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.

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&EE) applications where risk management is a core component, well-characterized 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.

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. Depending on statistical modeling choices, downscaled projections have been shown to vary significantly terms of accuracy and reliability. 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 for statistical downscaling of climate variables. DeepSD augments SRCNN with multi-scale input channels to maximize predictability in statistical downscaling. We provide a comparison with Bias Correction Spatial Disaggregation as well as three Automated-Statistical Downscaling approaches in 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.

Maintaining population diversity has long been considered fundamental to the effectiveness of evolutionary algorithms. Recently, with the advent of novelty search, there has been an increasing interest in sustaining behavioral diversity by using both fitness and behavioral novelty as separate search objectives. However, since the novelty objective explicitly rewards diverging from other individuals, it can antagonize the original fitness objective that rewards convergence toward the solution(s). As a result, fostering behavioral diversity may prevent proper exploitation of the most interesting regions of the behavioral space, and thus adversely affect the overall search performance. In this paper, we argue that an antagonism between behavioral diversity and fitness can indeed exist in semantic genetic programming applied to symbolic regression. Minimizing error draws individuals toward the target semantics but promoting novelty, defined as a distance in the semantic space, scatters them away from it. We introduce a less conflicting novelty metric, defined as an angular distance between two program semantics with respect to the target semantics. The experimental results show that this metric, in contrast to the other considered diversity promoting objectives, allows to consistently improve the performance of genetic programming regardless of whether it employs a syntactic or a semantic search operator.

We investigate the use of Deep Neural Networks for the classification of image datasets where texture features are important for generating class-conditional discriminative representations. To this end, we first derive the size of the feature space for some standard textural features extracted from the input dataset and then use the theory of Vapnik-Chervonenkis dimension to show that hand-crafted feature extraction creates low-dimensional representations which help in reducing the overall excess error rate. As a corollary to this analysis, we derive for the first time upper bounds on the VC dimension of Convolutional Neural Network as well as Dropout and Dropconnect networks and the relation between excess error rate of Dropout and Dropconnect networks. The concept of intrinsic dimension is used to validate the intuition that texture-based datasets are inherently higher dimensional as compared to handwritten digits or other object recognition datasets and hence more difficult to be shattered by neural networks. We then derive the mean distance from the centroid to the nearest and farthest sampling points in an n-dimensional manifold and show that the Relative Contrast of the sample data vanishes as dimensionality of the underlying vector space tends to infinity.

Satellite image classification is a challenging problem that lies at the crossroads of remote sensing, computer vision, and machine learning. Due to the high variability inherent in satellite data, most of the current object classification approaches are not suitable for handling satellite datasets. The progress of satellite image analytics has also been inhibited by the lack of a single labeled high-resolution dataset with multiple class labels. The contributions of this paper are twofold -- (1) first, we present two new satellite datasets called SAT-4 and SAT-6, and (2) then, we propose a classification framework that extracts features from an input image, normalizes them and feeds the normalized feature vectors to a Deep Belief Network for classification. On the SAT-4 dataset, our best network produces a classification accuracy of 97.95% and outperforms three state-of-the-art object recognition algorithms, namely - Deep Belief Networks, Convolutional Neural Networks and Stacked Denoising Autoencoders by ~11%. On SAT-6, it produces a classification accuracy of 93.9% and outperforms the other algorithms by ~15%. Comparative studies with a Random Forest classifier show the advantage of an unsupervised learning approach over traditional supervised learning techniques. A statistical analysis based on Distribution Separability Criterion and Intrinsic Dimensionality Estimation substantiates the effectiveness of our approach in learning better representations for satellite imagery.

