Search for: All records

As next-generation scientific instruments and simulations generate ever larger datasets, there is a growing need for high-performance computing (HPC) techniques that can provide timely and accurate analysis. With artificial intelligence (AI) and hardware breakthroughs at the forefront in recent years, interest in using this technology to perform decision-making tasks with continuously evolving real-world datasets has increased. Digital twinning is one method in which virtual replicas of real-world objects are modeled, updated, and interpreted to perform such tasks. However, the interface between AI techniques, digital twins (DT), and HPC technologies has yet to be thoroughly investigated despite the natural synergies between them. This paper explores the interface between digital twins, scientific computing, and machine learning (ML) by presenting a consistent definition for the digital twin, performing a systematic analysis of the literature to build a taxonomy of ML-enhanced digital twins, and discussing case studies from various scientific domains. We identify several promising future research directions, including hybrid assimilation frameworks and physics-informed techniques for improved accuracy. Through this comprehensive analysis, we aim to highlight both the current state-of-the-art and critical paths forward in this rapidly evolving field. 

Abstract. The Mississippi River is a critical waterway in the United States, and hydrologic variability along its course represents a perennial threat to trade, agriculture, industry, the economy, and communities. The Community Earth System Model version 1 (CESM1) complements observational records of river discharge by providing fully coupled output from a state-of-the-art earth system model that includes a river transport model. These simulations of past, historic, and projected river discharge have been widely used to assess the dynamics and causes of changes in the hydrology of the Mississippi River basin. Here, we compare observations and reanalysis datasets of key hydrologic variables to CESM1 output within the Mississippi River basin to evaluate model performance and bias. We show that the seasonality of simulated river discharge in CESM1 is shifted 2–3 months late relative to observations. This offset is attributed to seasonal biases in precipitation and runoff in the region. We also evaluate performance of several CMIP6 models over the Mississippi River basin, and show that runoff in other models — notably CESM2 — more closely simulates the seasonal trends in the reanalysis data. Our results have implications for model selection when assessing hydroclimate variability on the Mississippi River basin, and show that the seasonal timing of runoff can vary widely between models.  Our findings imply that continued improvements in the representation of land surface hydrology in earth system models may improve our ability to assess the causes and consequences of environmental change on terrestrial water resources and major river systems globally. 

Abstract. Annually laminated lake sediment can track paleoenvironmental change at high resolution where alternative archives are often not available. However,information about the chronology is often affected by indistinct and intermittent laminations. Traditional chronology building struggles with thesekinds of laminations, typically failing to adequately estimate uncertainty or discarding the information recorded in the laminations entirely,despite their potential to improve chronologies. We present an approach that overcomes the challenge of indistinct or intermediate laminations andother obstacles by using a quantitative lamination quality index combined with a multi-core, multi-observer Bayesian lamination sedimentation modelthat quantifies realistic under- and over-counting uncertainties while integrating information from radiometric measurements (210Pb,137Cs, and 14C) into the chronology. We demonstrate this approach on sediment of indistinct and intermittently laminatedsequences from alpine Columbine Lake, Colorado. The integrated model indicates 3137 (95 % highest probability density range: 2753–3375) varveyears with a cumulative posterior distribution of counting uncertainties of −13 % to +7 %, indicative of systematic observerunder-counting. Our novel approach provides a realistic constraint on sedimentation rates and quantifies uncertainty in the varve chronology byquantifying over- and under-counting uncertainties related to observer bias as well as the quality and variability of the sediment appearance. The approachpermits the construction of a chronology and sedimentation rates for sites with intermittent or indistinct laminations, which are likely moreprevalent than sequences with distinct laminations, especially when considering non-lacustrine sequences, and thus expands the possibilities ofreconstructing past environmental change with high resolution.