Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “algorithm selection”

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.

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

At least 1,369 records · Page 76

The CONTIN algorithm and its application to determine the size distribution of microgel suspensions

We review a powerful regularization method, known as CONTIN, for obtaining the size distribution of colloidal suspensions from dynamic light scattering data. We show that together with the so-called L-curve criterion for selecting the optimal regularization parameter, the method correctly describes the average size and size distribution of microgel suspensions independently characterized using small-angle neutron scattering. In contrast, we find that when using the default regularization process, where the regularizer is selected via the “probability to reject” method, the results are not as satisfactory.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Feature selection based on PCA and the transient identification in nuclear power plants - 289

For the sake of enhancing safety and achieving more economic benefits, it is very important to timely and correctly identify the transients during the operation of nuclear power plants. There are thousands of monitoring signals in a nuclear power plant. It is difficult to complete the calculations timely if all the related signals are chosen for an algorithm of transient identification. Moreover, the irrelevant state variables may also do harm to the correctness of the identification process. Thus, one of the key tasks for transient identification is to figure out effective methods to extract the features from state variables. In this paper, we will address an efficient method for feature selection and transient identification based on principal component analysis (PCA). The effectiveness of the method are illustrated by the tests based on the data from the simulator designed for the High Temperature gas-cooled Reactor Pebble-bed Module (HTR-PM), which is under construction in China and the construction will be completed around 2017. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Single-energy computed tomography-based pulmonary perfusion imaging: Proof-of-principle in a canine model

Purpose: Radiotherapy (RT) that selectively avoids irradiating highly functional lung regions may reduce pulmonary toxicity, which is substantial in lung cancer RT. Single-energy computed tomography (CT) pulmonary perfusion imaging has several advantages (e.g., higher resolution) over other modalities and has great potential for widespread clinical implementation, particularly in RT. The purpose of this study was to establish proof-of-principle for single-energy CT perfusion imaging. Methods: Single-energy CT perfusion imaging is based on the following: (1) acquisition of end-inspiratory breath-hold CT scans before and after intravenous injection of iodinated contrast agents, (2) deformable image registration (DIR) for spatial mapping of those two CT image data sets, and (3) subtraction of the precontrast image data set from the postcontrast image data set, yielding a map of regional Hounsfield unit (HU) enhancement, a surrogate for regional perfusion. In a protocol approved by the institutional animal care and use committee, the authors acquired CT scans in the prone position for a total of 14 anesthetized canines (seven canines with normal lungs and seven canines with diseased lungs). The elastix algorithm was used for DIR. The accuracy of DIR was evaluated based on the target registration error (TRE) of 50 anatomic pulmonary landmarks per subject for 10 randomly selected subjects as well as on singularities (i.e., regions where the displacement vector field is not bijective). Prior to perfusion computation, HUs of the precontrast end-inspiratory image were corrected for variation in the lung inflation level between the precontrast and postcontrast end-inspiratory CT scans, using a model built from two additional precontrast CT scans at end-expiration and midinspiration. The authors also assessed spatial heterogeneity and gravitationally directed gradients of regional perfusion for normal lung subjects and diseased lung subjects using a two-sample two-tailed t-test. Results: The mean TRE (and standard deviation) was 0.6 ± 0.7 mm (smaller than the voxel dimension) for DIR between pre contrast and postcontrast end-inspiratory CT image data sets. No singularities were observed in the displacement vector fields. The mean HU enhancement (and standard deviation) was 37.3 ± 10.5 HU for normal lung subjects and 30.7 ± 13.5 HU for diseased lung subjects. Spatial heterogeneity of regional perfusion was found to be higher for diseased lung subjects than for normal lung subjects, i.e., a mean coefficient of variation of 2.06 vs 1.59 (p = 0.07). The average gravitationally directed gradient was strong and significant (R{sup 2} = 0.99, p < 0.01) for normal lung dogs, whereas it was moderate and nonsignificant (R{sup 2} = 0.61, p = 0.12) for diseased lung dogs. Conclusions: This canine study demonstrated the accuracy of DIR with subvoxel TREs on average, higher spatial heterogeneity of regional perfusion for diseased lung subjects than for normal lung subjects, and a strong gravitationally directed gradient for normal lung subjects, providing proof-of-principle for single-energy CT pulmonary perfusion imaging. Further studies such as comparison with other perfusion imaging modalities will be necessary to validate the physiological significance.

