SEARCH · Search NASA

Results for “spatial regression”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 235 records · Page 13

All sources NASA NTRS DOE OSTI

NASA NTRS2020

Review: Strategies for Using Satellite-Based Products in Modeling PM2.5 and Short-Term Pollution Episodes

Short-term air pollution episodes motivate improved understanding of the association between air pollution and acute morbidity and mortality episodes, and triggers required mitigation plans. A variety of methods have been employed to estimate exposure to air pollution episodes, including GIS-based dispersion models, interpolation between sparse monitoring sites, land-use regression models, optimization models, line- or area-dispersion plume models, and models using information from imaging satellites, often including land-use and meteorological variables. There has been increasing use of satellite-borne aerosol products for assessing short-term air quality events. They provide better spatial coverage, but currently at the price of low temporal coverage and rather crude spatial resolution. This brief review of using satellite data for modeling short-term air quality and pollution events. The review can be pursued as a practical guide for modeling air quality with satellite-based products, as it includes important questions that should be considered in both the study design as well as the model development stages. Progress in this field is detailed and includes published models and their use in environmental and health studies. Both current and future satellite-borne capabilities are covered. It also provides links to access and download relevant datasets and some R code for data processing and modeling.

Meytar Sorek-Hamer↗

NASA NTRS1983

Remote sensing investigations of wetland biomass and productivity for global biosystems research

Monitoring biomass of wetlands ecosystems can provide information on net primary production and on the chemical and physical status of wetland soils relative to anaerobic microbial transformation of key elements. Multispectral remote sensing techniques successfully estimated macrophytic biomass in wetlands systems. Regression models developed from ground spectral data for predicting Spartina alterniflora biomass over an entire growing season include seasonal variations in biomass density and illumination intensity. An independent set of biomass and spectral data were collected and the standing crop biomass and net primary productivity were estimated. The improved spatial, radiometric and spectral resolution of th LANDSAT-4 Thematic Mapper over the LANDSAT MSS can greatly enhance multispectral techniques for estimating wetlands biomass over large areas. These techniques can provide the biomass data necessary for global ecology studies.

Harkisky, M.↗

NASA NTRS2018

Calibration of the DSCOVR EPIC Visible and NIR Channels using MODIS Terra and Aqua Data and EPIC Lunar Observations

The unique position of the Deep Space Climate Observatory (DSCOVR) Earth Polychromatic Imaging Camera (EPIC) at the Lagrange 1 point makes an important addition to the data from currently operating low Earth orbit observing instruments. EPIC instrument does not have an onboard calibration facility. One approach to its calibration is to compare EPIC observations to the measurements from polar-orbiting radiometers. Moderate Resolution Imaging Spectroradiometer (MODIS) is a natural choice for such comparison due to its well-established calibration record and wide use in remote sensing. We use MODIS Aqua and Terra L1B 1km reflectances to infer calibration coefficients for four EPIC visible and NIR channels: 443, 551, 680 and 780 nm. MODIS and EPIC measurements made between June 2015 and 2016 are employed for comparison. We first identify favorable MODIS pixels with scattering angle matching temporarily collocated EPIC observations. Each EPIC pixel is then spatially collocated to a subset of the favorable MODIS pixels within 25 km radius. Standard deviation of the selected MODIS pixels as well as of the adjacent EPIC pixels is used to find the most homogeneous scenes. These scenes are then used to determine calibration coefficients using a linear regression between EPIC counts/sec and reflectances in the close MODIS spectral channels. We present thus inferred EPIC calibration coefficients and discuss sources of uncertainties. The lunar EPIC observations are used to calibrate EPIC O2 absorbing channels (688 and 764 nm), assuming that there is a small difference between moon reflectances separated by approx.10 nm in wavelength provided the calibration factors of the red (680 nm) and near-IR (780 nm) are known from comparison between EPIC and MODIS.

