Interface learning of multiphysics and multiscale systems

Complex natural or engineered systems comprise multiple characteristic scales, multiple spatiotemporal domains, and even multiple physical closure laws. To address such challenges, we introduce an interface learning paradigm and put forth a data-driven closure approach based on memory embedding to provide physically correct boundary conditions at the interface. To enable the interface learning for hyperbolic systems by considering the domain of influence and wave structures into account, we put forth the concept of upwind learning toward a physics-informed domain decomposition. The promise of the proposed approach is shown for a set of canonical illustrative problems. We highlight that high-performance computing environments can benefit from this methodology to reduce communication costs among processing units in emerging machine-learning-ready heterogeneous platforms toward exascale era.

A high-resolution Asia-Pacific regional coupled prediction system with dynamically downscaling coupled data assimilation

A regional coupled prediction system for the Asia-Pacific (AP-RCP) (38°E-180°, 20°S-60°N) area has been established. The AP-RCP system consists of WRF-ROMS (Weather Research and Forecast, and Regional Ocean Model System) coupled models combined with local observational information through dynamically downscaling coupled data assimilation (CDA). The system generates 18-day forecasts for the atmosphere and ocean environment on a daily quasi-operational schedule at Pilot National Laboratory for Marine Science and Technology (Qingdao) (QNLM), consisting of 2 different-resolution coupled models: 27 km WRF coupled with 9 km ROMS, 9 km WRF coupled with 3 km ROMS, while a version of 3 km WRF coupled with 3 km ROMS is in a test mode. This study is a first step to evaluate the impact of high-resolution coupled model with dynamically downscaling CDA on the extended-range predictions, focusing on forecasts of typhoon onset, improved precipitation and typhoon intensity forecasts as well as simulation of the Kuroshio current variability associated with mesoscale oceanic activities. The results show that for realizing the extended-range predictability of atmospheric and oceanic environment characterized by statistics of mesoscale activities, a fine resolution coupled model resolving local mesoscale phenomena with balanced and coherent coupled initialization is a necessary first step. The next challenges include improving the planetary boundary physics and the representation of air-sea and air-land interactions to enable the model to resolve kilometer or sub-kilometer processes.

Adversarial super-resolution of climatological wind and solar data

Accurate and high-resolution data reflecting different climate scenarios are vital for policy makers when deciding on the development of future energy resources, electrical infrastructure, transportation networks, agriculture, and many other societally important systems. However, state-of-the-art long-term global climate simulations are unable to resolve the spatiotemporal characteristics necessary for resource assessment or operational planning. We introduce an adversarial deep learning approach to super resolve wind velocity and solar irradiance outputs from global climate models to scales sufficient for renewable energy resource assessment. Using adversarial training to improve the physical and perceptual performance of our networks, we demonstrate up to a [Formula: see text] resolution enhancement of wind and solar data. In validation studies, the inferred fields are robust to input noise, possess the correct small-scale properties of atmospheric turbulent flow and solar irradiance, and retain consistency at large scales with coarse data. An additional advantage of our fully convolutional architecture is that it allows for training on small domains and evaluation on arbitrarily-sized inputs, including global scale. We conclude with a super-resolution study of renewable energy resources based on climate scenario data from the Intergovernmental Panel on Climate Change's Fifth Assessment Report.

Impact of Air Conditioning Systems on the Outdoor Thermal Environment during Summer in Berlin, Germany

This study investigates the effect of anthropogenic heat emissions from air conditioning systems (AC) on air temperature and AC energy consumption in Berlin, Germany. We conduct simulations applying the model system CCLM/DCEP-BEM, a coupled system of the mesoscale climate model COSMO-CLM (CCLM) and the urban Double Canyon Effect Parameterization scheme with a building energy model (DCEP-BEM), for a summer period of 2018. The DCEP-BEM model is designed to explicitly compute the anthropogenic heat emissions from urban buildings and the heat flux transfer between buildings and the atmosphere. We investigate two locations where the AC outdoor units are installed: either on the wall of a building (VerAC) or on the rooftop of a building (HorAC). AC waste heat emissions considerably increase the near-surface air temperature. Compared to a reference scenario without AC systems, the VerAC scenario with a target indoor temperature of 22 ∘C results in a temperature increase of up to 0.6K. The increase is more pronounced during the night and for urban areas. The effect of HorAC on air temperature is overall smaller than in VerAC. With the target indoor temperature of 22 ∘C, an urban site’s daily average AC energy consumption per floor area of a room is 9.1W m2, which is 35% more than that of a suburban site. This energy-saving results from the urban heat island effect and different building parameters between both sits. The maximum AC energy consumption occurs in the afternoon. When the target indoor temperature rises, the AC energy consumption decreases at a rate of about 16% per 2 K change in indoor temperature. The nighttime near-surface temperature in VerAC scenarios shows a declining trend (0.06K per 2 K change) with increasing target indoor temperature. This feature is not obvious in HorAC scenarios which further confirms that HorAC has a smaller impact on near-surface air temperature.

