A. R. Michaelis

Prediction of Continental-Scale Evapotranspiration by Combining MODIS and AmeriFlux Data Through Support Vector Machine

Application of remote sensing data to extrapolate evapotranspiration (ET) measured at eddy covariance flux towers is a potentially powerful method to estimate continental-scale ET. In support of this concept, we used meteorological and flux data from the AmeriFlux network and an inductive machine learning technique called support vector machine (SVM) to develop a predictive ET model. The model was then applied to the conterminous U.S. In this process, we first trained the SVM to predict 2000-2002 ET measurements from 25 AmeriFlux sites using three remotely sensed variables [land surface temperature, enhanced vegetation index (EVI), and land cover] and one ground-measured variable (surface shortwave radiation). Second, we evaluated the model performance by predicting ET for 19 flux sites in 2003. In this independent evaluation, the SVM predicted ET with a root-mean-square error (rmse) of 0.62 mm/day (approximately 23% of the mean observed values) and an R2 of 0.75. The rmse from SVM was significantly smaller than that from neural network and multiple-regression approaches in a cross-validation experiment. Among the explanatory variables, EVI was the most important factor. Indeed, removing this variable induced an rmse increase from 0.54 to 0.77 mm/day. Third, with forcings from remote sensing data alone, we used the SVM model to predict the spatial and temporal distributions of ET for the conterminous U.S. for 2004. The SVM model captured the spatial and temporal variations of ET at a continental scale

A flexible, integrated system for generating meteorological surfaces derived from point sources across multiple geographic scales

Abstract The generation of meteorological surfaces from point-source data is a difficult but necessary step required for modeling ecological and hydrological processes across landscapes. To date, procedures to acquire, transform, and display meteorological information geographically have been specifically tailored to individual studies. Here we offer a flexible, integrated system that employs a relational database to store point information, a modular system incorporating a choice of weather data interpolation methods, and a matrix inversion method that speeds computer calculations to display information on grids of any specified size, all with minimal user intervention. We demonstrate the power of this integrated approach by cross-validating projected daily meteorological surfaces derived from ∼1200 weather stations distributed across the continental United States for a year. We performed cross-validations for five meteorological variables (solar radiation, minimum and maximum temperatures, humidity, and precipitation) with a truncated Gaussian filter, ordinary kriging and inverse distance weighting and achieved comparable success among all interpolation methods. Cross-validation computation time for ordinary kriging was reduced from 1 h to 3 min when we incorporated the matrix inversion method. We demonstrated the systems flexibility by displaying results at 8-km resolution for the continental USA and at one-degree resolution for the globe.

Downscaled Climate Projections Suitable for Resource Management

The general circulation model (GCM) experiments conducted under the Coupled Model Intercomparison Project Phase 5 (CMIP5) [Taylor et al., 2012], which is being conducted in preparation for the Intergovernmental Panel on Climate Changes Fifth Assessment Report, provide fundamental data sets for assessing the effects of global climate change. However, efforts to assess regional or local effects of the projected changes in climate are often impeded by the coarse spatial resolution of the GCM outputs, as well as potential local or regional biases in GCM outputs [Fowler et al., 2007].

A Semiautomated Probabilistic Framework for Tree-Cover Delineation From 1-m NAIP Imagery Using a High-Performance Computing Architecture

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.

River Temperature Forecasting: A Coupled-Modeling Framework for Management of River Habitat

Humans have substantially altered the thermal regimes of freshwater habitats worldwide, with significant environmental consequences. There is a critical need for a comprehensive modeling framework for forecasting the downstream impacts of two of the most common anthropogenic structures that alter river water temperatures: 1) dams that selectively release water from thermally stratified reservoirs, and 2) power generating stations and industrial plants that use river water for once-through cooling. These facilities change the thermal dynamics of the downstream waters through a complex interaction of water release volume and temperature and the subsequent exchange with the environment downstream. In order to stay within the downstream temperature limits imposed by regulatory agencies, managers must monitor not just release volumes and temperatures, but also need to be able to forecast the thermal impacts of their day-to-day operations on habitat which may be hundreds of kilometers downstream. Here we describe a coupled modeling framework that links mesoscale weather and ecological models to generate inputs for a physically-based water temperature model for monitoring and forecasting river temperatures downstream from these facilities at fine spatiotemporal scales. We provide an example of how this modeling framework is being applied to a water allocation decision support system (DSS) for the management of Endangered Species Act (ESA) listed salmon species in the Sacramento River in California.

