A novel deep learning approach for typhoon-induced storm surge modeling through efficient emulation of wind and pressure fields.

Modeling typhoon-induced storm surges requires 10-m wind and sea level pressure fields as forcings, commonly obtained using parametric models or a fully dynamical simulation by numerical weather prediction (NWP) models. The parametric models are generally less accurate than the full-physics models of the NWP, but they are often preferred owing to their computational efficiency facilitating rapid uncertainty quantification. Here, we propose using a deep learning method based on generative adversarial networks (GAN) to translate the parametric model outputs into a more realistic atmospheric forcings structure resembling the NWP model results. Additionally, we introduce lead-lag parameters to incorporate a forecasting feature in our model. Thirty-four historical typhoon events from 1981 to 2012 are selected to train the GAN, followed by storm surge simulations for the four most recent events. The proposed method efficiently transforms the parametric model into realistic forcing fields by a standard desktop computer within a few seconds. The results show that the storm surge model accuracy with forcings generated by GAN is comparable to that of the NWP model and outperforms the parametric model. Our novel GAN model offers an alternative for rapid storm forecasting and can potentially combine varied data, such as those from satellite images, to improve the forecasts further.

Analysis of COVID-19 Spread in Tokyo through an Agent-Based Model with Data Assimilation.

In this paper, we introduce an agent-based model together with a particle filter approach to study the spread of COVID-19. Investigations are mainly performed on the metropolis of Tokyo, but other prefectures of Japan are also briefly surveyed. A novel method for evaluating the effective reproduction number is one of the main outcomes of our approach. Other unknown parameters are also evaluated. Uncertain quantities, such as, for example, the probability that an infected agent develops symptoms, are tested and discussed, and the stability of our computations is examined. Detailed explanations are provided for the model and for the assimilation process.

Machine learning-based tsunami inundation prediction derived from offshore observations.

The world's largest and densest tsunami observing system gives us the leverage to develop a method for a real-time tsunami inundation prediction based on machine learning. Our method utilizes 150 offshore stations encompassing the Japan Trench to simultaneously predict tsunami inundation at seven coastal cities stretching ~100 km along the southern Sanriku coast. We trained the model using 3093 hypothetical tsunami scenarios from the megathrust (Mw 8.0-9.1) and nearby outer-rise (Mw 7.0-8.7) earthquakes. Then, the model was tested against 480 unseen scenarios and three near-field historical tsunami events. The proposed machine learning-based model can achieve comparable accuracy to the physics-based model with ~99% computational cost reduction, thus facilitates a rapid prediction and an efficient uncertainty quantification. Additionally, the direct use of offshore observations can increase the forecast lead time and eliminate the uncertainties typically associated with a tsunami source estimate required by the conventional modeling approach.

The first assimilation of Akatsuki single-layer winds and its validation with Venusian atmospheric waves excited by solar heating.

The planetary missions including the Venus Climate Orbiter 'Akatsuki' provide new information on various atmospheric phenomena. Nevertheless, it is difficult to elucidate their three-dimensional structures globally and continuously only from observations because satellite observations are considerably limited in time and space. We constructed the first 'objective analysis' of Venus' atmosphere by assimilating cloud-top horizontal winds on the dayside from the equator to mid-latitudes, which is frequently obtained from Akatsuki's Ultraviolet Imager (UVI). The three-dimensional structures of thermal tides, found recently to play a crucial role in maintaining the super rotation, are greatly improved by the data assimilation. This result is confirmed by comparison with Akatsuki's temperature observations. The momentum transport caused by the thermal tides and other disturbances are also modified by the wind assimilation and agrees well with those estimated from the UVI observations. The assimilated dataset is reliable and will be open to the public along with the Akatsuki observations for further investigation of Venus' atmospheric phenomena.

The Ensemble Mars Atmosphere Reanalysis System (EMARS) Version 1.0.

The Ensemble Mars Atmosphere Reanalysis System (EMARS) dataset version 1.0 contains hourly gridded atmospheric variables for the planet Mars, spanning Mars Year (MY) 24 through 33 (1999 through 2017). A reanalysis represents the best estimate of the state of the atmosphere by combining observations that are sparse in space and time with a dynamical model and weighting them by their uncertainties. EMARS uses the Local Ensemble Transform Kalman Filter (LETKF) for data assimilation with the GFDL/NASA Mars Global Climate Model (MGCM). Observations that are assimilated include the Thermal Emission Spectrometer (TES) and Mars Climate Sounder (MCS) temperature retrievals. The dataset includes gridded fields of temperature, wind, surface pressure, as well as dust, water ice, CO2 surface ice and other atmospheric quantities. Reanalyses are useful for both science and engineering studies, including investigations of transient eddies, the polar vortex, thermal tides and dust storms, and during spacecraft operations.

Statistical properties of global precipitation in the NCEP GFS model and TMPA observations for data assimilation.

There are many issues regarding the assimilation of satellite precipitation data into numerical models, including the non-Gaussian error distributions associated with precipitation, and large model and observation errors. As a result, it is not easy to improve the model forecast beyond a few hours by assimilating precipitation. To identify the challenges and propose practical solutions to assimilation of precipitation, statistics are calculated for global precipitation in a low-resolution NCEP Global Forecasting System (GFS) model and the TRMM Multisatellite Precipitation Analysis (TMPA). The samples are constructed using the same model with the same forecast period, observation variables, and resolution as planned in the follow-on GFS/TMPA precipitation assimilation experiments presented in the companion paper. The statistical results indicate that the T62 and T126 GFS models generally have positive bias in precipitation compared to the TMPA observations, and that the simulation of the marine stratocumulus precipitation is problematic in the T62 GFS model. It is necessary to apply to precipitation either the commonly used logarithm transformation or the newly proposed Gaussian transformation to obtain a better relationship between the model and observational precipitation. When the Gaussian transformations are separately applied to the model and observational precipitation, they serve as a bias correction that corrects the amplitude-dependent biases. In addition, using a spatially and/or temporally averaged precipitation variable, such as the 6-hour accumulated precipitation, should be advantageous for precipitation assimilation.

Modeling Sustainability: Population, Inequality, Consumption, and Bidirectional Coupling of the Earth and Human Systems.

Over the last two centuries, the impact of the Human System has grown dramatically, becoming strongly dominant within the Earth System in many different ways. Consumption, inequality, and population have increased extremely fast, especially since about 1950, threatening to overwhelm the many critical functions and ecosystems of the Earth System. Changes in the Earth System, in turn, have important feedback effects on the Human System, with costly and potentially serious consequences. However, current models do not incorporate these critical feedbacks. We argue that in order to understand the dynamics of either system, Earth System Models must be coupled with Human System Models through bidirectional couplings representing the positive, negative, and delayed feedbacks that exist in the real systems. In particular, key Human System variables, such as demographics, inequality, economic growth, and migration, are not coupled with the Earth System but are instead driven by exogenous estimates, such as UN population projections. This makes current models likely to miss important feedbacks in the real Earth-Human system, especially those that may result in unexpected or counterintuitive outcomes, and thus requiring different policy interventions from current models. The importance and imminence of sustainability challenges, the dominant role of the Human System in the Earth System, and the essential roles the Earth System plays for the Human System, all call for collaboration of natural scientists, social scientists, and engineers in multidisciplinary research and modeling to develop coupled Earth-Human system models for devising effective science-based policies and measures to benefit current and future generations.