60 APPLIED LIFE SCIENCES↗

A Selectivity based approach to Continuous Pattern Detection in Streaming Graphs

Cyber security is one of the most significant technical challenges in current times. Detecting adversarial activities, prevention of theft of intellectual properties and customer data is a high priority for corporations and government agencies around the world. Cyber defenders need to analyze massive-scale, high-resolution network flows to identify, categorize, and mitigate attacks involving networks spanning institutional and national boundaries. Many of the cyber attacks can be described as subgraph patterns, with prominent examples being insider infiltrations (path queries), denial of service (parallel paths) and malicious spreads (tree queries). This motivates us to explore subgraph matching on streaming graphs in a continuous setting. The novelty of our work lies in using the subgraph distributional statistics collected from the streaming graph to determine the query processing strategy. We introduce a ``Lazy Search" algorithm where the search strategy is decided on a vertex-to-vertex basis depending on the likelihood of a match in the vertex neighborhood. We also propose a metric named ``Relative Selectivity" that is used to select between different query processing strategies. Our experiments performed on real online news, network traffic stream and a synthetic social network benchmark demonstrate 10-100x speedups over non-incremental, selectivity agnostic approaches.

dynamic graphs↗

SU-E-J-30: Benchmark Image-Based TCP Calculation for Evaluation of PTV Margins for Lung SBRT Patients

Purpose: Tumor control probability (TCP) calculated with accumulated radiation doses may help design appropriate treatment margins. Image registration errors, however, may compromise the calculated TCP. The purpose of this study is to develop benchmark CT images to quantify registration-induced errors in the accumulated doses and their corresponding TCP. Methods: 4DCT images were registered from end-inhale (EI) to end-exhale (EE) using a “demons” algorithm. The demons DVFs were corrected by an FEM model to get realistic deformation fields. The FEM DVFs were used to warp the EI images to create the FEM-simulated images. The two images combined with the FEM DVF formed a benchmark model. Maximum intensity projection (MIP) images, created from the EI and simulated images, were used to develop IMRT plans. Two plans with 3 and 5 mm margins were developed for each patient. With these plans, radiation doses were recalculated on the simulated images and warped back to the EI images using the FEM DVFs to get the accumulated doses. The Elastix software was used to register the FEM-simulated images to the EI images. TCPs calculated with the Elastix-accumulated doses were compared with those generated by the FEM to get the TCP error of the Elastix registrations. Results: For six lung patients, the mean Elastix registration error ranged from 0.93 to 1.98 mm. Their relative dose errors in PTV were between 0.28% and 6.8% for 3mm margin plans, and between 0.29% and 6.3% for 5mm-margin plans. As the PTV margin reduced from 5 to 3 mm, the mean TCP error of the Elastix-reconstructed doses increased from 2.0% to 2.9%, and the mean NTCP errors decreased from 1.2% to 1.1%. Conclusion: Patient-specific benchmark images can be used to evaluate the impact of registration errors on the computed TCPs, and may help select appropriate PTV margins for lung SBRT patients.

60 APPLIED LIFE SCIENCES↗

Implementation and analysis of quantum computing application to Higgs boson reconstruction at the large Hadron Collider

With the advent of the High-Luminosity Large Hadron Collider (HL-LHC) era, high energy physics (HEP) event selection will require new approaches to rapidly and accurately analyze vast databases. The current study addresses the enormity of HEP databases in an unprecedented manner—a quantum search using Grover’s Algorithm (GA) on an unsorted database, ATLAS Open Data, from the ATLAS detector. A novel method to identify rare events at 13 TeV in CERN’s LHC using quantum computing (QC) is presented. As indicated by the Higgs boson decay channel H→ZZ*→4l , the detection of four leptons in one event may be used to reconstruct the Higgs boson and, more importantly, evince Higgs boson decay to some new phenomena, such as H→ZZ d →4l . Searching the dataset for collisions resulting in detection of four leptons using a Jupyter Notebook, a classical simulation of GA, and several quantum computers with multiple qubits, the current application was found to make the proper selection in the unsorted dataset. Quantum search efficacy was analyzed for the incoming HL-LHC by implementing the QC method on multiple classical simulators and IBM’s quantum computers with the IBM Qiskit Open Source Software. The current QC application provides a novel, high-efficiency alternative to classical database searches, demonstrating its potential utility as a rapid and increasingly accurate search method in HEP.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A Selectivity based approach to Continuous Pattern Detection in Streaming Graphs