Geogdzhayev, Igor V.↗

DOE OSTI2022

Optimizing Piezoelectric Nanocomposites by High‐Throughput Phase‐Field Simulation and Machine Learning

Abstract Piezoelectric nanocomposites with oxide fillers in a polymer matrix combine the merit of high piezoelectric response of the oxides and flexibility as well as biocompatibility of the polymers. Understanding the role of the choice of materials and the filler‐matrix architecture is critical to achieving desired functionality of a composite towards applications in flexible electronics and energy harvest devices. Herein, a high‐throughput phase‐field simulation is conducted to systematically reveal the influence of morphology and spatial orientation of an oxide filler on the piezoelectric, mechanical, and dielectric properties of the piezoelectric nanocomposites. It is discovered that with a constant filler volume fraction, a composite composed of vertical pillars exhibits superior piezoelectric response and electromechanical coupling coefficient as compared to the other geometric configurations. An analytical regression is established from a linear regression‐based machine learning model, which can be employed to predict the performance of nanocomposites filled with oxides with a given set of piezoelectric coefficient, dielectric permittivity, and stiffness. This work not only sheds light on the fundamental mechanism of piezoelectric nanocomposites, but also offers a promising material design strategy for developing high‐performance polymer/inorganic oxide composite‐based wearable electronics.

42 ENGINEERING↗

DOE OSTI2023

Machine learning models inaccurately predict current and future high-latitude C balances

The high-latitude carbon (C) cycle is a key feedback to the global climate system, yet because of system complexity and data limitations, there is currently disagreement over whether the region is a source or sink of C. Recent advances in big data analytics and computing power have popularized the use of machine learning (ML) algorithms to upscale site measurements of ecosystem processes, and in some cases forecast the response of these processes to climate change. Due to data limitations, however, ML model predictions of these processes are almost never validated with independent datasets. To better understand and characterize the limitations of these methods, we develop an approach to independently evaluate ML upscaling and forecasting. We mimic data-driven upscaling and forecasting efforts by applying ML algorithms to different subsets of regional process-model simulation gridcells, and then test ML performance using the remaining gridcells. In this study, we simulate C fluxes and environmental data across Alaska using ecosys, a process-rich terrestrial ecosystem model, and then apply boosted regression tree ML algorithms to training data configurations that mirror and expand upon existing AmeriFLUX eddy-covariance data availability. We first show that a ML model trained using ecosys outputs from currently-available Alaska AmeriFLUX sites incorrectly predicts that Alaska is presently a modeled net C source. Increased spatial coverage of the training dataset improves ML predictions, halving the bias when 240 modeled sites are used instead of 15. However, even this more accurate ML model incorrectly predicts Alaska C fluxes under 21st century climate change because of changes in atmospheric CO 2 , litter inputs, and vegetation composition that have impacts on C fluxes which cannot be inferred from the training data. Our results provide key insights to future C flux upscaling efforts and expose the potential for inaccurate ML upscaling and forecasting of high-latitude C cycle dynamics.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS1999

The Fringe-Imaging Skin Friction Technique PC Application User's Manual

A personal computer application (CXWIN4G) has been written which greatly simplifies the task of extracting skin friction measurements from interferograms of oil flows on the surface of wind tunnel models. Images are first calibrated, using a novel approach to one-camera photogrammetry, to obtain accurate spatial information on surfaces with curvature. As part of the image calibration process, an auxiliary file containing the wind tunnel model geometry is used in conjunction with a two-dimensional direct linear transformation to relate the image plane to the physical (model) coordinates. The application then applies a nonlinear regression model to accurately determine the fringe spacing from interferometric intensity records as required by the Fringe Imaging Skin Friction (FISF) technique. The skin friction is found through application of a simple expression that makes use of lubrication theory to relate fringe spacing to skin friction.

Zilliac, Gregory G.↗

DOE OSTI2021

Physics-Informed Machine Learning Models for Predicting the Progress of Reactive-Mixing

This paper presents a physics-informed machine learning (ML) framework to construct reduced-order models (ROMs) for reactive-transport quantities of interest (QoIs) based on high-fidelity numerical simu-lations. QoIs include species decay, product yield, and degree of mixing. The ROMs for QoIs are applied to quantify and understand how the chemical species evolve over time. First, high-resolution datasets for constructing ROMs are generated by solving anisotropic reaction-di?usion equations using a non-negative finite element formulation for di?erent input parameters. The reactive-mixing model input parameters are: time-scale associated with flipping of velocity, spatial-scale controlling small/large vortex structures of velocity, perturbation parameter of the vortex-based velocity, anisotropic dispersion strength/contrast, and molecular diffusion. Second, random forests, F-test, and mutual information criterion are used to evaluate the importance of model inputs/features with respect to QoIs. We observed that anisotropic dispersion strength/contrast is the most important feature and time-scale associated with flipping of velocity is the least important feature. Third, Support Vector Machines (SVM) and Support Vector Regression (SVR) are used to construct ROMs based on the model inputs. The constructed SVR-ROMs are then used to predict scaling of QoIs. We also present estimates and inequalities on the QoIs, which inform that the species decay, mix, and produce in an exponential fashion. These inequalities also inform that a radial basis function is the most suitable kernel for the SVM/SVR models for QoIs. It is observed that R2-score for SVR-ROMs on unseen data is greater than 0.9, implying that the SVR-ROMs are able to predict the reaction-diffusion system state reasonably well. Finally, in terms of the computational cost, the proposed SVM-ROMs are O(107) times faster than running a high-fidelity finite element simulation for evaluating QoIs. This makes the proposed ML-based ROMs attractive for reactive-transport sensing and real-time monitoring applications as they are significantly faster yet reasonably accurate.