Accurate tree-cover estimates are useful in deriving above-ground biomass density estimates from very high resolution (VHR) satellite imagery data. Numerous algorithms have been designed to perform tree-cover delineation in high-to-coarse-resolution satellite imagery, but most of them do not scale to terabytes of data, typical in these VHR data sets. In this paper, we present an automated probabilistic framework for the segmentation and classification of 1-m VHR data as obtained from the National Agriculture Imagery Program (NAIP) for deriving tree-cover estimates for the whole of Continental United States, using a high-performance computing architecture. The results from the classification and segmentation algorithms are then consolidated into a structured prediction framework using a discriminative undirected probabilistic graphical model based on conditional random field, which helps in capturing the higher order contextual dependence relations between neighboring pixels. Once the final probability maps are generated, the framework is updated and retrained by incorporating expert knowledge through the relabeling of misclassified image patches. This leads to a significant improvement in the true positive rates and reduction in false positive rates (FPRs). The tree-cover maps were generated for the state of California, which covers a total of 11 095 NAIP tiles and spans a total geographical area of 163 696 sq. miles. Our framework produced correct detection rates of around 88% for fragmented forests and 74% for urban tree-cover areas, with FPRs lower than 2% for both regions. Comparative studies with the National Land-Cover Data algorithm and the LiDAR high-resolution canopy height model showed the effectiveness of our algorithm for generating accurate high-resolution tree-cover maps.

Phenological information can be provided globally using remote sensing based time-series vegetation indices. Basic differences in the data and methods used can yield different results. This study analysed such differences in the phenological information, mainly onset of greenness (OG), estimated using the Enhanced Vegetation Index (EVI) from Moderate Resolution Imaging Spectroradiometer (MODIS) data and the Terrestrial Chlorophyll Index (MTCI) from Medium Resolution Imaging Spectrometer (MERIS) data. The two datasets were processed independently using different techniques to provide weekly estimates. Differences in the OG results were analysed for two years (2003 & 2006) and at four levels: a) full study area, b) within land cover classes, c) within core zones of each class and d) at the edge zones of each class. It was found that the trend of OG estimated from MODIS and MERIS were spatially similar, although not the same. From 15 Biome classes found in the study area the classes with the greatest differences were evergreen needle leaf, mixed forest and cropland. The differences were mainly due to the characteristic nature of the indices and also, to some extent, due to false internal flags in the algorithms.

This study explores the capability of satellite remote sensing to detect relatively rapid changes of vegetation cover in northern Fennoscandian regions in response to disturbance more generally, and insect defoliation damage in particular. The data used is a long term time series of leaf area index (LAI) at 8Km resolution derived from the Advanced Very High Resolution Radiometer (AVHRR) between 1982 and 2006, developed to be structurally consistent with the Moderate Resolution Imaging Spectrometer (MODIS) record. The study explores the potential of frequentist traditional statistics to detect disturbances at this coarse spatial resolution over the 25 years time series, and outlines the possibilities that Bayesian methods offer to improve the detection by including prior information on the profile of such disturbance events.

We have developed and tested a physically based approach, rooted in the theory of stochastic radiative transfer within vegetation, for generating a seamless time series of LAI and FPAR fields from AVHRR NDVI data and MODIS reflectance data. Our goal is to: (i) develop a theoretical formulation that results in consistent LAI and FPAR fields from satellite data of varying spectral and spatial resolutions, and (ii) assess the uncertainties in the generated fields due to differences in quality and information content of AVHRR and MODIS sensors. Our results suggests that single scattering albedo and overall relative uncertainty of input surface reflectances and/ or NDVI are configurable parameters in governing the accuracy of LAI/FPAR retrievals from NDVI as an input. Single scattering albedo accounts for differences in both spatial and spectral resolution and the implementation of the algorithm is performed over test areas such that: (i) the consistency requirements are met, (ii) the retrieval index is maximized, and (iii) the difference between reference LAI and LAI obtained from AVHRR NDVI is minimized.

Recent empirical studies have shown that multi-angle data can be useful for predicting canopy height, but the physical reason for this correlation was not understood. The research presented here puts forth a physical explanation for this phenomenon. We employ the use of Radiative Transfer, more specifically canopy spectral invariants, which can decouple spectral and structural parameters in a vegetation canopy. As a case study we compare canopy heights predicted from a multivariate analysis of 28 (7 cameras* 4 bands) and LVIS canopy heights and a multivariate analysis of 7 directional escape probabilities and LVIS canopy heights. We find that the 7 directional escape probabilities can provide approximately the same amount of information about canopy height as 28 spectral/angular reflectances from AirMISR. Finally we speculate that multi-angle data does not allow for extraction of canopy height but in fact requires synergy between Lidar sensors.