Supplementary Material for: Coordinated Global and Regional Climate Modelling

Size: px

Start display at page:

Download "Supplementary Material for: Coordinated Global and Regional Climate Modelling"

Ada Bruce
5 years ago
Views:

1 1 Supplementary Material for: Coordinated Global and Regional Climate Modelling 2 a. CanRCM4 NARCCAP Analysis As CanRCM4 is a new regional model that employs a physics package from a global model, it is important to provide an evaluation of its performance. The most straightforward way to do this is to compare its performance against the performance of RCMs documented in an established intercomparison project. Here we consider the performance of CanRCM4 against RCMs that participated in the North American Regional Climate Change Assessment Program (NARCCAP) (Mearns et al. 2009). Specifically, a CanRCM4 evaluation run is performed for the North America domain at 0.44 degree resolution for the period driven by the NCEP reanalysis R-2 (Kanamitsu et al. 2002). Following the study of Mearns et al. (2012), the modelled near surface (2 m) temperature (Tas) and total precipitation rate (Pr) in CanRCM4 is validated against the version 2.10 Climatic Research Unit (CRU2) monthly time series of temperature and precipitation (Mitchell 2008; Mitchell and Jones 2005), which includes elevation correction for temperature and precipitation (New et al. 2000). Figures S1 and S2 present June-July-August () and December-January-February () mean Tas biases relative to the CRU2 dataset for spectrally nudged CanRCM4 (panel a), spectrally unnudged CanRCM4 (panel b), and six other regional climate models (panels c h). Details related to the six additional RCMs, as well as a more in-depth analysis of their performance, is provided by Mearns et al. (2012). From Figs. S1 and S2 it can be seen that CanRCM4 displays a warm bias in the south central U.S. in and in the continent interior in, and a cold bias along the west coast in both seasons. The impact of spectral nudging in CanRCM4 causes a slight reduction in the amplitude of biases (particularly in ) but has little impact on their structure and distribution. This Tas bias is consistent with bias present in CanRCM4 s parent model CanAM4 when run in 1

2 the AMIP-type mode (i.e. with prescribed observed SSTs and sea ice distribution). It is apparent that CanRCM4 temperature biases displayed in Figs. S1 and S2 are reasonable relative to the other six RCMs considered. CanRCM4 temperature biases are further evaluated for a number of North America sub-regions, which are displayed in Fig. S3. Figure S4 presents a graphical summary of spatially averaged seasonal and annual mean Tas biases for each of these sub-regions for the considered RCMs together with the overall North America bias (labelled as All Regions ). The results indicate that CanRCM4 regional temperature biases generally fall within the inter-model spread of other NAR- CCAP RCMs and are not strongly influenced by spectral nudging. Figures S5 and S6 present and mean precipitation rate (Pr) biases relative to the CRU2 dataset for spectrally nudged CanRCM4 (panel a), spectrally unnudged CanRCM4 (panel b), and the six other RCMs (panels c h). CanRCM4 displays a dry bias in south central U.S. in and near the Gulf of Mexico in, and a wet bias along the west coast in both seasons and the east coast in, which display little sensitivity to spectral nudging. The CanRCM4 precipitation biases displayed in Figs. S5 and S6 are reasonable relative to the six other RCMs considered. The precipitation biases in CanRCM4 are also evaluated for the same sub-regions presented in Fig. S3. Their graphical summary is presented in Fig. S7 in the same format as for Tas regional biases in Fig. S4. Again, it is apparent that CanRCM4 regional precipitation biases generally fall within the range of those for the other six models. CanRCM4 suffers from the same systematic wet bias in the North-West and dry bias in the South-East as is displayed by the other NARCCAP RCMs. Overall, the analysis of Tas and Pr biases indicates that the performance of CanRCM4 generally falls within the range of the RCMs that participated in the NARCCAP. 2