Mudunuru, Maruti K.↗

DOE OSTI2020

Indicator Patterns of Forced Change Learned by an Artificial Neural Network

Abstract Many problems in climate science require the identification of signals obscured by both the “noise” of internal climate variability and differences across models. Following previous work, we train an artificial neural network (ANN) to predict the year of a given map of annual‐mean temperature (or precipitation) from forced climate model simulations. This prediction task requires the ANN to learn forced patterns of change amidst a background of climate noise and model differences. We then apply a neural network visualization technique (layerwise relevance propagation) to visualize the spatial patterns that lead the ANN to successfully predict the year. These spatial patterns thus serve as “reliable indicators” of the forced change. The architecture of the ANN is chosen such that these indicators vary in time, thus capturing the evolving nature of regional signals of change. Results are compared to those of more standard approaches like signal‐to‐noise ratios and multilinear regression in order to gain intuition about the reliable indicators identified by the ANN. We then apply an additional visualization tool (backward optimization) to highlight where disagreements in simulated and observed patterns of change are most important for the prediction of the year. This work demonstrates that ANNs and their visualization tools make a powerful pair for extracting climate patterns of forced change.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS1994

Determining the response of sea level to atmospheric pressure forcing using TOPEX/POSEIDON data

The static response of sea level to the forcing of atmospheric pressure, the so-called inverted barometer (IB) effect, is investigated using TOPEX/POSEIDON data. This response, characterized by the rise and fall of sea level to compensate for the change of atmospheric pressure at a rate of -1 cm/mbar, is not associated with any ocean currents and hence is normally treated as an error to be removed from sea level observation. Linear regression and spectral transfer function analyses are applied to sea level and pressure to examine the validity of the IB effect. In regions outside the tropics, the regression coefficient is found to be consistently close to the theoretical value except for the regions of western boundary currents, where the mesoscale variability interferes with the IB effect. The spectral transfer function shows near IB response at periods of 30 degrees is -0.84 +/- 0.29 cm/mbar (1 standard deviation). The deviation from = 1 cm /mbar is shown to be caused primarily by the effect of wind forcing on sea level, based on multivariate linear regression model involving both pressure and wind forcing. The regression coefficient for pressure resulting from the multivariate analysis is -0.96 +/- 0.32 cm/mbar. In the tropics the multivariate analysis fails because sea level in the tropics is primarily responding to remote wind forcing. However, after removing from the data the wind-forced sea level estimated by a dynamic model of the tropical Pacific, the pressure regression coefficient improves from -1.22 +/- 0.69 cm/mbar to -0.99 +/- 0.46 cm/mbar, clearly revealing an IB response. The result of the study suggests that with a proper removal of the effect of wind forcing the IB effect is valid in most of the open ocean at periods longer than 20 days and spatial scales larger than 500 km.

Fu, Lee-Lueng↗

DOE OSTI2020

Data-driven surrogates for high dimensional models using Gaussian process regression on the Grassmann manifold

