Progress Report Material

Progress Report Material

Overview

To quantify climate change over the UAE and Arabian Gulf region, we are running simulations using the Weather Research and Forecasting Model (WRF, Skamarock et al. 2008). WRF is a regional climate model that can realistically simulate present and future climate at localized spatial scales. WRF reads in observations (in this case, a coarse-scale gridded observational dataset), and simulates weather and climate on a finer-scale grid. Our computational domains are shown in Figure 1. The outer 36-km grid spacing domain covers much of the eastern hemisphere. Nested inside this domain is a 12-km domain covering the Arabian Gulf region. The innermost 4-km nested domain covers the UAE and surrounding areas.

Figure 1. Domains used in WRF simulations.

Initial and lateral boundary conditions for the WRF simulations are taken from the ERA-Interim reanalysis (Dee et al. 2011), which is the European Centre for Medium Range Weather Forecasting (ECMWF)’s latest atmospheric reanalysis. ERA-Interim is considered the best atmospheric reanalysis available at the present time. Output is available at ~0.7° grid spacing on 38 vertical levels. Sea surface temperature (SST) data is provided by the National Oceanic and Atmospheric Administration (NOAA) Optimum Interpolation (OISST) product (Reynolds et al. 2007).

The WRF simulations feature 40 vertical levels and are reinitialized every two weeks, with four-dimensional data assimilation (FDDA, Stauffer and Seaman 1994) used on the 36-km domain to keep the model solution from diverging from reality. Physical parameterization schemes, which simulate model sub-grid scale processes in an empirical and statistical manner, include the WSM 5-class microphysics scheme, the RRTM longwave radiation scheme, Dudhia shortwave scheme, MM5 surface layer scheme, NOAH land surface model, YSU PBL scheme, and the Grell-Devenyi convective scheme (36-km and 12-km domains only).

Validation

To determine the fidelity of the WRF simulations in simulating climate over the UAE, we ran two one-month test simulations for July and December 1995. Rainfall estimates from WRF are compared with observations obtained from (source?). Figure 2a shows WRF-simulated rainfall and model biases from observations for July 1995. Overall, WRF underestimates rainfall at most sites, with larger biases over the Hajar Mountains. Figure 2b shows the same map but for December 1995. Positive model biases surround the coastal areas, with negative biases again west of the Hajar Mountains.

Figure 2. (a) WRF rainfall biases (color-coded dots), WRF rainfall (mm color shading), and WRF terrain height (gray contours) for July 1995.

(b) Same as (a) except for December 1995.

Figure 3 shows daily rainfall accumulation time series for a sample of locations during December 1995. It is encouraging that WRF captures most rainfall events throughout the month, even though the rainfall amounts are biased. A similar figure for July 1995 is not shown because the entire monthly rainfall occurred over a four-day period (July 22-26).

We are still working on adjusting model setup parameters (particularly the parameterization schemes) in order to reduce the precipitation biases shown in Figs. 2 and 3.

Figure 3. Daily rainfall tendencies for selected locations for December 1995. Observed tendencies in blue and WRFtendencies in red.

Modeling Climate Change – Strategy

To estimate the climate change signal over the UAE and Arabian Gulf region, we will use a dynamical downscaling approach. In other words, we will combine the climate change signal simulated by a global climate model with the high spatial resolution of a regional climate model (WRF). The advantages of using downscaled WRF output rather than raw global climate model output to simulate climate change over the UAE and Arabian Gulf region can be seen in Figure 4. This figure shows the terrain height and land/sea mask from the National Center for Atmospheric Research (NCAR) Community Climate System Model Version 4 (CCSM4, Gent et al. 2011), a global climate model used to support the IPCC AR4 (top) and from the 4-km WRF domain (bottom). CCSM4 does not adequately represent the Hajar Mountains, and represents much of the UAE as being in the ocean. In order to simulate any meteorological processes that are affected by mountainous terrain and coastal effects, it is thus necessary to downscale climate model output as we are proposing.

Figure 4. Terrain height (m, color scale at bottom) and land/sea mask for CCSM4 (top) and 4-km WRF (bottom). Actual coastlines and political boundaries shown in black.

