A5048694535- Climate variability and models
- Meteorological Phenomena and Simulations
- Cryospheric studies and observations
- Explainable Artificial Intelligence (XAI)
- Hydrological Forecasting Using AI
- Scientific Computing and Data Management
- Hydrology and Drought Analysis
- Oceanographic and Atmospheric Processes
- Geophysics and Gravity Measurements
- Energy Efficient Wireless Sensor Networks
- Hydrology and Watershed Management Studies
- Computational Physics and Python Applications
- Atmospheric and Environmental Gas Dynamics
- Groundwater and Isotope Geochemistry
- Reservoir Engineering and Simulation Methods
- Climate change and permafrost
- Groundwater flow and contamination studies
- Climate Change and Geoengineering
- Ionosphere and magnetosphere dynamics
- Seismology and Earthquake Studies
- Soil Geostatistics and Mapping
- Ocean Acidification Effects and Responses
- Topic Modeling
- Infrastructure Resilience and Vulnerability Analysis
- Climate change impacts on agriculture
University of Virginia
2023-2025
Colorado State University
2021-2023
Irvine University
2017-2022
University of California, Irvine
2017-2022
University of Patras
2015-2018
Despite the increasingly successful application of neural networks to many problems in geosciences, their complex and nonlinear structure makes interpretation predictions difficult, which limits model trust does not allow scientists gain physical insights about problem at hand. Many different methods have been introduced emerging field eXplainable Artificial Intelligence (XAI), aim attributing network s prediction specific features input domain. XAI are usually assessed by using benchmark...
Abstract Extreme events such as heat waves and cold spells, droughts, heavy rain, storms are particularly challenging to predict accurately due their rarity chaotic nature, because of model limitations. However, recent studies have shown that there might be systemic predictability is not being leveraged, whose exploitation could meet the need for reliable predictions aggregated extreme weather measures on timescales from weeks decades ahead. Recently, numerous been devoted use artificial...
Abstract Storms include a range of weather events resulting in heavy liquid and solid precipitation high winds. These critically impact crops natural resources and, turn, health, economy, infrastructure safety. The intensity frequency the physical mechanisms triggering storms will most likely increase under global warming due to changing flows water energy atmosphere. Addressing storm threats holistically requires nexus approach that links climate change, infrastructure, human prosperity...
Abstract In extreme excess modeling, one fits a generalized Pareto (GP) distribution to rainfall excesses above properly selected threshold u . The latter is generally determined using various approaches, such as nonparametric methods that are intended locate the changing point between and nonextreme regions of data, graphical where studies dependence GP‐related metrics on level , Goodness‐of‐Fit (GoF) that, for certain significance, lowest GP model applicable. Here we review representative...
Convolutional neural networks (CNNs) have recently attracted great attention in geoscience due to their ability capture non-linear system behavior and extract predictive spatiotemporal patterns. Given black-box nature however, the importance of prediction explainability, methods explainable artificial intelligence (XAI) are gaining popularity as a means explain CNN decision-making strategy. Here, we establish an intercomparison some most popular XAI investigate fidelity explaining decisions...
Climate-driven changes in precipitation amounts and their seasonal variability are expected many continental-scale regions during the remainder of 21st century. However, much less is known about future predictability precipitation, an important earth system property relevant for climate adaptation. Here, on basis CMIP6 models that capture present-day teleconnections between previous-season sea surface temperature (SST), we show change to alter SST-precipitation relationships thus our ability...
Abstract Distribution mapping has been identified as the most efficient approach to bias‐correct climate model rainfall, while reproducing its statistics at spatial and temporal resolutions suitable run hydrologic models. Yet implementation based on empirical distributions derived from control samples (referred nonparametric distribution mapping) makes method's performance sensitive sample length variations, presence of outliers, resolution results, may lead biases, especially in extreme...
Reliable prediction of seasonal precipitation in the southwestern US (SWUS) remains a challenge with significant implications for economy, water security and ecosystem management region. Winter SWUS has been linked to several climate modes, including El Niño-Southern Oscillation (ENSO), limited predictive ability. Here we report evidence that late-summer sea surface temperature geopotential height anomalies close New Zealand exhibit higher correlation winter than ENSO, enhancing potential...
Abstract Methods of explainable artificial intelligence (XAI) are used in geoscientific applications to gain insights into the decision-making strategy neural networks (NNs), highlighting which features input contribute most a NN prediction. Here, we discuss our “lesson learned” that task attributing prediction does not have single solution. Instead, attribution results depend greatly on considered baseline XAI method utilizes—a fact has been overlooked literature. The is reference point...