This paper introduces a surrogate modeling scheme based on Grassmannian manifold learning to be used for cost-efficient predictions of high-dimensional stochastic systems. The method exploits subspace-structured features of each solution by projecting it onto a Grassmann manifold. This point-wise linear dimensionality reduction harnesses the structural information to assess the similarity between solutions at different points in the input parameter space. The method utilizes a solution clustering approach in order to identify regions of the parameter space over which solutions are sufficiently similarly such that they can be interpolated on the Grassmannian. In this clustering, the reduced-order solutions are partitioned into disjoint clusters on the Grassmann manifold using the eigen-structure of properly defined Grassmannian kernels and, the Karcher mean of each cluster is estimated. Then, the points in each cluster are projected onto the tangent space with origin at the corresponding Karcher mean using the exponential mapping. For each cluster, a Gaussian process regression model is trained that maps the input parameters of the system to the reduced solution points of the corresponding cluster projected onto the tangent space. Using this Gaussian process model, the full-field solution can be efficiently predicted at any new point in the parameter space. In certain cases, the solution clusters will span disjoint regions of the parameter space. In such cases, for each of the solution clusters we utilize a second, density-based spatial clustering to group their corresponding input parameter points in the Euclidean space. The proposed method is applied to two numerical examples. Here, the first is a nonlinear stochastic ordinary differential equation with uncertain initial conditions where the surrogate is used to predict the time history solution. The second involves modeling of plastic deformation in a model amorphous solid using the Shear Transformation Zone theory of plasticity, where the proposed surrogate is used to predict the full strain field of a material specimen under large shear strains.

42 ENGINEERING↗

NASA NTRS1984

Evaluation of spatial, radiometric and spectral Thematic Mapper performance for coastal studies

The effect different wetland plant canopies have upon observed reflectance in Thematic Mapper bands is examined. The three major vegetation canopy types (broadleaf, gramineous and leafless) produce unique spectral responses for a similar quantity of live biomass. Biomass estimates computed from spectral data were most similar to biomass estimates determined from harvest data when models developed for a specific canopy were used. Precise determination of regression coefficients for each canopy type and modeling changes in the coefficients with various combinations of canopy types are being tested. The multispectral band scanner vegetation index estimates are very similar to the vegetation index estimates.

Klemas, V.↗

NASA NTRS2021

Capturing the Footprints of Ground Motion in the Spatial Distribution of Rainfall-Induced Landslides

The coupled effect of earthquakes and rainfall is rarely investigated in landslide susceptibility assessments although it could be crucial to predict landslide occurrences. This is even more critical in the context of early warning systems and especially in cases of extreme precipitation regimes in post-seismic conditions, where the rock masses are already damaged due to the ground shaking. Here, we investigate this concept by accounting for the legacy of seismic ground shaking in rainfall-induced landslide (RFIL) scenarios. We do this to identify whether ground shaking plays a role in the susceptibility to post-seismic rainfall-induced landslides and to identify whether this legacy effect persists through time. With this motivation, we use binary logistic regression and examine time series of landslides associated with four earthquakes occurred in Indonesia: 2012 Sulawesi (M(sub w) = 6.3), 2016 Reuleut (M(sub w) = 6.5), 2017 Kasiguncu (M(sub w) = 6.6) and 2018 Palu (M(sub w) = 7.5) earthquakes. The dataset includes one co-seismic and three post-seismic landslide inventories for each earthquake. We use the peak ground acceleration map of the last strongest earthquake in each case as a predisposing factor of landslides representing the effect of ground shaking. We observe that, at least for the study areas under consideration and in a probabilistic context, the earthquake legacy contributes to increase the post-seismic RFIL susceptibility. This positive contribution decays through time. Specifically, we observe that ground motion is a significant predisposing factor controlling the spatial distribution of RFIL in the post-seismic period 110 days after an earthquake. We also show that this effect dissipates within 3 years at most.

Landslides↗

NASA NTRS2012

An Automated Algorithm to Screen Massive Training Samples for a Global Impervious Surface Classification

An algorithm is developed to automatically screen the outliers from massive training samples for Global Land Survey - Imperviousness Mapping Project (GLS-IMP). GLS-IMP is to produce a global 30 m spatial resolution impervious cover data set for years 2000 and 2010 based on the Landsat Global Land Survey (GLS) data set. This unprecedented high resolution impervious cover data set is not only significant to the urbanization studies but also desired by the global carbon, hydrology, and energy balance researches. A supervised classification method, regression tree, is applied in this project. A set of accurate training samples is the key to the supervised classifications. Here we developed the global scale training samples from 1 m or so resolution fine resolution satellite data (Quickbird and Worldview2), and then aggregate the fine resolution impervious cover map to 30 m resolution. In order to improve the classification accuracy, the training samples should be screened before used to train the regression tree. It is impossible to manually screen 30 m resolution training samples collected globally. For example, in Europe only, there are 174 training sites. The size of the sites ranges from 4.5 km by 4.5 km to 8.1 km by 3.6 km. The amount training samples are over six millions. Therefore, we develop this automated statistic based algorithm to screen the training samples in two levels: site and scene level. At the site level, all the training samples are divided to 10 groups according to the percentage of the impervious surface within a sample pixel. The samples following in each 10% forms one group. For each group, both univariate and multivariate outliers are detected and removed. Then the screen process escalates to the scene level. A similar screen process but with a looser threshold is applied on the scene level considering the possible variance due to the site difference. We do not perform the screen process across the scenes because the scenes might vary due to the phenology, solar-view geometry, and atmospheric condition etc. factors but not actual landcover difference. Finally, we will compare the classification results from screened and unscreened training samples to assess the improvement achieved by cleaning up the training samples. Keywords:

