Assessment of Sea Ice Simulations in the CMIP5

9 The historical simulations of sea ice during 1979 to 2005 by the Coupled Model 10 Intercomparison Project Phase 5 (CMIP5) are compared with satellite observations, 11 Global Ice-Ocean Modeling and Assimilation System (GIOMAS) output data and 12 Pan-Arctic Ice Ocean Modeling and Assimilation System (PIOMAS) output data in 13 this study. Forty-nine models, almost all of the CMIP5 climate models and Earth 14 System Models with historical simulation, are used. For the Antarctic, multi-model 15 ensemble mean (MME) results can give good climatology of sea ice extent (SIE), but 16 the linear trend is incorrect. The linear trend of satellite-observed Antarctic SIE is 17 1.29×10 km decade; only 1/7 CMIP5 models show increasing trends, and the 18 linear trend of CMIP5 MME is negative (-3.36×10 km decade). For the Arctic, 19 both climatology and linear trend are better reproduced. Sea ice volume (SIV) is also 20 evaluated in this study, and this is a first attempt to evaluate the SIV in all CMIP5 21 models. Compared with the GIOMAS and PIOMAS data, the SIV values in both 22 Antarctic and Arctic are too small, especially for the Antarctic in spring and winter. 23 The GIOMAS Antarctic SIV in September is 19.1×10 km, while the corresponding 24 Antarctic SIV of CMIP5 MME is 13.0×10 km, almost 32% less. The Arctic SIV of 25 CMIP5 in April is 27.1×10 km, which is also less than the PIOMAS SIV (29.5× 26 10 km). This means that the sea ice thickness simulated in CMIP5 is too thin 27 although the SIE is fairly well simulated. 28


30
The Coupled Model Intercomparison Project Phase 5 (CMIP5) provides a very useful 31 platform for studying climate change. Simulations and projections by more than 60 32 and 6.62 million km 2 (highligthed in Table 1 with bold font), respectively. BNU-ESM 140 simulated February SIE is even larger than MIROC5 simulated September SIE. Large 141 SIE spread and small MME SIE errors indicate that we should use as many models as 142 we can when using CMIP5 outputs. observations. Although CMIP5 simulated MME SIE fits the observations well, MME 153 spatial map of SIC fits the observations not so well. MME SICs in the Weddell Sea, 154 the Bellingshausen Sea and the Amundsen Sea are too little. In September, most 155 CMIP5 models have better performance than that in February, and MME SIC also has 156 better spatial pattern. anthropogenic forcing. From Table 1 we can see that several models (highligthed in Table 1   The trends of observed Antarctic SIC have large spatial differences (Fig. 3), but the 180 simulated Antarctic SIC trends are almost decreasing everywhere (Fig. 4). Figure 3   181 shows that decreasing SIC is mainly in the Antarctic Peninsula, which is one of the 182 three high-latitude areas showing rapid regional warming over the last 50 years 183 (Vaughan et al., 2003). SIC also decreases in the Bellingshausen Sea and the 184 Amundsen Sea in summer and autumn. The increasing SIC is mainly in the Ross Sea 185 all year round and in the Weddell Sea in summer and autumn. Figure 4 clearly shows 186 that CMIP5 MME SIC has decreasing trend everywhere except in the coast of the 187 Amundsen Sea and in part of the Ross Sea in spring and winter.

188
SIV depends on both sea ice coverage and sea ice thickness. SIV is more directly tied 189 to climate forcing than SIE. So, SIV is an important climate indicator in climate study.

190
The observed sea ice thickness records are mainly from submarine, aircraft and

196
These observations provide modelers with useful validation of their models. But, these data are not easily used to long-term simulation validations by now because these data are not too long enough. Here, we use GIOMAS data, which is from a 199 global ice-ocean model (Zhang and Rothrock, 2003) with data assimilation capability.

