The North American Monsoon precipitation response to climate warming at convection-permitting scales

Abstract

Global climate models project a mean drying of the North American Monsoon (NAM) at grid spacings where convection is not explicitly resolved. We investigate the response of the NAM to climate warming at convection-permitting scales, utilizing output from a series of 13-year continental scale regional climate model runs performed at 4-km grid spacing. These include a control run forced by reanalysis and a pseudo global warming run that applies a mean perturbation to the boundaries derived from end-of-century changes projected by an ensemble of global climate models. NAM precipitation (June–September) shows averaged increases of 38.6% for portions of Arizona and western New Mexico. Over central Mexico, daily mean precipitation shows increases of 8.13%. Increases in rainfall amount are primarily associated with mean increases in precipitation intensity that overcome reductions in precipitation frequency. This increase in precipitation intensity is attributed to higher precipitable water concentrations, enhanced near-surface horizontal moisture fluxes off the Gulf of California, and stronger moisture flux convergence along the mountain peaks. Results from an offline semi-Lagrangian tracer model reveal that moisture contributions from the local land surface are significant in the control climate and produce proportionally less of the total precipitation in the warmed climate. Instead, contributions from nearby oceanic sources, notably the Gulf of California and Gulf of Mexico, increase.

Appendix A comparison against CMIP5

A direct comparison of the downscaled WRF CONUS precipitation response for the NAM region against the CMIP5 ensemble members used to derive the PGW perturbation is carried out here. We compare the PGW-CTR difference in total seasonal accumulated precipitation (June-September) averaged across the 2002–2013 period to the difference in total seasonal accumulated precipitation amongst 16 of the CMIP5 ensemble members (Table 1) used to construct the original CONUS PGW perturbation as outlined in Liu et al. (2017). Precipitation fields from the remaining 3 ensemble members not used here were unavailable at the time of analysis. We calculate the CMIP5 difference in accumulated seasonal precipitation by first regridding the accumulated precipitation field to a common 0.25$^{\circ }$ grid centered over the core NAM region and then taking the mean of the difference between the 1976–2005 and 2071–2100 periods. These periods are kept consistent to what was used to construct the end-of-century perturbation used to create the PGW initialized fields and lateral boundary conditions. The regridding of each individual CMIP5 ensemble member to a common grid is done using bilinear interpolation. We also regrid the CONUS WRF data from a 4 km native curvilinear grid to the same 0.25$^{\circ }$ grid using a first order conservative interpolation method.

The difference in accumulated seasonal precipitation for each individual CMIP5 member is shown in Fig. 16. Across most members, there is a persistent drying signature within close proximity to the SMO and its western foothills.

Comparison of the PGW-CTR difference in accumulated seasonal precipitation for CONUS (on both the native and 0.25$^{\circ }$ degree grid) and the end-of-century-historical difference across all of the selected CMIP5 members is shown in Fig. 17. When regridded onto a 0.25$^{\circ }$ grid, the magnitude of the CONUS precipitation change diminishes for most locations, particularly west of the SMO and along the Mogollon Rim in central Arizona (Fig. 17b). However, the overall sign of the response is well preserved, with simulated increases in seasonal precipitation still present across much of the domain. The CMIP5 mean ensemble difference contrasts with results from the downscaled CONUS simulations and shows a broad reduction in rainfall for much of the NAM domain with the strongest drying occurring along the western foothills of the SMO (Fig. 17c). The magnitude of this drying is small relative to the increase in precipitation found in the CONUS simulations, and suggests overall little change in accumulated NAM seasonal precipitation, similar to conclusions from earlier studies highlighting insignificant changes in total rainfall within CMIP5 owing to a redistribution of rainfall later in the season (Cook and Seager 2013; Seth et al. 2013; He et al. 2020; Wang et al. 2021; Hernandez and Chen 2022). Additionally, it is unclear whether the contrast in the sign of the response between the downscaled CONUS simulations and the CMIP5 ensemble used to generate the PGW perturbation can be attributed to a tendency for better resolved local features to act against the large-scale drying signature (similar to earlier work from Meyer and Jin (2017)), or whether the limitations of the PGW experiment in being unable to simulate large-scale circulation and storm track changes are responsible. More work directly comparing downscaled RCMs to GCMs is needed to explore this further.