Tan, Bin↗

NASA NTRS2019

Evaluating Combinations of Sentinel-2 Data and Machine-Learning Algorithms for Mangrove Mapping in West Africa

Creating a national baseline for natural resources, such as mangrove forests, and monitoring them regularly often requires a consistent and robust methodology. With freely available satellite data archives and cloud computing resources, it is now more accessible to conduct such large-scale monitoring and assessment. Yet, few studies examine the reproducibility of such mangrove monitoring frameworks, especially in terms of generating consistent spatial extent. Our objective was to evaluate a combination of image processing approaches to classify mangrove forests along the coast of Senegal and The Gambia. We used freely available global satellite data (Sentinel-2), and cloud computing platform (Google Earth Engine) to run two machine learning algorithms, random forest (RF), and classification and regression trees (CART). We calibrated and validated the algorithms using 800 reference points collected using high-resolution images. We further re-ran 10 iterations for each algorithm, utilizing unique subsets of the initial training data. While all iterations resulted in thematic mangrove maps with over 90% accuracy, the mangrove extent ranges between 827-2807 km2 for Senegal and 245-1271 km2 for The Gambia with one outlier for each country. We further report "Places of Agreement" (PoA) to identify areas where all iterations for both methods agree (506.6 km2 and 129.6 km2 for Senegal and The Gambia, respectively), thus have a high confidence in predicting mangrove extent. While we acknowledge the time- and cost-effectiveness of such methods for the landscape managers, we recommend utilizing them with utmost caution, as well as post-classification on-the-ground checks, especially for decision making.

Mondal, Pinki↗

DOE OSTI2021

Probabilistic Day-Ahead Forecasting Using an Analog Ensemble Approach for Wind Farm Grid Services

Wind resource assessment and wind power forecasting are used in research and industry to anticipate future power output at scales ranging from individual wind turbines to entire wind farms. Probabilistic day-ahead wind forecasting is useful for anticipating how a wind farm could potentially participate in the day-ahead market by providing upper and lower bounds for expected power generation, thus informing grid operators of its uncertainty. Understanding this uncertainty is part of a larger project focused on building a platform that combines efforts in weather forecasting, aerodynamic and economic modeling to create maximum value of a wind plant to better provide services to the grid. This effort is also known as the Atmosphere to Electrons to Grid (A2E2G) project. One method for producing a probabilistic forecast is through the analog ensemble approach (Delle Monache et al., 2011). This method leverages historical forecasts and their corresponding observations as a training data set from which future forecasts can be made. For some future forecast, the most similar historical forecasts (analogs) are identified on a regular time basis such as once per a 3-hour window. The most similar analogs, based on a metric such as root mean square error (RMSE), are recorded and their corresponding verifying observations are used as an ensemble member for this future forecast. Prior work in this area demonstrates improvements over raw Numerical Weather Prediction (NWP) forecasts and shows skill similar to techniques such as logistic regression and machine learning (Delle Monache et al., 2013; Alessandrini et al., 2015). Here, we take the High-Resolution Rapid Refresh model (HRRR) day-ahead forecast (0-36 hours) to create a probabilistic day-ahead forecast using an analog ensemble approach. The HRRR has an hourly temporal resolution, with a spatial resolution of 3 km. The 12 UTC HRRR model run is downloaded every day for one year from August 2019 - July 2020, with the first 11 months serving as a bank of analogs from which the forecasting algorithm can create a probabilistic forecast. Once downloaded, the original HRRR forecast is temporally interpolated to 5-minutes, aligning with both the temporal resolution of the observations as well as the timescale relevant for day-ahead power forecasts. The forecast is validated at the M2 tower at the Flatirons Campus of the National Renewable Energy Laboratory (NREL) at a typical wind turbine height of 80 m. Variables such as wind speed, wind direction, and turbulence intensity are incorporated into the probabilistic forecast model and weighted according to their relative importance to the forecast. Based on metrics such as mean bias error (MBE), mean absolute error (MAE), and root mean square error, the analog ensemble forecast outperforms the raw HRRR forecast during the testing period of July 2020. Figure 1 illustrates an example day-ahead forecast compared against the verifying observations. The general variability and ramps are captured throughout the day, with potential to further improve the analog ensemble model through machine learning techniques.

