Data Imputation, Scaling, and Homogenization (DISH): Drought and Employment Nexus

Abstract:

Quantifying the nexus between natural resource availability and its impact on the socioeconomic status of communities is hindered by relevant data and information being scattered across disparate data sources in heterogenous spatiotemporal resolutions and formats. Tools to ingest these data into common frameworks (i.e. data wrangling and data cleansing) would help us understand the extent of resource disparity and the interventions needed to reduce them. The Data Imputation, Scaling, and Homogenization (DISH) project, part of UT Austin’s Planet Texas 2050 program, aims to create a set of robust, widely applicable tools to reduce data friction and improve productivity. To test these tools, we combine unemployment and drought index data from 2010 through 2015 for the Brazos River basin, Texas to assess when, where, and to what extent drought conditions in Texas contributed to unemployment across a wide spectrum of communities and stakeholders. This was done by integrating drought index shapefiles from the U.S. Drought Monitor with population and employment estimates from the American Community Survey (ACS) and monthly employment statistics available from the U.S. Census Bureau and the U.S. Bureau of Labor Statistics (BLS), respectively. Spatially, we apply statistical and machine learning approaches to perform up/downscaling and data fusion to link drought indexes with census tracts in shapefile and gridded forms. Temporally, we homogenize ACS and BLS employment data into a monthly timestep at the census tract scale using the “disaggregated census share method” developed and used by the BLS. From this example, we show how harmonized datasets can aid researchers working at the nexus of socioeconomics and water, or virtually any other natural resource-related stock.