Modified Approach to Reduce GCM Bias in Downscaled Precipitation: A Study in Ganga River Basin

Reanalysis data is widely used to develop predictor-predictand models, which are further used to downscale coarse gridded general circulation models (GCM) data at a local scale. However, large variability in the downscaled product using different GCMs is still a big challenge. The first objective of this study was to assess the performance of reanalysis data to downscale precipitation using different GCMs. High bias in downscaled precipitation was observed using different GCMs, so a different downscaling approach is proposed in which historical data of GCM was used to develop a predictor-predictand model. The earlier approach is termed “Re-Obs” and the proposed approach as “GCM-Obs”. Both models were assessed using mathematical derivation and generated synthetic series. The intermodal bias in different GCMs downscaled precipitation using Re-Obs and GCM-Obs model was also checked. Coupled Model Inter-comparison Project-5 (CMIP5) data of ten different GCMs was used to downscale precipitation in different urbanized, rural, and forest regions in the Ganga river basin. Different measures were used to represent the relative performances of one downscaling approach over other approach in terms of closeness of downscaled precipitation with observed precipitation and reduction of bias using different GCMs. The effect of GCM spatial resolution in downscaling was also checked. The model performance, convergence, and skill score were computed to assess the ability of GCM-Obs and Re-Obs models. The proposed GCM-Obs model was found better than Re-Obs model to statistically downscale GCM. It was observed that GCM-Obs model was able to reduce GCM-Observed and GCM-GCM bias in the downscaled precipitation in the Ganga river basin.

Added Value of Atmosphere-Ocean Coupling in a Century-Long Regional Climate Simulation

A twentieth century-long coupled atmosphere-ocean regional climate simulation with COSMO-CLM (Consortium for Small-Scale Modeling, Climate Limited-area Model) and NEMO (Nucleus for European Modelling of the Ocean) is studied here to evaluate the added value of coupled marginal seas over continental regions. The interactive coupling of the marginal seas, namely the Mediterranean, the North and the Baltic Seas, to the atmosphere in the European region gives a comprehensive modelling system. It is expected to be able to describe the climatological features of this geographically complex area even more precisely than an atmosphere-only climate model. The investigated variables are precipitation and 2 m temperature. Sensitivity studies are used to assess the impact of SST (sea surface temperature) changes over land areas. The different SST values affect the continental precipitation more than the 2 m temperature. The simulated variables are compared to the CRU (Climatic Research Unit) observational data, and also to the HOAPS/GPCC (Hamburg Ocean Atmosphere Parameters and Fluxes from Satellite Data, Global Precipitation Climatology Centre) data. In the coupled simulation, added skill is found primarily during winter over the eastern part of Europe. Our analysis shows that, over this region, the coupled system is dryer than the uncoupled system, both in terms of precipitation and soil moisture, which means a decrease in the bias of the system. Thus, the coupling improves the simulation of precipitation over the eastern part of Europe, due to cooler SST values and in consequence, drier soil.

Long-Term Atmospheric Changes in a Convection-Permitting Regional Climate Model Hindcast Simulation over Northern Germany and the German Bight

Long-term atmospheric changes are a result of complex interactions on various spatial scales. In this study, we examine the long-term variability of the most important meteorological variables in a convection-permitting regional climate model simulation. A consistent, gridded data set from 1948 to 2014 was computed using the regional climate model COSMO-CLM with a very high convection-permitting resolution at a grid distance of 2.8 km, for a region encompassing the German Bight and Northern Germany. This is one of the very first atmospheric model simulations with such high resolution, and covering several decades. Using a very high-resolution hindcast, this study aims to extend knowledge of the significance of regional details for long-term variability and multi-decadal trends of several meteorological variables such as wind, temperature, cloud cover, precipitation, and convective available potential energy (CAPE). This study demonstrates that most variables show merely large decadal variability and no long-term trends. The analysis shows that the most distinct and significant positive trends occur in temperature and in CAPE for annual mean values as well as for extreme events. No clear and no significant trend is detectable for the annual sum of precipitation and for extreme precipitation. However, spatial structures in the trends remain weak.