numerical weather prediction↗

DOE OSTI2021

Excess deaths reveal the true spatial, temporal and demographic impact of COVID-19 on mortality in Ecuador

Accepted Background In early 2020, Ecuador reported one of the highest surges of per capita deaths across the globe. Methods We collected a comprehensive dataset containing individual death records between 2015 and 2020, from the Ecuadorian National Institute of Statistics and Census and the Ecuadorian Ministry of Government. We computed the number of excess deaths across time, geographical locations and demographic groups using Poisson regression methods. Results Between 1 January and 23 September 2020, the number of excess deaths in Ecuador was 36 402 [95% confidence interval (CI): 35 762–36 827] or 208 per 100 000 people, which is 171% of the expected deaths in that period in a typical year. Only 20% of the excess deaths are attributable to confirmed COVID-19 deaths. Strikingly, in provinces that were most affected by COVID-19 such as Guayas and Santa Elena, the all-cause deaths are more than double the expected number of deaths that would have occurred in a normal year. The extent of excess deaths in men is higher than in women, and the number of excess deaths increases with age. Indigenous populations had the highest level of excess deaths among all ethnic groups. Conclusions Overall, the exceptionally high level of excess deaths in Ecuador highlights the enormous burden and heterogeneous impact of COVID-19 on mortality, especially in older age groups and Indigenous populations in Ecuador, which was not fully revealed by COVID-19 death counts. Together with the limited testing in Ecuador, our results suggest that the majority of the excess deaths were likely to be undocumented COVID-19 deaths.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2024

A mapped dataset of surface ocean acidification indicators in large marine ecosystems of the United States

Mapped monthly data products of surface ocean acidification indicators from 1998 to 2022 on a 0.25° by 0.25° spatial grid have been developed for eleven U.S. large marine ecosystems (LMEs). The data products were constructed using observations from the Surface Ocean CO 2 Atlas, co-located surface ocean properties, and two types of machine learning algorithms: Gaussian mixture models to organize LMEs into clusters of similar environmental variability and random forest regressions (RFRs) that were trained and applied within each cluster to spatiotemporally interpolate the observational data. The data products, called RFR-LMEs, have been averaged into regional timeseries to summarize the status of ocean acidification in U.S. coastal waters, showing a domain-wide carbon dioxide partial pressure increase of 1.4 ± 0.4 μatm yr -1 and pH decrease of 0.0014 ± 0.0004 yr -1 . RFR-LMEs have been evaluated via comparisons to discrete shipboard data, fixed timeseries, and other mapped surface ocean carbon chemistry data products. Regionally averaged timeseries of RFR-LME indicators are provided online through the NOAA National Marine Ecosystem Status web portal.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Short-lead seasonal precipitation forecast in northeastern Brazil using an ensemble of artificial neural networks

This study assesses the deterministic and probabilistic forecasting skill of a 1-month-lead ensemble of Artificial Neural Networks (EANN) based on low-frequency climate oscillation indices. The predictand is the February-April (FMA) rainfall in the Brazilian state of Ceará, which is a prominent subject in climate forecasting studies due to its high seasonal predictability. Additionally, the study proposes combining the EANN with dynamical models into a hybrid multi-model ensemble (MME). The forecast verification is carried out through a leave-one-out cross-validation based on 40 years of data. The EANN forecasting skill is compared with traditional statistical models and the dynamical models that compose Ceará’s operational seasonal forecasting system. A spatial comparison showed that the EANN was among the models with the smallest Root Mean Squared Error (RMSE) and Ranked Probability Score (RPS) in most regions. Moreover, the analysis of the area-aggregated reliability showed that the EANN is better calibrated than the individual dynamical models and has better resolution than Multinomial Logistic Regression for above-normal (AN) and below-normal (BN) categories. It is also shown that combining the EANN and dynamical models into a hybrid MME reduces the overconfidence of the extreme categories observed in a dynamically-based MME, improving the reliability of the forecasting system.

54 ENVIRONMENTAL SCIENCES↗