JDI Viewer: The joint deficit index (JDI) is a statistical tool that encapsulates the joint probability distribution of droughts over a 12-month period (in steps of one month). This viewer shows the evolution of this index in both space and time. Users are asked to provide:
  • A starting time (year and month)
  • An ending time (year and month)
The viewer then provides static views of the how the JDI evolves over Indiana over the user-specified time period. This viewer is useful for advanced users are researchers who are interested in performing higher-order computations about the various statistics of precipitation. Research purposes of this viewer include
  • characterization of spatio-temporal extents of droughts
  • identification of drought triggers
  • association with other hydrologic variables
Water Deficit Viewer The water deficit viewer is developed from the JDI as an example application. The viewer is useful for visualizing the changing pattern of water deficit over the state of Indiana for a specified time horizon (in months). Users are prompted for the following information:
  1. A starting time (year and month)
  2. An ending time (year and month)
  3. The time horizon (time in months, ranging from 1 to 12 months)
  4. Amount of water deficit or probability of recovery
Two maps are created simultaneously for each month. The first shows the expected water shortage as a function of time over the time horizon selected by the user. The second map shows the chance of recovering from a drought during this time window. These maps change with each month as observed precipitation causes us to update these estimates.

This viewer is useful for a variety of users who are interested in short term (one to several months) projection of water deficit estimates over a region. It is useful for impact studies. Some of the questions that can be answered from this information are: How should farmers plan for irrigation scheduling? What kind of a crop yield is likely? Are we going to have enough water to meet drinking water demands, industry needs, or recreational uses?