3 46 b. CanRCM4/CanESM2 Appreciable Difference Analysis - No Spectral Nudging As an initial test of the potential influence of spectral nudging in CanRCM4 future projections, an additional 5 historical simulations and 5 future RCP4.5 simulations were performed in the North American CORDEX domain with a version of CanRCM4 in which interior spectral nudging was turned off. The appreciable difference analysis presented in Section 5c was then repeated with these new simulations. Figs. 5 7 of that section are reproduced here in Figs. S8 S10, respectively. In general, the area covered by statistically significant differences in the RCM and GCM response means is marginally reduced when spectral nudging is turned off. This is expected as internal variability of the large scales in the RCM relative to the driving GCM should generally increase due to chaotic divergence and enhanced upscale influence. For the lower order diagnostics of total precipitation, P, and near-surface temperature, T, a comparison of Figs. 5 and 6 to S8 and S9 reveal striking similarities. With respect to total precipitation, the precipitation enhancement in the RCM response along the west coast of North America remains statistically robust. Larger differences in the RCM and GCM responses occur at the Eastern boundary of the domain in the unnudge experiments. These are likely associated with artifacts at the outflow boundary, which are expected to be exacerbated in the spectrally unnudged configuration but further analysis would be required to verify this. For the higher-order diagnostic of 10-year return values of annual extremes of 24-hr precipitation amounts, P10, there occur significant differences in panels c and d between Fig. 7 and S10. The higher internal variability of the spectrally unnudged RCM configuration results in a domain fraction of statistical significance that is reduced from 7% to 6%. As this is essentially equal to the average areal fraction expected to be labelled significant simply by chance, it provides additional 3

4 68 69 evidence that much larger ensemble sizes are required to discern appreciable differences in such higher order statistics. 70 References Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Amer. Met. Soc., 83, , doi: /BAMS Mearns, L. O., W. J. Gutowski, R. Jones, L. Y. Leung, S. McGinnis, A. M. B. Nunes, and Y. Qian, 2009: A regional climate change assessment program for North America. Eos, Trans. Amer. Geophys. Union, 90, doi: /2009eo Mearns, L. O., and Coauthors, 2012: The North American Regional Climate Change Assessment Program. Bull. Amer. Met. Soc., doi: /bams-d Mitchell, T. D., 2008: CRU TS Tech. rep., University of East Anglia Climatic Research Unit. Available online at ts Mitchell, T. D., and P. D. Jones, 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25, New, M., M. Hulme, and P. D. Jones, 2000: Representing twentieth century space-time climate variability. Part II: Development of monthly grids of terrestrial surface climate. J. Clim., 13,

5 LIST OF FIGURES Fig. S1. June-July-August mean 2m temperature biases for relative to CRU2 observations for (a) spectrally nudged CanRCM4, (b) spectrally unnudged CanRCM4, and (c) (d) six additional NARCCAP RCMs Fig. S2. Same as Fig. S1 but for December-January-February mean 2 m temperature biases Fig. S3. Fig. S4. Fig. S5. Definition of North American subregions used to evaluated local regional biases displayed in Figs. S4 and S Annual and seasonal mean 2m temperature biases relative to the CRU2 observations over the period in the various subregions defined in Fig. S June-July-August mean precipitation rate biases for relative to CRU2 observations for (a) spectrally nudged CanRCM4, (b) spectrally unnudged CanRCM4, and (c) (d) six additional NARCCAP RCMs Fig. S6. Sam as Fig. S5 but for December-January-February mean precipitation rate biases Fig. S7. Fig. S8. Fig. S9. Annual and seasonal average precipitation biases relative to the CRU2 observations over the period in the various subregions defined in Fig. S March-April-May ensemble response mean of total precipitation between the two periods and based on 5-member ensembles of CanESM2 (panel a) and Can- RCM4 (panel b) for the future scenario RCP4.5. In this experiment, CanRCM4 employs no spectral nudging in its interior. The difference and standardized difference in the response means are presented in panels c and d respectively. In panels e and f the response differences in panels c and d are displayed only where the difference in the response means are statistically significant at the 95% confidence level. The domain/land fraction of statistically significant differences is 15%/15% As in Fig. S8 except for near surface temperature response T. The domain/land fraction of statistically significant differences is 44%/55% Fig. S10. As in Fig. S8 except for the response of 10-year return values of annual extremes of 24-hr precipitation amounts shown only over land. The land fraction of statistically significant differences is 6%

6 Precipitation Biases a) CanRCM4 b) CanRCM4 (not nudged) c) ECP2 d) HRM3 e) RCM3 f) CRCM4 g) WRFG h) MM5I a b 4 2 d e 0 f g h F IG. S1. June-July-August mean 2m temperature biases for relative to CRU2 observations for (a) 115 spectrally nudged CanRCM4, (b) spectrally unnudged CanRCM4, and (c) (d) six additional NARCCAP RCMs. 6 mm day-1 c

7 2m Temperature Biases a) CanRCM4 b) CanRCM4 (not nudged) c) ECP2 d) HRM3 e) RCM3 f) CRCM4 g) WRFG h) MM5I a b c d e 0 Celsius 2-2 f g h F IG. S2. Same as Fig. S1 but for December-January-February mean 2 m temperature biases. 7

8 F IG. S3. Definition of North American subregions used to evaluated local regional biases displayed in Figs. S4 and S7. 8

