On HRM3 (a.k.a. HadRM3P, a.k.a. PRECIS) North American simulations Bill Gutowski (on behalf of Richard Jones, Simon Tucker and Wilfran Moufouma-Okia) © Crown copyright Met Office

Contents Results from the HadCM3-driven climate (change) simulation Results from the NCEP-driven climate simulation Comparison of NCEP and ERA-Interim simulations and boundary conditions Concluding remarks © Crown copyright Met Office

Results from the HadCM3-driven climate simulations © Crown copyright Met Office

Experimental Set-up The 50km resolution HadRM3P was nested within the HadCM3 GCM, and run in two time-slices: 1968-2000 and 2038-2070 under SRES A2 emission scenario Domain size is 171x146 - the interior domain corresponds to the NARCCAP region The outer 8 grid boxes were discarded along with the first two years of the model output data whilst the interior solution was spinning up The simulations provided 30 year baseline and future periods for analysis © Crown copyright Met Office

How to assess the RCM performance in simulating the current climate HadRM3P HadCM3 Observations consistency realism Compare like with like RCM only has skill at spatial scales resolved by its grid Aggregate or interpolate RCM or observed data Can not compare individual RCM years with corresponding observed years when GCM-driven (same reason as with GCM) Errors are a combination of three errors: 1) Physical errors in the GCM affecting the LBCs 2) RCM/GCM consistency errors 3) Physical errors in the RCM © Crown copyright Met Office

Large-scale consistency between HadRM3P and HadCM3 Mean sea level pressure 700 hPa advection of humidity HadRM3P HadCM3 These results are computed for mean JJA 1971-2000, and on the GCM grid © Crown copyright Met Office

The realism of HadRM3P and HadCM3 Models realistically capture the mean winter precipitation Similarities between HadRM3P and HadCM3 biases Important differences occurs in areas of complex orography Mean DJF 1971-2000 precipitation and anomalies © Crown copyright Met Office

Summer mean precipitation is also well captured by the two models HadRM3P biases are largely reduced over domain, Mean JJA 1971-2000 precipitation and anomalies © Crown copyright Met Office

There is a consistency in the anticipated temperature change signal of HadRM3PvsHadCM3 RCM GCM DJF MAM JJA Projected seasonal changes in temperature, between 1971-2000 and 2041-2070 © Crown copyright Met Office

Results from the NCEP and ERA-Interim driven climate simulation © Crown copyright Met Office

HadRM3P precipitation compared to CRU climatology – NCEP boundary conditions Winter Summer CRU NCEP-driven RCM RCM bias © Crown copyright Met Office 1981-2000 Mean seasonal precipitation and biases

HadRM3P JJA/ DJF temperature biases: Top – ERA-In boundaries Bottom – NCEP boundaries Bias pattern similar – magnitude much greater with NCEP boundaries © Crown copyright Met Office

Difference in DJF surface temperature simulation when using NCEP rather than ERA-Interim boundaries HadRM3P DJF temperature bias reduced by >5K when using ERA-Interim boundary conditions Similar signal seen in summer © Crown copyright Met Office

NCEP–ERA-Interim temperature (left and middle) and Rh 1000hPa 200hPa 1000hPa NCEP–ERA-Interim temperature (left and middle) and Rh 925hPa 150hPa 925hPa NCEP warmer and moister on western inflow boundary 850hPa 100hPa 850hPa © Crown copyright Met Office

Summary of reanalysis downscaling analysis HadRM3P-NCEP has by far the largest seasonal temperature biases (see BAMS article) HadRM3P has no systematic tendency for significant positive temperatures biases (Xu et el., 2006 over China, Marengo et al., 2009 over South America, Kamga et Buscalet, 2006, over Africa) Previous studies have used ECMWF boundary conditions, ERA-15, ERA-40 or ERA-Interim reanalyses HadRM3P-ERA-Interim data produces significantly lower (5K) temperature biases - and similar in magnitude to the other RCMs. © Crown copyright Met Office

Some reasons for differences in the downscaled reanalyses At the boundary of the NARCCAP domain the NCEP data are both warmer and moister in the lower troposphere and in the upper troposphere/stratosphere – directly increasing temperature and increasing downward longwave radaiation (from higher atmospheric humidities and, in winter, by increased cloud cover) In winter the warmer temperatures lead to reduced snow cover thus enhanced solar radiation absorption at the surface and in spring/summer lead to drier soils, lower evaporative cooling and lower cloud amounts enhancing surface solar radiation © Crown copyright Met Office

Concluding remarks on GCM downscaling HadRM3P model successfully downscaled the climate projection from HadCM3 over North America HadRM3P-HadCM3 simulates realistically the mean surface features of the twentieth century climate The regional climate shows a good agreement with the large-scale driving fields and adds value to the GCM © Crown copyright Met Office

Concluding remarks on reanalysis downscaling It is important not to base strong conclusions on a model's performance (either good or bad) on the basis of simulations driven by a single set of reanalyses. The quality of a reanalysis should be assessed before using it to validate an RCM (Cerezo-Mota et al concluded the NCEP reanalyses were deficient for assessing RCM reliability in southern N. America) HadRM3P is the only non-North American model in NARCCAP and would have been reformulated if it had been developed using NCEP boundary conditions © Crown copyright Met Office