Cyber security is one of the most significant technical challenges in current times. Detecting adversarial activities, prevention of theft of intellectual properties and customer data is a high priority for corporations and government agencies around the world. Cyber defenders need to analyze massive-scale, high-resolution network flows to identify, categorize, and mitigate attacks involving net- works spanning institutional and national boundaries. Many of the cyber attacks can be described as subgraph patterns, with promi- nent examples being insider infiltrations (path queries), denial of service (parallel paths) and malicious spreads (tree queries). This motivates us to explore subgraph matching on streaming graphs in a continuous setting. The novelty of our work lies in using the subgraph distributional statistics collected from the streaming graph to determine the query processing strategy. We introduce a “Lazy Search" algorithm where the search strategy is decided on a vertex-to-vertex basis depending on the likelihood of a match in the vertex neighborhood. We also propose a metric named “Relative Selectivity" that is used to se- lect between different query processing strategies. Our experiments performed on real online news, network traffic stream and a syn- thetic social network benchmark demonstrate 10-100x speedups over selectivity agnostic approaches.

graphs, cybersecurity↗

Individual pore and interconnection size analysis of macroporous ceramic scaffolds using high-resolution X-ray tomography

The pore interconnection size of β-tricalcium phosphate scaffolds plays an essential role in the bone repair process. Although, the μCT technique is widely used in the biomaterial community, it is rarely used to measure the interconnection size because of the lack of algorithms. In addition, discrete nature of the μCT introduces large systematic errors due to the convex geometry of interconnections. We proposed, verified and validated a novel pore-level algorithm to accurately characterize the individual pores and interconnections. Specifically, pores and interconnections were isolated, labeled, and individually analyzed with high accuracy. The technique was verified thoroughly by visually inspecting and verifying over 3474 properties of randomly selected pores. This extensive verification process has passed a one-percent accuracy criterion. Scanning errors inherent in the discretization, which lead to both dummy and significantly overestimated interconnections, have been examined using computer-based simulations and additional high-resolution scanning. Then accurate correction charts were developed and used to reduce the scanning errors. Only after the corrections, both the μCT and SEM-based results converged, and the novel algorithm was validated. Material scientists with access to all geometrical properties of individual pores and interconnections, using the novel algorithm, will have a more-detailed and accurate description of the substitute architecture and a potentially deeper understanding of the link between the geometric and biological interaction. - Highlights: •An algorithm is developed to analyze individually all pores and interconnections. •After pore isolating, the discretization errors in interconnections were corrected. •Dummy interconnections and overestimated sizes were due to thin material walls. •The isolating algorithm was verified through visual inspection (99% accurate). •After correcting for the systematic errors, algorithm was validated successfully.

36 MATERIALS SCIENCE↗

Tile-Based Fisher-Ratio Software for Improved Feature Selection Analysis of Comprehensive Two-Dimensional Gas Chromatography Time-of-Flight Mass Spectrometry Data

Two-dimensional (2D) gas chromatography coupled with time-of-flight mass spectrometry (GC × GC – TOFMS) is a highly capable instrumental platform that produces complex and information-rich multi-dimensional chemical data. The complex data can be overwhelming, especially when many samples (of various sample classes) are analyzed with multiple injections for each sample. Thus, the data must be analyzed in such a way to extract the most meaningful information. The pixel-based and peak table-based algorithmic use of Fisher ratios has been used successfully in the past to reduce the multi-dimensional data down to those chemical compounds that are changing between classes relative to those that are not (i.e., chemical feature selection). We report on the initial development of a computationally fast novel tile-based Fisher-ratio software that addresses challenges due to 2D retention time misalignment without explicitly aligning the data, which is a problem for both pixel-based and peak table- based methods. Concurrently, the tile-based Fisher-ratio software maximizes the sensitivity contrast of true positives against a background of potential false positives and noise. To study this software, eight compounds, plus one internal standard, were spiked into diesel at various concentrations. The tile-based F-ratio software was able to discover all spiked analytes, within the complex diesel sample matrix with thousands of potential false positives, in each possible concentration comparison, even at the lowest absolute spiked analyte concentration ratio of 1.06.

