In our previous work [1] we show the effectiveness of learning based schemes to emulate numerical simulations. Specifically, we explore the use of GANs to model Snow Water Equivalent (SWE). In this work, we present further analysis of these results, with a model now trained on 30 years of SWE data compared to 10 years for the previous. First, we study the model outputs with station data and find the outputs to closely match the station data. Second, we study the uncertainty of the cGAN based results. We study the correlation between noise and the outputs generated. We hope these contributions can help the community leverage the power of these deep learning models for future works.

While deep learning has shown tremendous success in a wide range of domains, it remains a grand challenge to incorporate physical principles in a systematic manner to the design, training, and inference of such models. In this paper, we aim to predict turbulent flow by learning its highly nonlinear dynamics from spatiotemporal velocity fields of large-scale fluid flow simulations of relevance to turbulence modeling and climate modeling. We adopt a hybrid approach by marrying two well-established turbulent flow simulation techniques with deep learning. Specifically, we introduce trainable spectral filters in a coupled model of Reynolds-averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES), followed by a specialized U-net for prediction. Our approach, which we call Turbulent-Flow Net, is grounded in a principled physics model, yet offers the flexibility of learned representations. We compare our model with state-of-the-art baselines and observe significant reductions in error for predictions 60 frames ahead. Most importantly, our method predicts physical fields that obey desirable physical characteristics, such as conservation of mass, whilst faithfully emulating the turbulent kinetic energy field and spectrum, which are critical for accurate prediction of turbulent flows.

In this study we propose a new method to simulate hyper-realistic urban patterns using Generative Adversarial Networks trained with a global urban land-use inventory. We generated a synthetic urban “universe” that qualitatively reproduces the complex spatial organization observed in global urban patterns, while being able to quantitatively recover certain key high-level urban spatial metrics.

Urban planning applications (energy audits, investment, etc.) require an understanding of built infrastructure and its environment, i.e., both low-level, physical features (amount of vegetation, building area and geometry etc.), as well as higher-level concepts such as land use classes (which encode expert understanding of socio-economic end uses). This kind of data is expensive and labor-intensive to obtain, which limits its availability (particularly in developing countries). We analyze patterns in land use in urban neighborhoods using large-scale satellite imagery data (which is available worldwide from third-party providers) and state-of-the-art computer vision techniques based on deep convolutional neural networks. For supervision, given the limited availability of standard benchmarks for remote-sensing data, we obtain ground truth land use class labels carefully sampled from open-source surveys, in particular the Urban Atlas land classification dataset of $20$ land use classes across $~300$ European cities. We use this data to train and compare deep architectures which have recently shown good performance on standard computer vision tasks (image classification and segmentation), including on geospatial data. Furthermore, we show that the deep representations extracted from satellite imagery of urban environments can be used to compare neighborhoods across several cities. We make our dataset available for other machine learning researchers to use for remote-sensing applications.

As a way to match peaks in demand to available supply in real-time on the power grid, energy utility companies employ Demand-Response (DR) strategies. With the recent deployment of advanced metering infrastructure collecting highly granular (sub-hourly) data on consumption from millions of users system operators may now understand how demand arises down to the individual level. In this paper we present an application of a dynamic model that describes residential users' thermally-sensitive consumption using hourly electricity and weather readings. We build rich profiles at individual and group levels that may be used to inform DR programs that focus on temperature-dependent consumption such as heating and cooling. We learn individual thermal profiles for a large user sample, and describe the seasonal and time-of-day distribution of thermal regimes. Finally, we build aggregate thermal profiles for geographically-homogeneous groups of different size. We argue that aggregation leads to regularity in temperature response of energy consumption, and characterize empirically the critical size of groups needed to achieve limiting regularity.