Reducing uncertainty in local climate projections

Planning for adaptation to climate change requires accurate climate projections. 6 Recent studies have shown that the uncertainty in global mean surface temperature 7 projections can be considerably reduced by using historical observations. However, 8 the transposition of these new results to the local scale is not yet available. Here we 9 adapt an innovative statistical method that combines the latest generation of climate 10 model simulations, global observations, and local observations to reduce uncertainty 11 in local temperature projections. By taking advantage of the tight links between local 12 and global temperature, we can derive the local implications of global constraints. 13 The model uncertainty is reduced by 30% up to 50% at any location worldwide, 14 allowing to substantially improve the quantiﬁcation of risks associated with future 15 climate change. A rigorous evaluation of these results within a perfect model frame- 16 work indicates a robust skill, leading to a high conﬁdence in our constrained climate 17 projections. 18


19
As the global mean temperature keeps rising and climate change intensifies, there is 20 a growing demand for local scale monitoring of current and future climate change. 21 Assessing and planning the adaptation to the expected unprecedented impacts of 22 climate change on humans activities, ecosystems and the biosphere as a whole, require 23 an accurate local information with well calibrated uncertainties. This need relates to 24 estimates of warming to date and the future warming in response to set of scenarios 25 of future greenhouse gas emissions. 26 There is now clear evidence that the recent increase of the average Earth's tempera-27 ture is mostly due to human activities [1]. Concurrently, the anthropogenic influence

54
Here we assess how much uncertainty in local temperature projections can be re-55 duced. We first take advantage of the tight links that exist between local climate 56 and global mean surface temperature (GMST) [18,19]. Specifically, we describe 57 the local implications of the recent advances in the reduction of the uncertainties 58 in GMST projections. We then provide a set of local-scale temperature projections, 59 which encapsulate another source of information: the observed local warming to date. 60 If compared to the global mean temperature record, local observations are typically 61 more affected by internal variability and measurement uncertainty. Yet, they still 62 provide a useful source of information on both past and future trends, particularly 63 4 over some specific regions. We discuss how much these two types of observations 64 (global and local) narrow uncertainty on future warming ranges. Such a reduction 65 is expected to provide more accurate information that becomes critical for policy-66 makers in the local climate risk management [20], as well as for the climate science  GMST annual observations (black points) are used to constrain concatenated historical and SSP5-8.5 scenario simulations of GMST. The unconstrained (pink) and constrained (red) ranges stand for the 5-95% confidence interval of the forced response as estimated from 28 CMIP6 models. The thick pink (red) line stands for the ensemble mean (best estimate). All values are anomalies with respect to the 1850-1900 period. (b) Intermodel correlation between simulated GMST trends over the 1850-2019 period and local temperature trends over the 2020-2100 period. Stippling indicates regions with non-significant correlation (p-value > 0.05 based on a two-sided Student's t-test). 7 Constrain local climate projections with global ob-91 servations 92 Climate models exhibit a strong correlation between current GMST changes and 93 future local warming over most regions of the globe (Figure 1b). To take such a 94 relationship into account, we extend the KCC method to constrain local temper-  ranges is very close to the constrained GMST ranges shown in Figure 1a (not shown).

162
The addition of local information can also clearly modify the warming pattern at +2   can be criticised [34]. Therefore, using a subset of models qualified as independent a 232 priori, or weighting the models in this way [35,36], before applying the observational 233 constraint, may provide even more reliable results.

234
Our results demonstrate that available observations already offer valuable informa-235 tion to sharpen climate projections. As the climate system will continue to change   Figure 1 GMST time series and its correlation with local temperature. (a) GMST annual observations (black points) are used to constrain concatenated historical and SSP5-8.5 scenario simulations of GMST. The unconstrained (pink) and constrained (red) ranges stand for the 5{95% con dence interval of the forced response as estimated from 28 CMIP6 models. The thick pink (red) line stands for the ensemble mean (best estimate). All values are anomalies with respect to the 1850{1900 period. (b) Intermodel correlation between simulated GMST trends over the 1850-2019 period and local temperature trends over the 2020-2100 period. Stippling indicates regions with non-signi cant correlation (p-value > 0:05 based on a twosided Student's t-test). Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 4
CRPSS for the constrained temperature projections over the 2081-2100 period within the perfect model framework. Red (green) boxplots (one for each location, see blue points in Fig. 1b)   Mean temperature change at a +2 oC GMST warming. Best estimate of the constrained local temperature changes in the Local+GMST case. Similarly to Figure 2b, the constrained and unconstrained temperature ranges are shown for several world capitals cities over the 1850-2100 period. All values are anomalies with respect to the 1850{1900 period. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.