To perform the downscaling, we will run WRF for the entire 1980-2100 time period, at ~12 km grid spacing. Initial and lateral boundary conditions will come fromCCSM4, which has 1.25° latitude / 0.942° grid spacing and 26 vertical levels, with output available at 6-hourly frequency. Input from the CCSM4 Historical simulation would be used for 1980-2005 as a present-day baseline. We then use input from the RCP8.5 simulation for 2006-2100. Additionally, we “branch off” a simulation using RCP4.5 input for 2051-2100. Our ground-truth simulation is WRF using input from the ERA-Interim reanalysis, 1980-2005. Within each of the 1981-2005 Historical time period, 2006-2050 RCP8.5 time period, 2051-2100 RCP4.5 time period, and 2051-2100 RCP8.5 time period, we select ~5 year time slices where we run WRF at 4 km grid spacing. This output is used to statistically downscale WRF to 4 km grid spacing for near-surface temperature and precipitation at monthly intervals for the entire 1980-2100 time period (with RCP4.5/8.5 branches at 2050). Figure 5 shows a schematic of the simulation time tracks.

Figure 5. Schematic of simulations run with WRF and forcing scenarios used. Gray boxes represent ~5 year time slices where 4-km dynamic downscaling is performed. Ordinate is arbitrary and not to scale.

The climate change estimates simulated from WRF are based on CCSM4 (through initial and lateral boundary conditions), and because CCSM4 contains regional-scale biases due to the coarse spatial resolution, these climate model biases infiltrate the WRF simulations, leading to unsatisfactory results. To remedy these biases, it is common to bias correct the climate model input to regional-scale models like WRF. Here we will use a recently-developed bias correction method that corrects for the mean bias from CCSM4, but allows synoptic-scale and climate-scale variability to change in the future (Xu and Yang 2012; Done et al. 2013; Bruyére et al. 2013). Revised CCSM4 output, correcting for the mean bias, is produced by summing the average 6-hourly annual cycle (the Reynolds averaged mean term) from ERA-Interim (1981-2005) and a perturbation term (the Reynolds averaged eddy term) from CCSM4:

whereoverbar terms are the mean climatology, primed terms are perturbations from the climatology, and CCSMR is the revised (bias-corrected) CCSM4 output. We then run the 12-km WRF simulations for 1980-2100, as described above, using the corrected CCSM4 output as input. To finally downscale to 4-km grid spacing, we will use a hybrid dynamical-statistical downscaling approach, where we dynamically downscale to 4 km time slices of approximately 5 years for the four time periods of interest (Historical 1981-2005, RCP8.5 2006-2050, RCP4.5 2051-2100, and RCP8.5 2051-2100). We then use a to-be-determined statistical downscaling method to downscale the entire 1981-2100 time period (with RCP4.5/8.5 branching at 2050) to 4 km. This would be done for near-surface air temperature and precipitation at monthly timescales. This final output at 4-km grid spacing for both present day and future time periods will be used to assess regional climate change over the UAE and Arabian Gulf regions.

References

Bruyére CL, Done JM, Holland GJ, Fredrick S. 2013.Bias corrections of global models for regional climate simulations of high-impact weather.Clim. Dyn., doi:10.1007/s00382-013-2011-6.

Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda MA, Balsamo G, Bauer P, Bechtold P, Beljaars ACM, van de Berg L, Bidlot J, Bormann N, Delsol C, Dragani R, Fuentes M, Geer AJ, Haimberger L, Healy SB, Hersbach H, Hólm EV, Isaksen L, Kållberg P, Köhler M, Matricardi M, McNally AP, Monge-Sanz BM, Morcrette JJ, Park BK, Peubey C, de Rosnay P, Tavolato C, Thépaut JN, Vitart F. 2011. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc.137: 553-597.

Done JM, Holland GJ, Bruyére CL, Leung LR, Suzuki-Parker A. 2013. Modeling high-impact weather and climate: lessons from a tropical cyclone perspective.Climatic Change, doi:10.1007/s10584-013-0954-6.

Gent PR, Danabasoglu G, Donner LJ, Holland MM, Hunke EC, Jayne SR, Lawrence DM, Neale RB, Rasch PJ, Vertenstein M, Worley PH, Yang Z-L, Zhang M. 2011.The Community Climate System Model version 4.J. Climate,24: 4973-4991, doi: 10.1175/2011JCLI4083.1.

Reynolds RW, Smith TM, Liu C, Chelton DB, Casey KS, and Schlax MG. 2007. Daily high-resolution-blended analyses for sea surface temperature. J. Climate20: 5473-5496.

Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Duda M, Huang XY, Wang W, Powers JG. 2008. `A description of the Advanced Research WRF Version 3`.NCAR Tech Notes-475+ STR.

Stauffer DR, Seaman NL. 1994. Multiscale four-dimensional data assimilation. J. Appl. Meteorol.33: 416-434.

Xu Z, Yang Z-L. 2012. An improved dynamical downscaling method with GCM bias corrections and its validation with 30 years of climate simulations. J. Climate 25: 6271-6286, doi:10.1175/JCLI-D-12-00005.1.

1