The impacts of climate change are felt by most critical systems, such as infrastructure, ecological systems, and power-plants. However, contemporary Earth System Models (ESM) are run at spatial resolutions too coarse for assessing effects this localized. Local scale projections can be obtained using statistical downscaling, a technique which uses historical climate observations to learn a low-resolution to high-resolution mapping. The spatio-temporal nature of the climate system motivates the adaptation of super-resolution image processing techniques to statistical downscaling. In our work, we present DeepSD, a generalized stacked super resolution convolutional neural network (SRCNN) framework with multi-scale input channels for statistical downscaling of climate variables. A comparison of DeepSD to four state-of-the-art methods downscaling daily precipitation from 1 degree (~100km) to 1/8 degrees (~12.5km) over the Continental United States. Furthermore, a framework using the NASA Earth Exchange (NEX) platform is discussed for downscaling more than 20 ESM models with multiple emission scenarios.

Statistical downscaling of global climate models (GCMs) allows researchers to study local climate change effects decades into the future. A wide range of statistical models have been applied to downscaling GCMs but recent advances in machine learning have not been explored. In this paper, we compare four fundamental statistical methods, Bias Correction Spatial Disaggregation (BCSD), Ordinary Least Squares, Elastic-Net, and Support Vector Machine, with three more advanced machine learning methods, Multi-task Sparse Structure Learning (MSSL), BCSD coupled with MSSL, and Convolutional Neural Networks to downscale daily precipitation in the Northeast United States. Metrics to evaluate of each method's ability to capture daily anomalies, large scale climate shifts, and extremes are analyzed. We find that linear methods, led by BCSD, consistently outperform non-linear approaches. The direct application of stateof-the-art machine learning methods to statistical downscaling does not provide improvements over simpler, longstanding approaches. * vandal.t@husky.neu.edu † evan.kodra@risq.io ‡ a.ganguly@neu.edu limiting models to coarse spatial and temporal scale projections. While the coarse scale projections are useful in understanding climate change at a global and continental level, regional and local understanding is limited. Most often, the critical systems society depends on exist at the regional and local scale, where projections are most limited. Downscaling techniques are applied to provide climate projections at finer spatial scales, exploiting GCMs to build higher resolution outputs. Statistical and dynamical are the two classes of techniques used for downscaling. The statistical downscaling (SD) approach aims to learn a statistical relationship between coarse scale climate variables (ie. GCMs) and high resolution observations. The other approach, dynamical downscaling, joins the coarse grid GCM projections with known local and regional processes to build Regional Climate Models (RCMs). RCMs are unable to generalize from one region to another as the parameters and physical processes are tuned to specific regions. Though RCMs are useful for hypothesis testing, their lack of generality across regions and extensive computational resources required are strong disadvantages.

No abstract

Deep Learning (DL) methods have been transforming computer vision with innovative adaptations to other domains including climate change. For DL to pervade Science and Engineering (S&E) applications where risk management is a core component, wellcharacterized uncertainty estimates must accompany predictions. However, S&E observations and model-simulations often follow heavily skewed distributions and are not well modeled with DL approaches, since they usually optimize a Gaussian, or Euclidean, likelihood loss. Recent developments in Bayesian Deep Learning (BDL), which attempts to capture uncertainties from noisy observations, aleatoric, and from unknown model parameters, epistemic, provide us a foundation. Here we present a discrete-continuous BDL model with Gaussian and lognormal likelihoods for uncertainty quantification (UQ). We demonstrate the approach by developing UQ estimates on "DeepSD", a super-resolution based DL model for Statistical Downscaling (SD) in climate applied to precipitation, which follows an extremely skewed distribution. We find that the discrete-continuous models outperform a basic Gaussian distribution in terms of predictive accuracy and uncertainty calibration. Furthermore, we find that the lognormal distribution, which can handle skewed distributions, produces quality uncertainty estimates at the extremes. Such results may be important across S&E, as well as other domains such as finance and economics, where extremes are often of significant interest. Furthermore, to our knowledge, this is the first UQ model in SD where both aleatoric and epistemic uncertainties are characterized.

