Table 1.
Analysis summary.
Table 2.
Global climate model specifications and sensitivities.
Fig 1.
The topography of the CONUS in GFDL-ESM2M (left) and in CanESM2 (right).
The red boxes delineate the Northern Great Plains region adopted in this study, over which E0 is spatially averaged at each daily time step.
Fig 2.
Historical mean E0 across the different E0 formulations.
Climatological mean E0 (mm) across the CONUS for the MJJAS period in GFDL-ESM2M (first three plots in left column) and CanESM2 (right column) as estimated by the Penman-Monteith (first row), Hargreaves-Samani (second row), and Priestley-Taylor (third row) formulations for the 1976–2005 period. The bottom left plot shows observed mean MJJAS pan evaporation across the CONUS from 228 stations which had at least 20 years of data between 1950 and 2001.
Fig 3.
Historical coefficient of variation across the different E0 formulations.
E0 coefficient of variation (CV, times 100 for %) across the CONUS for MJJAS in GFDL-ESM2M (first three plots in left column) and CanESM2 (right column) as estimated by the Penman-Monteith (first row), Hargreaves-Samani (second row), and Priestley-Taylor (third row) formulations for the 1976–2005 period. The bottom left plot shows the CV of observed MJJAS pan evaporation across the CONUS from 228 stations which had at least 20 years of data between 1950 and 2001.
Fig 4.
Projected changes (%) in E0 by 2050 across the different E0 formulations.
Percent change in mean MJJAS E0 from Penman-Monteith (top row), Hargreaves-Samani (center), and Priestley-Taylor (bottom) formulations for GFDL-ESM2M (left) and CanESM2 (right) by 2050 (2036–2065) relative to the historical (1976–2005) period. Grid cells where the change is not statistically significant (i.e., p > 0.05) are masked out in white.
Fig 5.
21st century trends in E0 across the different E0 formulations.
Trends in MJJAS E0 projected from Penman-Monteith (green), Hargreaves-Samani (blue), and Priestley-Taylor (red) formulations driven by GFDL-ESM2M and CanESM2 data for the Northern Great Plains. The top row shows seasonal totals in mm, center row shows E0 anomalies as % of the 1976–2005 mean, and bottom row shows standardized E0 anomalies.
Fig 6.
Same as Fig 5 but using only the Penman-Monteith formulation and variations made to (a) radiation inputs (left column): Strict FAO56 with fixed albedo (0.23) and parameterized longwave radiation (green), fixed albedo and GCM-modeled longwave radiation (red), and GCM-modeled net shortwave and longwave radiation (blue); and (b) to crop selection and/or daily versus monthly timescales for wind speed (right column): Tall reference crop and daily wind (green), tall reference crop and monthly wind (brown), short reference crop and daily wind (blue), short reference crop and monthly wind (yellow).
Analysis based on GFDL-ESM2M only.
Fig 7.
21st century trends in EDDI and SPEI.
Comparison of 12-week EDDI and SPEI computed with the three E0 formulations for each day between 1950–2100 for GFDL-ESM2M and CanESM2 for the Northern Great Plains region. Daily EDDI or SPEI values are binned into specific percentile categories (spanning between driest and wettest categories) relative to the historical (1976–2005) distribution.
Fig 8.
Uncertainties in changing drought risk by 2050 based on E0 formulation and GCM selection.
Comparison of changes in 12-week EDDI and SPEI values for August 31 by 2050 (relative to the 1976–2005 mean) between the two GCMs based on the different E0 formulations considered in this study. Filled circles show the mean change, box plots show 25th, 50th and 75th percentiles, and whiskers show 5th and 95th percentiles. Confidence intervals shown here about the mean projected change are estimated based on Monte Carlo resampling. Positive changes in EDDI and negative changes in SPEI signify increases in drought intensity.