Effects of spatial and temporal climatic variability on terrestrial carbon and water fluxes in the Pacific Northwest, USA

The Pacific Northwest (PNW) of the conterminous United States is characterized by large variations in climate and topography, and provides an ideal geographic domain for studying interactions between regional climate and vegetation dynamics. We examined vegetation carbon (C) and water dynamics along PNW climate and topographic gradients using a process-based biogeochemical model, BIOME-BGC, the algorithms of which form bases for a fully-prognostic treatment of carbon and nitrogen cycles in Land Community Model (CLM). Simulation experiments were used to (1) analyze spatial and temporal variability of terrestrial carbon (C) stocks and flux, (2) investigate primary climatic variables controlling the variability, and (3) predict effects of future climate projections on vegetation productivity and water flux variables including evapotranspiration and water supply. The model experiments focused on two 18-year (1980-1997 and 2088-2105) simulations using future climate predictions for A2 (+4.2 ^oC, -7% precipitation) and B2 (1.6 ^oC, +11% precipitation) emissions scenarios through year 2100. Our results show large west to east spatial variations in C and water fluxes and C stocks associated with regional topography and distance from coastal areas. Interannual variability of net primary productivity (NPP) and evapotranspiration (ET) are 57% and 33%, respectively, of the 18-year mean annual fluxes for 1980-1997. The annual NPP and ET are positively correlated with precipitation but inversely proportional to vapor pressure deficit; this suggests that modeled NPP and ET are predominantly water limited in the PNW. The A2 scenario results in higher NPP and ET of 23% and 10%, respectively, and 15% lower water outflow. The B2 scenario results in higher NPP and ET of 12% and 15%, respectively, and 2% lower water outflow, despite projected increases in precipitation. Simulation experiments indicate that most PNW ecosystems are water limited, and that annual water outflow will decrease under both drier (A2) and wetter (B2) scenarios. However, higher elevations with high snowpacks of long duration may buffer the loss of water resources in some areas, even if precipitation is lower.

DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution

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.

Selection and quality assessment of Landsat data for the North American forest dynamics forest history maps of the US

ABSTRACT Using the NASA Earth Exchange platform, the North American Forest Dynamics (NAFD) project mapped forest history wall-to-wall, annually for the contiguous US (1986–2010) using the Vegetation Change Tracker algorithm. As with any effort to identify real changes in remotely sensed time-series, data gaps, shifts in seasonality, misregistration, inconsistent radiometry and cloud contamination can be sources of error. We discuss the NAFD image selection and processing stream (NISPS) that was designed to minimize these sources of error. The NISPS image quality assessments highlighted issues with the Landsat archive and metadata including inadequate georegistration, unreliability of the pre-2009 L5 cloud cover assessments algorithm, missing growing-season imagery and paucity of clear views. Assessment maps of Landsat 5–7 image quantities and qualities are presented that offer novel perspectives on the growing-season archive considered for this study. Over 150,000+ Landsat images were considered for the NAFD project. Optimally, one high quality cloud-free image in each year or a total of 12,152 images would be used. However, to accommodate data gaps and cloud/shadow contamination 23,338 images were needed. In 220 specific path-row image years no acceptable images were found resulting in data gaps in the annual national map products.

A physically based approach in retrieving vegetation Leaf Area Index from Landsat surface reflectance data

In this study, we aim to generate global 30-m Leaf Area Index (LAI) from Landsat surface reflectance data using the radiative transfer theory of canopy spectral invariants which facilitates parameterization of the canopy spectral bidirectional reflectance factor (BRF). Furthermore, canopy spectral invariants introduce an efficient way for incorporating multiple bands for retrieving LAI. We incorporate a 3-band retrieval scheme including the Red, NIR and SWIR bands, the SWIR band being specifically useful in low LAI regions and thus compensating for background effects. The initial results have satisfactory agreement with MODIS LAI, although with spatially more detailed structure and variability. A future exercise will be to introduce field measured LAI estimates to minimize the differences between model-simulated LAIs and in-situ observations.

Deploying a quantum annealing processor to detect tree cover in aerial imagery of California

Quantum annealing is an experimental and potentially breakthrough computational technology for handling hard optimization problems, including problems of computer vision. We present a case study in training a production-scale classifier of tree cover in remote sensing imagery, using early-generation quantum annealing hardware built by D-wave Systems, Inc. Beginning within a known boosting framework, we train decision stumps on texture features and vegetation indices extracted from four-band, one-meter-resolution aerial imagery from the state of California. We then impose a regulated quadratic training objective to select an optimal voting subset from among these stumps. The votes of the subset define the classifier. For optimization, the logical variables in the objective function map to quantum bits in the hardware device, while quadratic couplings encode as the strength of physical interactions between the quantum bits. Hardware design limits the number of couplings between these basic physical entities to five or six. To account for this limitation in mapping large problems to the hardware architecture, we propose a truncation and rescaling of the training objective through a trainable metaparameter. The boosting process on our basic 108- and 508-variable problems, thus constituted, returns classifiers that incorporate a diverse range of color- and texture-based metrics and discriminate tree cover with accuracies as high as 92% in validation and 90% on a test scene encompassing the open space preserves and dense suburban build of Mill Valley, CA.