Search

A new detection and attribution method is presented and applied to the global mean surface air temperature (GSAT) from 1900 to 2014. The method aims at attributing the climate changes to the variations of greenhouse gases, anthropogenic aerosols, and natural forcings. A convolutional neural network (CNN) is trained using the simulated GSAT from historical and single‐forcing simulations of 12 climate models. Then, we perform a backward optimization with the CNN to estimate the attributable GSAT changes. Such a method does not assume additivity in the effects of the forcings. The uncertainty in the attributable GSAT is estimated by sampling different starting points from single‐forcing simulations and repeating the backward optimization. To evaluate this new method, the attributable GSAT changes are also calculated using the regularized optimal fingerprinting (ROF) method. Using synthetic non‐additive data, we first find that the neural network‐based method estimates attributable changes better than ROF. When using GSAT data from climate model, the attributable anomalies are similar for both methods, which might reflect that the influence of forcing is mainly additive for the GSAT. However, we found that the uncertainties given both methods are different. The new method presented here can be adapted and extended in future work, to investigate the non‐additive changes found at the local scale or on other physical variables. Plain Language Summary: In order to design effective adaptation policies, it is essential to have reliable estimates of the effect of anthropogenic activities on the climate. For that purpose, a new attribution method based on a neural network is designed and evaluated. The method estimates the past global mean surface air temperatures anomalies caused by the changes in the greenhouse gases concentration, the variation of anthropogenic aerosols, and the variations driven by naturally occurring phenomena. To build this estimation, the data from observations and climate models are used. This methodology is compared with another state‐of‐the‐art method. The results of both methods are evaluated and discussed. The proposed method provides better estimations in the case of large non‐additivity of the causes of climate change and can be applied to other physical variables or at the regional scale. In the case of the global mean surface air temperature, the method presented provides estimation similar to other methods. Key Points: We present a non linear method based on neural network to attribute the global mean surface air temperature variability to different forcingsWe use a convolutional neural network associated with a backward optimization to estimate the climate response to the different external forcingsThe attributable forcings are consistent with those obtained using another state‐of‐the‐art method [ABSTRACT FROM AUTHOR]

European climate variability is shaped by atmospheric dynamics over the North Atlantic and local processes. Better understanding their future seasonality is essential to anticipate changes in weather conditions for human and natural systems. We explore atmospheric seasonality over 1979–2017 and 1979–2100 with seasonal circulation regimes (SCRs), by clustering year‐round daily fields of Z500 from the ERA‐Interim reanalysis and 12 Coupled Model Intercomparison Project phase 5 (CMIP5) climate models (historical and RCP8.5 runs). The spatial and temporal variability of SCR structures and associated patterns of surface air temperature are investigated. Climate models have biases but reproduce structures and evolutions of SCRs similar to the reanalysis over 1979–2017: decreasing frequency of winter conditions (starting later and ending earlier in the year) and the opposite for summer conditions. These changes are stronger over 1979–2100 than over 1979–2017, associated with a large increase of North Atlantic seasonal mean Z500 and temperature. When using more SCRs (more freedom in definition of seasonality), the changes over 1979–2100 correspond to a long‐term swap between SCRs, resulting in similar structures (annual cycle and spatial patterns) relative to the evolution of seasonal mean Z500 and temperature. To understand whether the evolution of SCRs is linked to uniform warming, or to changes in circulation patterns, we remove the calendar trend in the Z500 regional average to define SCRs based on detrended data (d‐SCRs). The temporal properties of d‐SCRs appear almost constant whereas their spatial patterns change, indicating that the calendar Z500 regional trend drives the evolutions of SCRs, and that changing spatial patterns in d‐SCRs account for the heterogeneity of this trend. Our study suggests that historical winter conditions will continue to decrease in the future while historical summer conditions continue to increase. However, it also suggests that the spatial and temporal patterns of SCRs would remain similar, relatively to the year‐round Z500 increase. [ABSTRACT FROM AUTHOR]