Lessons from Inter-Comparison of Decadal Climate Simulations and Observations for the Midwest U.S. and Great Lakes Region

Even with advances in climate modeling, meteorological impact assessment remains elusive, and decision-makers are forced to operate with potentially malinformed predictions. In this article, we investigate the dependence of the Weather Research and Forecasting (WRF) model simulated precipitation and temperature at 12- and 4-km horizontal resolutions and compare it with 32-km NARR data and 1/16th-degree gridded observations for the Midwest U.S. and Great Lakes region from 1991 to 2000. We used daily climatology, inter-annual variability, percentile, and dry days as metrics for inter-comparison for precipitation. We also calculated the summer and winter daily seasonal minimum, maximum, and average temperature to delineate the temperature trends. Results showed that NARR data is a useful precipitation product for mean warm season and summer climatological studies, but performs extremely poorly for winter and cold seasons for this region. WRF model simulations at 12- and 4-km horizontal resolutions were able to capture the lake-effect precipitation successfully when driven by observed lake surface temperatures. Simulations at 4-km showed negative bias in capturing precipitation without convective parameterization but captured the number of dry days and 99th percentile precipitation extremes well. Overall, our study cautions against hastily pushing for increasingly higher resolution in climate studies, and highlights the need for the careful selection of large-scale boundary forcing data.

Blocking representation in the ERA-Interim driven EURO-CORDEX RCMs

While Regional Climate Models (RCMs) have been shown to yield improved simulations compared to General Circulation Model (GCM), their representation of large-scale phenomena like atmospheric blocking has been hardly addressed. Here, we evaluate the ability of RCMs to simulate blocking situations present in their reanalysis driving data and analyse the associated impacts on anomalies and biases of European 2-m air temperature (TAS) and precipitation rate (PR). Five RCM runs stem from the EURO-CORDEX ensemble while three RCMs are WRF models with different nudging realizations, all of them driven by ERA-Interim for the period 1981-2010. The detected blocking systems are allocated to three sectors of the Euro-Atlantic region, allowing for a characterization of distinctive blocking-related TAS and PR anomalies. Our results indicate some misrepresentation of atmospheric blocking over the EURO-CORDEX domain, as compared to the driving reanalysis. Most of the RCMs showed fewer blocks than the driving data, while the blocking misdetection was negligible for RCMs strongly conditioned to the driving data. A higher resolution of the RCMs did not improve the representation of atmospheric blocking. However, all RCMs are able to reproduce the basic anomaly structure of TAS and PR connected to blocking. Moreover, the associated anomalies do not change substantially after correcting for the misrepresentation of blocking in RCMs. The overall model bias is mainly determined by pattern biases in the representations of surface parameters during non-blocking situations. Biases in blocking detections tend to have a secondary influence in the overall bias due to compensatory effects of missed blockings and non-blockings. However, they can lead to measurable effects in the presence of a strong blocking underestimation.

Local difference measures between complex networks for dynamical system model evaluation

A faithful modeling of real-world dynamical systems necessitates model evaluation. A recent promising methodological approach to this problem has been based on complex networks, which in turn have proven useful for the characterization of dynamical systems. In this context, we introduce three local network difference measures and demonstrate their capabilities in the field of climate modeling, where these measures facilitate a spatially explicit model evaluation.Building on a recent study by Feldhoff et al. [8] we comparatively analyze statistical and dynamical regional climate simulations of the South American monsoon system [corrected]. types of climate networks representing different aspects of rainfall dynamics are constructed from the modeled precipitation space-time series. Specifically, we define simple graphs based on positive as well as negative rank correlations between rainfall anomaly time series at different locations, and such based on spatial synchronizations of extreme rain events. An evaluation against respective networks built from daily satellite data provided by the Tropical Rainfall Measuring Mission 3B42 V7 reveals far greater differences in model performance between network types for a fixed but arbitrary climate model than between climate models for a fixed but arbitrary network type. We identify two sources of uncertainty in this respect. Firstly, climate variability limits fidelity, particularly in the case of the extreme event network; and secondly, larger geographical link lengths render link misplacements more likely, most notably in the case of the anticorrelation network; both contributions are quantified using suitable ensembles of surrogate networks. Our model evaluation approach is applicable to any multidimensional dynamical system and especially our simple graph difference measures are highly versatile as the graphs to be compared may be constructed in whatever way required. Generalizations to directed as well as edge- and node-weighted graphs are discussed.