Abstract The Southern Ocean is a region of high surface nutrient content, reflecting an inefficient biological carbon pump. variability, predictability, and causes changes in these levels on interannual to decadal time scales remain unclear. We employ deep learning approach, specifically Temporal Convolution Attention Neural Network (TCANN), conduct multi‐year forecasting based oceanic physical drivers. TCANN successfully replicates testing data with prediction skill extending at least 4...
The Madden-Julian Oscillation (MJO) is the leading mode of intra-seasonal climate variability, having profound impacts on a wide range weather and phenomena. Here, we use wavelet-based spectral Principal Component Analysis (wsPCA) to evaluate skill 20 state-of-the-art CMIP6 models in capturing magnitude dynamics MJO. By construction, wsPCA has ability focus desired frequencies capture each propagative physical with one principal component (PC). We show that MJO contribution total variability...
Abstract To improve the level skill of climate models (CMs) in reproducing statistics daily rainfall at a basin level, two types statistical approaches have been suggested. One is correction CM outputs based on historical series precipitation. The other, usually referred to as downscaling, use stochastic conditionally simulate series, large‐scale atmospheric forcing from CMs. While promising, latter approach attracted reduced attention recent years, since developed downscaling schemes...
Abstract Due to its importance for water availability in the tropics and subtropics, efficient tracking of seasonal long‐term shifts intertropical convergence zone (ITCZ) is great value. Current approaches, which are based on changes annual mean single variables, ignore intra‐annual dynamics, while more sophisticated methods computationally intensive. Here we propose a new probabilistic framework track ITCZ, location maximum precipitation minimum outgoing longwave radiation overlapping...
Abstract Stratospheric aerosol injection (SAI) has been proposed as a possible response option to limit global warming and its societal consequences. However, the climate impacts of such intervention are unclear. Here, an explainable artificial intelligence (XAI) framework is introduced quantify how distinguishable SAI might be from pre‐deployment climate. A suite neural networks trained on Earth system model data learn distinguish between pre‐ post‐deployment periods across variety...
Understanding the physical drivers of seasonal hydroclimatic variability and improving predictive skill remains a challenge with important socioeconomic environmental implications for many regions around world. Physics-based deterministic models show limited ability to predict precipitation as lead time increases, due imperfect representation processes incomplete knowledge initial conditions. Similarly, statistical methods drawing upon established climate teleconnections have low prediction...
Spectral PCA (sPCA), in contrast to classical PCA, offers the advantage of identifying organized spatio-temporal patterns within specific frequency bands and extracting dynamical modes. However, unavoidable tradeoff between resolution robustness PCs leads high sensitivity noise overfitting, which limits interpretation sPCA results. We propose herein a simple non-parametric implementation using continuous analytic Morlet wavelet as robust estimator cross-spectral matrices with good...
Precipitation prediction at seasonal timescales is important for planning and management of water resources as well preparedness hazards such floods, droughts wildfires. Quantifying predictability quite challenging a consequence large number potential drivers, varying antecedent conditions, small sample size high-quality observations available timescales, that in turn, increases uncertainty the risk model overfitting. Here, we introduce generalized probabilistic framework to account these...
Abstract Many of our generation’s most pressing environmental science problems are wicked problems, which means they cannot be cleanly isolated and solved with a single “correct” answer. The NSF AI Institute for Research on Trustworthy in Weather, Climate, Coastal Oceanography (AI2ES) seeks to address such by developing synergistic approaches team scientists from three disciplines: (including atmospheric, ocean, other physical sciences), artificial intelligence (AI), social including risk...
The accelerated progress of artificial intelligence (AI) has popularized deep learning models across domains, yet their inherent opacity poses challenges, notably in critical fields like healthcare, medicine and the geosciences. Explainable AI (XAI) emerged to shed light on these "black box" models, helping decipher decision making process. Nevertheless, different XAI methods yield highly explanations. This inter-method variability increases uncertainty lowers trust networks' predictions. In...
Atmospheric AI modeling is increasingly reliant on complex machine learning (ML) techniques and high-dimensional gridded inputs to develop models that achieve high predictive skill. Complex deep architectures such as convolutional neural networks transformers are trained model highly non-linear atmospheric phenomena coastal fog [1], tornadoes [2], severe hail [3]. The input data typically in the form of spatial composed multiple channels satellite imagery, numerical weather prediction...
Every year, astronomers from around the world submit research proposals to Atacama Large Millimeter Array (ALMA), largest radio telescope array in world. The aim of current work is streamline proposal process for submitting projects ALMA by suggesting frequency ranges that may be relevant their based on text. We introduce a pipeline supervised and unsupervised machine learning models, each using various representations title abstract an incoming proposal. First, logistic regression filters...