chemometrics↗

THE ROADMAP FOR UNIFICATION IN GALAXY GROUP SELECTION. I. A SEARCH FOR EXTENDED X-RAY EMISSION IN THE CNOC2 SURVEY

X-ray properties of galaxy groups can unlock some of the most challenging research topics in modern extragalactic astronomy: the growth of structure and its influence on galaxy formation. Only with the advent of the Chandra and XMM-Newton facilities have X-ray observations reached the depths required to address these questions in a satisfactory manner. Here we present an X-ray imaging study of two patches from the CNOC2 spectroscopic galaxy survey using combined Chandra and XMM-Newton data. A state of the art extended source finding algorithm has been applied, and the resultant source catalog, including redshifts from a spectroscopic follow-up program, is presented. The total number of spectroscopically identified groups is 25 spanning a redshift range 0.04-0.79. Approximately 50% of CNOC2 spectroscopically selected groups in the deeper X-ray (RA14h) field are likely X-ray detections, compared to 20% in the shallower (RA21h) field. Statistical modeling shows that this is consistent with expectations, assuming an expected evolution of the L{sub X} -M relation. A significant detection of a stacked shear signal for both spectroscopic and X-ray groups indicates that both samples contain real groups of about the expected mass. We conclude that the current area and depth of X-ray and spectroscopic facilities provide a unique window of opportunity at z approx 0.4 to test the X-ray appearance of galaxy groups selected in various ways. There is at present no evidence that the correlation between X-ray luminosity and velocity dispersion evolves significantly with redshift, which implies that catalogs based on either method can be fairly compared and modeled.

79 ASTRONOMY AND ASTROPHYSICS↗

Electron postmastectomy chest wall plus comprehensive nodal irradiation technique using Electron Monte Carlo dose algorithm

For left-sided postmastectomy patients requiring chest wall plus comprehensive nodal irradiation, sometimes traditional techniques such as partial wide tangents or electron/tangents combination are not feasible due to abnormal chest wall contour or heart position or unusually wide excision scar. We developed electron chest wall irradiation technique using Electron Monte Carlo (EMC) dose algorithm that will achieve heart sparing with acceptable ipsilateral lung dose, minimal contralateral lung, and breast dose. Ten left-sided postmastectomy patients with very challenging anatomy were selected for this dosimetry study. The en face electron fields were designed from a single isocenter and gantry angle with different energy beams using different cutouts that matched on the skin. Smaller energy was used in the central thin chest wall part and higher energy in the medial internal mammary nodes (IMN) area, superior part of the thick chest wall, and/or axillary nodal area. The electron fields were matched to the photon supraclavicular field in the superior region. Daily field junctions were used to feather the match lines between all the fields. Electron field dose calculations were done with Monte Carlo. Five patients' chest wall fields were planned with 6/9MeVcombination, 1 with 6/12 MeV, 2 with 9/12 MeV, 1 with 9/16 MeV, and 1 with 6/9/12 MeV. Institutional criteria of prescription dose of 50 Gy for target volumes and normal tissue dose were met with this technique in spite of the challenging anatomy. Mean heart dose averaged 3.0 Gy ± 0.8 Gy. For ipsilateral lung, V20Gy and V5Gy averaged 33.2% ± 4.5% and 64.6% ± 9.6%, respectively. For contralateral lung, V5Gy averaged 5.1% ± 5.0%. For contralateral breast, V5Gy averaged 3.3% ± 3.1%. The electron chest wall irradiation technique using EMC dose algorithm can provide adequate dose coverage to the chest wall, IMNs, and/or axillary nodes while achieving heart sparing with acceptable ipsilateral lung dose, minimal contralateral lung, and breast dose.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Distributed Control Strategy to Achieve Synchronized Operation of an Islanded MG

To seamlessly transition a microgrid (MG) from islanded to grid-connected mode, it is necessary to synchronize the magnitude, frequency, and phase of the MG voltage to the voltage of the main grid. In this paper, we present a distributed control strategy to achieve synchronized operation of an islanded MG supported by multiple controllable distributed generators (DGs). The proposed method utilizes pinning-based consensus algorithm to ensure explicit coordination between magnitude, frequency and phase angle regulation, while ensuring proportional power sharing. System frequency is regulated by all the DGs in proportion to their capacity, while a selected DG eliminates the phase and magnitude regulation errors. Controller design criteria is based on small-signal stability analysis. Here, the proposed control strategy is implemented in hardware controllers and its effectiveness is demonstrated using a real-time hardware-in-the-loop MG testbed.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