No abstract

Private businesses in sectors, such as food, energy, and retail, as well as public sector and federal agencies are interested in the predictive understanding of weather impacts on crop yield, which is an important aspect of food security. Scientific literature has mainly examined how crop yield is impacted by growing season-averaged weather indices. Although a few studies did consider weather extremes in their analysis, their scope was either restricted to measuring their conditional relationship with yield or the extreme event types considered were limited. Selection of regression models, whether the more commonly used linear approaches or nonlinear methods, have not been appropriately justified in this context. Here, we develop data-driven methods to examine two interrelated hypotheses for improved scientific understanding and enhanced predictive modeling. The first hypothesis, that extreme weather indices have a statistically significant information content in them is found to be valid based on linear and nonlinear methods for pairwise dependence. The second hypothesis, examines the value addition of nonlinear regression methods, and suggests that linear approaches may not alone be adequate. The results of this study can inform scientific understanding, generation and relevance of indices and end-to-end risk assessment systems in the context of climate impacts on crop yield. An immediate application may be in the context of NASA Earth Exchange (NEX) which facilitates the generation and dissemination of impacts relevant weather data and indices using a multitude of satellite-derived data sets and model outputs.

A provisional surface reflectance (SR) product from the Advanced Himawari Imager (AHI) on-board the new generation geostationary satellite (Himawari-8) covering the period between July 2015 and December 2018 is made available to the scientific community. The Multi-Angle Implementation of Atmospheric Correction (MAIAC) algorithm is used in conjunction with time series Himawari-8 AHI observations to generate 1-km gridded and tiled land SR every 10 minutes during day time. This Himawari-8 AHI SR product includes retrieved atmospheric properties (e.g., aerosol optical depth at 0.47µm and 0.51µm), spectral surface reflectance (AHI bands 1-6), parameters of the RTLS BRDF model, and quality assurance flags. Product evaluation shows that Himawari-8 AHI data on average yielded 35% more cloud-free, valid pixels in a single day when compared to available data from the low earth orbit (LEO) satellites Terra/Aqua with MODIS sensor. Comparisons of Himawari-8 AHI SR against corresponding MODIS SR products (MCD19A1) over a variety of land cover types with the similar viewing geometry show high consistency between them, with correlation coefficients (r) being 0.94 and 0.99 for red and NIR bands, respectively. The high-frequency geostationary data are expected to facilitate studies of ecosystems on daily to diurnal time scales, complementing observations from networks such as the FLUXNET.

Generative models have the capacity to model and generate new examples from a dataset and have an increasingly diverse set of applications driven by commercial and academic interest. In this work, we present an algorithm for learning a latent variable generative model via generative adversarial learning where the canonical uniform noise input is replaced by samples from a graphical model. This graphical model is learned by a Boltzmann machine which learns low-dimensional feature representation of data extracted by the discriminator. A quantum processor can be used to sample from the model to train the Boltzmann machine. This novel hybrid quantum-classical algorithm joins a growing family of algorithms that use a quantum processor sampling subroutine in deep learning, and provides a scalable framework to test the advantages of quantum-assisted learning. For the latent space model, fully connected, symmetric bipartite and Chimera graph topologies are compared on a reduced stochastically binarized MNIST dataset, for both classical and quantum sampling methods. The quantum-assisted associative adversarial network successfully learns a generative model of the MNIST dataset for all topologies. Evaluated using the Fréchet inception distance and inception score, the quantum and classical versions of the algorithm are found to have equivalent performance for learning an implicit generative model of the MNIST dataset. Classical sampling is used to demonstrate the algorithm on the LSUN bedrooms dataset, indicating scalability to larger and color datasets. Though the quantum processor used here is a quantum annealer, the algorithm is general enough such that any quantum processor, such as gate model quantum computers, may be substituted as a sampler.