9 Regional 2m Temperature Anomalies All Regions 5 NW Forest NE Forest 0 Celsius -5 Pacific Rockies Plains Great Lakes California SW USA Gulf Coastal Atlantic FIG. S4. Annual and seasonal mean 2m temperature biases relative to the CRU2 observations over the period in the various subregions defined in Fig. S3. 9

10 Precipitation Biases a) CanRCM4 b) CanRCM4 (not nudged) c) ECP2 d) HRM3 e) RCM3 f) CRCM4 g) WRFG h) MM5I a b 4 2 d e 0 f g h F IG. S5. June-July-August mean precipitation rate biases for relative to CRU2 observations for (a) 121 spectrally nudged CanRCM4, (b) spectrally unnudged CanRCM4, and (c) (d) six additional NARCCAP RCMs. 10 mm day-1 c

11 Precipitation Biases a) CanRCM4 b) CanRCM4 (not nudged) c) ECP2 d) HRM3 e) RCM3 f) CRCM4 g) WRFG h) MM5I a b 4 2 d e 0 f g h -2-4 F IG. S6. Sam as Fig. S5 but for December-January-February mean precipitation rate biases. 11 mm day-1 c

12 Regional Precipitation Anomalies All Regions NW Forest NE Forest mm day Pacific Rockies Plains Great Lakes California SW USA Gulf Coastal Atlantic FIG. S7. Annual and seasonal average precipitation biases relative to the CRU2 observations over the period in the various subregions defined in Fig. S3. 12

Precipitation Response PȝGCM PȝRCM a b (PȝRCMí PȝGCM ) / PıGCM PȝRCM í

PȝGCM } SIG f e Standard Deviations 124 F IG. S8.

the two periods 1986 2005 125 and 2081 2100 based on 5-member ensembles

The difference 127 and standardized difference in the response means are

In panels e 128 and f the response differences in panels c and d are

13 Precipitation Response PȝGCM PȝRCM a b (PȝRCMí PȝGCM ) / PıGCM PȝRCM í PȝGCM d c Standard Deviations { (PȝRCMí PȝGCM ) / PıGCM } SIG { PȝRCM í PȝGCM } SIG f e Standard Deviations 124 F IG. S8. March-April-May ensemble response mean of total precipitation between the two periods and based on 5-member ensembles of CanESM2 (panel a) and CanRCM4 (panel b) for the future 126 scenario RCP4.5. In this experiment, CanRCM4 employs no spectral nudging in its interior. The difference 127 and standardized difference in the response means are presented in panels c and d respectively. In panels e 128 and f the response differences in panels c and d are displayed only where the difference in the response means 129 are statistically significant at the 95% confidence level. The domain/land fraction of statistically significant 130 differences is 15%/15%. 13

Screen Temperature Response TȝGCM TȝRCM a b

d c oc Standard Deviations { (TȝRCMí TȝGCM

14 Screen Temperature Response TȝGCM TȝRCM a b oc oc TȝRCM í TȝGCM (TȝRCMí TȝGCM ) / TıGCM d c oc Standard Deviations { (TȝRCMí TȝGCM ) / TıGCM } SIG { TȝRCM í TȝGCM } SIG f e Standard Deviations F IG. S9. As in Fig. S8 except for near surface temperature response T. The domain/land fraction of statistically significant differences is 44%/55% 14

10yr Return Values of Annual Extreme P P10ȝRCM P10ȝGCM a b P10ȝRCM í

P10ȝGCM } SIG f e 133 134 {(P10ȝRCM í P10ȝGCM) / P10ıGCM } SIG Standard

S8 except for the response of 10-year return values of annual extremes of

15 10yr Return Values of Annual Extreme P P10ȝRCM P10ȝGCM a b P10ȝRCM í P10ȝGCM (P10ȝRCM í P10ȝGCM) / P10ıGCM d c Standard Deviations { P10ȝRCM í P10ȝGCM } SIG f e {(P10ȝRCM í P10ȝGCM) / P10ıGCM } SIG Standard Deviations F IG. S10. As in Fig. S8 except for the response of 10-year return values of annual extremes of 24-hr precipitation amounts shown only over land. The land fraction of statistically significant differences is 6% 15

Dynamical Core: GEM (Cote et al. 1998)/GEM-LAM (Zadra et al. 2008) Physics Package: CanAM4 (von Salzen et al. 2013) CanAM4.

CCCma CCCmaGlobal Global and and Regional Regional Climate ClimateModels Models Dynamical Core: GEM (Cote et al. 1998)/GEM-LAM (Zadra et al. 2008) Physics Package: CanAM4 (von Salzen et al. 2013) CanAM4