An Efficient Sweep-Based Solver for the S N Equations on High-Order Meshes

In this study, we propose a graph-based sweep algorithm for solving the steady-state, monoenergetic discrete ordinates on meshes of high-order (HO) curved mesh elements. Our spatial discretization consists of arbitrarily HO discontinuous Galerkin finite elements using upwinding at mesh element faces. To determine mesh element sweep ordering, we define a directed, weighted graph whose vertices correspond to mesh elements and whose edges correspond to mesh element upwind dependencies. This graph is made acyclic by removing select edges in a way that approximately minimizes the sum of removed edge weights. Once the set of removed edges is determined, transport sweeps are performed by lagging the upwind dependency associated with the removed edges. The proposed algorithm is tested on several two-dimensional and three-dimensional meshes composed of HO curved mesh elements.

97 MATHEMATICS AND COMPUTING↗

Benchmark Imagery Project: Selection of Real World Images for Creating Composite Images of Facilities

The "Benchmark Imagery" project is a DOE NA-22-funded effort whose three-year goal is to create a suite of imagery (both real and composite) that can be used to test geospatial algorithms used for extracting semantic content from overhead images of industrial facilities. A composite image is an image that contains both real elements such as photographic imagery and synthetic elements such as 3D models. Composite images have the advantage of being customizable and distributable, and they can contain complete ground truth. It is our goal to generate composite images which appear realistic to both human and algorithm. When generating composite images of facilities we use real overhead imagery to increase realism and decrease effort. To construct all elements with purely synthetic textures would require too much effort and still may not look as realistic as a real image. We use real time 3D rendering software "SceneWorks" for rendering composite images. In SceneWorks, a scene is constructed by placing 3D models of facility components on top of a base terrain. A base terrain is a 3D model constructed from digital elevation mapping (DEM) data and then textured with overhead orthorectified imagery. The base terrain is constructed to be geospatially accurate using a real location’s source data (DEM and Imagery.) This document outlines the selection criteria in selecting appropriate location and source data for constructing the base terrain. Imagery should have resolution equal or better than desired composite image resolution. The imagery should have no seams. Seams can occur when two or more different imagery data sets exist in a single area with neither set covering the entire area. Elevation data (DEM) resolution generally should be 1/9 arc second or better depending on the variability of elevation. The imagery must contain a flat clearing of land large enough for multiple facility types. There should be no high-rise buildings anywhere in the base terrain.

42 ENGINEERING↗

Neutrino event selection in the MicroBooNE liquid argon time projection chamber using Wire-Cell 3D imaging, clustering, and charge-light matching

An accurate and efficient event reconstruction is required to realize the full scientific capability of liquid argon time projection chambers (LArTPCs). The current and future neutrino experiments that rely on massive LArTPCs create a need for new ideas and reconstruction approaches. Wire-Cell, proposed in recent years, is a novel tomographic event reconstruction method for LArTPCs. The Wire-Cell 3D imaging approach capitalizes on charge, sparsity, time, and geometry information to reconstruct a topology-agnostic 3D image of the ionization electrons prior to pattern recognition. A second novel method, the many-to-many charge-light matching, then pairs the TPC charge activity to the detected scintillation light signal, thus enabling a powerful rejection of cosmic-ray muons in the MicroBooNE detector. A robust processing of the scintillation light signal and an appropriate clustering of the reconstructed 3D image are fundamental to this technique. In this paper, we describe the principles and algorithms of these techniques and their successful application in the MicroBooNE experiment. A quantitative evaluation of the performance of these techniques is presented. Using these techniques, a 95% efficient pre-selection of neutrino charged-current events is achieved with a 30-fold reduction of non-beam-coincident cosmic-ray muons, and about 80% of the selected neutrino charged-current events are reconstructed with at least 70% completeness and 80% purity.