Early summer (May–June) is the season where the surface air temperature (SAT) variability is largest and may result in extreme temperature conditions over Pakistan. Therefore, we analysed the early summer interannual SAT variability over Pakistan for the period 1981–2018 using observational dataset and model simulations. We noted that upper‐level anticyclonic circulation anomalies over Pakistan, which are associated with upper to the middle tropospheric descending motion, favour clear skies with an increase in net shortwave radiation that leads to extreme surface warming over the region. Moreover, the El Niño–Southern Oscillation (ENSO) based on Niño3.4 SST index is negatively associated with SAT anomalies over Pakistan. The cold ENSO phase favours the positive geopotential height anomalies due to the strengthening of the Walker circulation that enhance the sinking motion and result in less cloudiness leading to the extreme higher surface temperature conditions over Pakistan, while the opposite happens in the warm ENSO phase. Moreover, the Saudi‐KAU Atmospheric Global Climate Model (AGCM) simulated large‐scale patterns that are in good agreement with the observations. Sensitivity experiments with the AGCM confirm that cold SST anomalies in the ENSO region significantly favour above‐normal SAT anomalies over Pakistan, while the opposite happens in the warm ENSO phase. These results are important to understand and potentially predict regional heatwaves over the South Asian region, particularly over Pakistan. [ABSTRACT FROM AUTHOR]

Climate reconstructions by data assimilation need to accommodate for sensitivities to proxies and prior estimates because models are uncertain and proxies are spatiotemporally limited. This study examines these sensitivities using multiple climate model simulations and different combinations of proxies (i.e., corals, ice cores, and tree‐ring cellulose). Experiments were conducted using an offline data assimilation approach; the results showed annual variations in the global distribution of surface air temperature and precipitation amount from 850 to 2000. Standard deviations of surface air temperature and precipitation amount during the entire period differed by up to 36% due to prior estimates. Experiments with different types of proxies showed that the El Niño‐like distribution of positive anomalies in the central to eastern tropical Pacific may only be adequately reproduced in experiments with corals, and not experiments without corals. The correlation coefficient of the NINO.3 index from 1971 to 2000 between experiments with corals and the Japanese 55‐year Reanalysis (JRA‐55) was 0.81 at maximum. By contrast, the correlation coefficient between experiments without corals and JRA‐55 was a maximum of 0.19. Key Points: Climate reconstruction by data assimilation illustrates annual variations in surface air temperature and precipitation from 850 to 2,000Spatiotemporal differences in climate reconstruction were compared based on prior estimatesThe impact of proxies on the results of past reproduced El Niño cases from 1971 to 2000 was validated [ABSTRACT FROM AUTHOR]

In this study, a dynamically downscaled regional climate model (RegCM4.3) is used to study the Indian summer monsoon (ISM) surface air temperature over the South‐Asia CORDEX domain using six convection schemes during 1986–2010. The spatial and temporal variability of mean surface air temperature has been analysed with reference to the India Meteorological Department (IMD) analysis data using various statistical scores. The sensitivity experiments in selecting the best convective parameterized schemes have been performed in simulating the surface air temperature during the summer monsoon season (June–September) over India and its five sub‐regions such as Northwest India, Northcentral India, West Peninsular India, Eastern Peninsular India, and Southern Peninsular India. The model results show the tendency of overestimation of surface air temperature mainly in four cumulus parameterization schemes (CPSs) that is, Tiedtke, Emanuel, Mix98, and Mix99 of RegCM4.3 during the JJAS, where Grell and Kuo CPSs show better agreement with the IMD data. Overall, Grell CPS has a close resemblance to the observation data with a minimum root mean square error, mean absolute error, lowest mean absolute percentage error (MAPE), and higher correlation coefficient. The model simulated results have also been investigated further using modified Nash Sutcliffe efficiency and modified Willmott's degree of index. These analyses confirm the potentiality of the Grell CPS followed by the Kuo CPS in simulating interannual variability of the surface air temperature over Indian and its five sub‐regions. The MAPE in Grell and Kuo CPSs are 0.004 and 0.013°C during monsoon season over India, respectively. The inter‐scheme difference in simulating surface air temperature is linked with the generation of low cloud convection and warming‐induced atmospheric moisture advection in the schemes. Therefore, Emanuel, Tiedtke, and Mix98 CPSs have shown a persistent nature of overestimation in surface air temperature variability during JJAS. It is also inferred that after removing the systematic mean bias from the RegCM4.3 model simulated outputs; the skill of Emanuel, Mix98, and Mix99 could be useful over the Indian subcontinent except for the southern peninsular region. [ABSTRACT FROM AUTHOR]