200
What we should keep in mind is that GIOMAS sea ice thickness is not from the GIOMAS data (Fig. 5). The annual mean SIV from GIOMAS is 11.02×10 3 km 3 , 215 but CMIP5 MME SIV is only 7.73×10 3 km 3 (Table 1). In February, Antarctic SIV 216 from GIOMAS is 1.9×10 3 km 3 , while the CMIP5 MME is 2.7×10 3 km 3 . In 217 September, GIOMAS SIV is 19.1×10 3 km 3 , while CMIP5 MME is only 13.0×10 3 218 km 3 , almost 32% less than the GIOMAS. We can also see from Figure 5a that the 219 model spread of Antarctic SIV in CMIP5 is very large. The one standard deviation of 220 modeled SIV is much larger than 15% of the GIOMAS data in every month. We 221 checked the correlation between SIE RMS error and SIV RMS error, and we can find 222 that the models with small SIE RMS errors always have small SIV RMS errors (Table  decreasing trend Table 1 with bold font).  Table 1 with bold font). Arctic SIE amplitudes from CMIP5 251 models also have large spread. GISS-E2-R-CC has the largest amplitude with the 252 value of 16.73 million km 2 , and FGOAL-g2 has the smallest amplitude with the value 253 of only 3.35 million km 2 (highligthed in Table 2 with bold font). Compared with Table 2).

256
CMIP5 MME SIE shows a decreasing trend that is consistent with the satellite 257 observation, though the decreasing rate is a little smaller than that of the observation between summer and autumn is, however, larger than that of the observations. The 266 main reason is CMIP5-simulated SIE has small reduction in summer, especially in 267 July (Fig. 7). Satellite-observed SIE decreasing rate is 5.22% per decade in July, while 268 the CMIP5-simulated decreasing rate is 3.54% per decade. The largest decreasing rate 269 is in September; the observed trend is -8.61% per decade and the simulated trend is is too little Arctic SIV all year round and too large model spread (Fig. 10). In spring, 284 the Arctic has the largest SIV. Long-term mean PIOMAS SIV is maximum in April 285 with 29.5 × 10 3 km 3 , and the corresponding CMIP5 MME is 27.1 × 10 3 km 3 .

315
Our results show that the Arctic sea ice simulations are better than the Antarctic sea 316 ice simulations, and SIE simulations are better than SIV simulations. CMIP5 MME 317 SIV is too less in winter and spring because the sea ice thickness in CMIP5 models is 318 too thin in winter and spring compared with the GIOMAS and PIOMAS data. In the 319 Antarctic, MME can reproduce good mean state and monthly amplitude for SIE, but 320 for SIV MME mean state and amplitude are smaller. In the Arctic, MME can 321 reproduce good mean state and monthly amplitude for both SIE and SIV. CMIP5 trends. Can these few CMIP5 models give correct Antarctic sea ice trend? If we use 338 these eight CMIP5 models to plot Antarctic SIC trends (not shown) as in Fig. 4, we 339 will find that these eight CMIP5 model mean SIC trends have different spatial 340 patterns with the observations (Fig. 3) Table 1. Antarctic sea ice metrics in CMIP5 models, satellite observations and GIOMAS dataset. Column (a) is mean annual SIE in million km 2 .

496
Column (b) is monthly SIE amplitude in million km 2 . Column (c) is standard deviation of detrended monthly SIE anomaly in million km 2 .

497
Column (d) is linear trend in monthly SIE in 10 5 km 2 decade -1 , and the value in parentheses is 95% confidence level. Column (e) is monthly SIE 498 root mean square error in million km 2 . Column (f) is mean annual SIV in 10 3 km 3 . Column (g) is monthly SIV amplitude in 10 3 km 3 . Column (h) 499 is standard deviation of detrended monthly SIV anomaly in 10 3 km 3 . Column (i) is linear trend in monthly SIV in 10 3 km 3 decade -1 , and the 500 value in parentheses is 95% confidence level. Column (j) is monthly SIV root mean square error in 10 3 km 3 .

501
Data sources or CMIP5 models Column (b) is monthly SIE amplitude in million km 2 . Column (c) is standard deviation of detrended monthly SIE anomaly in million km 2 .

504
Column (d) is linear trend in monthly SIE in 10 5 km 2 decade -1 , and the value in parentheses is 95% confidence level. Column (e) is monthly SIE 505 root mean square error in million km 2 . Column (f) is mean annual SIV in 10 3 km 3 . Column (g) is monthly SIV amplitude in 10 3 km 3 . Column (h) 506 is standard deviation of detrended monthly SIV anomaly in 10 3 km 3 . Column (i) is linear trend in monthly SIV in 10 3 km 3 decade -1 , , and the 507 value in parentheses is 95% confidence level. Column (j) is monthly SIV root mean square error in 10 3 km 3 .