3D imaging↗

SU-F-J-200: An Improved Method for Event Selection in Compton Camera Imaging for Particle Therapy

Purpose: The uncertainty in the beam range in particle therapy limits the conformality of the dose distributions. Compton scatter cameras (CC), which measure the prompt gamma rays produced by nuclear interactions in the patient tissue, can reduce this uncertainty by producing 3D images confirming the particle beam range and dose delivery. However, the high intensity and short time windows of the particle beams limit the number of gammas detected. We attempt to address this problem by developing a method for filtering gamma ray scattering events from the background by applying the known gamma ray spectrum. Methods: We used a 4 stage Compton camera to record in list mode the energy deposition and scatter positions of gammas from a Co-60 source. Each CC stage contained a 4×4 array of CdZnTe crystal. To produce images, we used a back-projection algorithm and four filtering Methods: basic, energy windowing, delta energy (ΔE), or delta scattering angle (Δθ). Basic filtering requires events to be physically consistent. Energy windowing requires event energy to fall within a defined range. ΔE filtering selects events with the minimum difference between the measured and a known gamma energy (1.17 and 1.33 MeV for Co-60). Δθ filtering selects events with the minimum difference between the measured scattering angle and the angle corresponding to a known gamma energy. Results: Energy window filtering reduced the FWHM from 197.8 mm for basic filtering to 78.3 mm. ΔE and Δθ filtering achieved the best results, FWHMs of 64.3 and 55.6 mm, respectively. In general, Δθ filtering selected events with scattering angles < 40°, while ΔE filtering selected events with angles > 60°. Conclusion: Filtering CC events improved the quality and resolution of the corresponding images. ΔE and Δθ filtering produced similar results but each favored different events.

60 APPLIED LIFE SCIENCES↗

Adaptation of a cubic smoothing spline algortihm for multi-channel data stitching at the National Ignition Facility

Some diagnostics at the National Ignition Facility (NIF), including the Gamma Reaction History (GRH) diagnostic, require multiple channels of data to achieve the required dynamic range. These channels need to be stitched together into a single time series, and they may have non-uniform and redundant time samples. We chose to apply the popular cubic smoothing spline technique to our stitching problem because we needed a general non-parametric method. We adapted one of the algorithms in the literature, by Hutchinson and deHoog, to our needs. The modified algorithm and the resulting code perform a cubic smoothing spline fit to multiple data channels with redundant time samples and missing data points. The data channels can have different, time-varying, zero-mean white noise characteristics. The method we employ automatically determines an optimal smoothing level by minimizing the Generalized Cross Validation (GCV) score. In order to automatically validate the smoothing level selection, the Weighted Sum-Squared Residual (WSSR) and zero-mean tests are performed on the residuals. Further, confidence intervals, both analytical and Monte Carlo, are also calculated. In this paper, we describe the derivation of our cubic smoothing spline algorithm. We outline the algorithm and test it with simulated and experimental data.

42 ENGINEERING↗

Quantifying Climate and Catchment Control on Hydrological Drought in the Continental United States

The evolution of hydrological drought events is a result of complex (nonlinear) interactions between climate and catchment processes. To investigate such nonlinear relationship, we integrated a machine learning modeling framework based on the random forest (RF) algorithms with an interpretation framework to quantify the role of climate and catchment controls on hydrological drought. More particularly, our framework interprets a built RF machine-learning model to identify dominant variables and visualize their functional dependence and interaction effects on hydrological drought characteristics utilizing concepts of minimal depth, interactive depth, and partial dependence. We test our proposed modeling framework based on a set of 652 continental United States catchments with minimal human interference for a period of 1979–2010. Application of this framework indicated presence of three distinct drought regimes, which includes, Regime 1: droughts with longer duration, less frequent and lesser intensity; Regime 2: droughts with moderate duration, moderate frequency, and moderate intensity; and Regime 3: droughts with shorter duration, more frequent, and more intense. RF algorithm was able to accurately model the drought characteristics (intensity, duration, and number of events) for all the three drought regimes as a function of selected variables. It was observed that the type of dominant variables as well as their nonlinear functional relationship with hydrological droughts characteristics can vary between three selected regimes. Finally, our interpretation framework indicated that catchment characteristics have a significant role in controlling the hydrologic drought for catchments (regime 1), whereas both climate and catchment characteristics control hydrological drought in regimes 2 and 3.

54 ENVIRONMENTAL SCIENCES↗