Designing accurate emulators for scientific processes using calibration-driven deep models.

Predictive models that accurately emulate complex scientific processes can achieve speed-ups over numerical simulators or experiments and at the same time provide surrogates for improving the subsequent analysis. Consequently, there is a recent surge in utilizing modern machine learning methods to build data-driven emulators. In this work, we study an often overlooked, yet important, problem of choosing loss functions while designing such emulators. Popular choices such as the mean squared error or the mean absolute error are based on a symmetric noise assumption and can be unsuitable for heterogeneous data or asymmetric noise distributions. We propose Learn-by-Calibrating, a novel deep learning approach based on interval calibration for designing emulators that can effectively recover the inherent noise structure without any explicit priors. Using a large suite of use-cases, we demonstrate the efficacy of our approach in providing high-quality emulators, when compared to widely-adopted loss function choices, even in small-data regimes.

Improved surrogates in inertial confinement fusion with manifold and cycle consistencies.

Neural networks have become the method of choice in surrogate modeling because of their ability to characterize arbitrary, high-dimensional functions in a data-driven fashion. This paper advocates for the training of surrogates that are 1) consistent with the physical manifold, resulting in physically meaningful predictions, and 2) cyclically consistent with a jointly trained inverse model; i.e., backmapping predictions through the inverse results in the original input parameters. We find that these two consistencies lead to surrogates that are superior in terms of predictive performance, are more resilient to sampling artifacts, and tend to be more data efficient. Using inertial confinement fusion (ICF) as a test-bed problem, we model a one-dimensional semianalytic numerical simulator and demonstrate the effectiveness of our approach.

Scalable Topological Data Analysis and Visualization for Evaluating Data-Driven Models in Scientific Applications.

With the rapid adoption of machine learning techniques for large-scale applications in science and engineering comes the convergence of two grand challenges in visualization. First, the utilization of black box models (e.g., deep neural networks) calls for advanced techniques in exploring and interpreting model behaviors. Second, the rapid growth in computing has produced enormous datasets that require techniques that can handle millions or more samples. Although some solutions to these interpretability challenges have been proposed, they typically do not scale beyond thousands of samples, nor do they provide the high-level intuition scientists are looking for. Here, we present the first scalable solution to explore and analyze high-dimensional functions often encountered in the scientific data analysis pipeline. By combining a new streaming neighborhood graph construction, the corresponding topology computation, and a novel data aggregation scheme, namely topology aware datacubes, we enable interactive exploration of both the topological and the geometric aspect of high-dimensional data. Following two use cases from high-energy-density (HED) physics and computational biology, we demonstrate how these capabilities have led to crucial new insights in both applications.

Polybrominated diphenyl ethers in bald (Haliaeetus leucocephalus) and golden (Aquila chrysaetos) eagles from Washington and Idaho, USA.

Little is known about the exposure and accumulation of polybrominated diphenyl ethers (PBDEs) in the 2 species of eagles inhabiting North America. The authors analyzed the livers of 33 bald eagles and 7 golden eagles collected throughout Washington and Idaho, USA, for 51 PBDE congeners. Total PBDEs ranged from 2.4 ng/g to 9920 ng/g wet weight. Bald eagles and eagles associated with large urban areas had the highest PBDE concentrations; golden eagles and eagles from more sparsely populated areas had the lowest concentrations. Congener patterns in the present study (brominated diphenyl ether [BDE]-47, BDE-99, BDE-100, BDE-153, and BDE-154 dominating concentrations) were similar to those reported for other bird species, especially raptors. However, the authors also found elevated contributions of BDE-209 in golden eagles and BDE-77 in both species. Patterns in bald eagle samples reflected those in fillets of fish collected from the same general locations throughout Washington, suggesting local exposure to runoff-based contamination, whereas patterns in golden eagle samples suggest a difference in food chain uptake facilitated by atmospheric transport and deposition of BDE-209 and its degradation products into the terrestrial system. Data from the present study represent some of the first reported on burdens of PBDEs in juvenile and adult eagles from North America. The high PBDE liver concentrations associated with large metropolitan areas and accumulation of deca-BDE congeners are a cause for concern.

Lead exposure and poisoning of songbirds using the Coeur d'Alene River Basin, Idaho, USA.

Previous studies have found widespread Pb poisoning of waterfowl in the Coeur d'Alene River Basin in northern Idaho, USA, which has been contaminated by mining and smelting activities. We studied the exposure of ground-feeding songbirds to Pb, sampling 204 American robins (Turdus migratorius), song sparrows (Melospiza melodia), and Swainson's thrushes (Catharus ustulatus) throughout the basin. These songbirds had mean blood Pb concentrations (mg/kg, dry weight) of less than 0.19 at a reference area (25 mg Pb/kg soil), 1.09 at moderately contaminated sites (170 to 1300 mg Pb/kg soil), and 2.06 at highly contaminated sites (2000 to 5000 mg Pb/kg soil). Based on guidelines for evaluating blood Pb in birds, 6% of robins from the highly contaminated sites had background concentrations, 24% were subclinically poisoned, 52% were clinically poisoned, and 18% were severely clinically poisoned with Pb. Blood Pb concentrations were lower in song sparrows than in robins and lowest in Swainson's thrushes. More than half of the robins and song sparrows from all contaminated sites and more than half of the Swainson's thrushes from highly contaminated sites showed at least 50% inhibition of the activity of the enzyme δ-aminolevulinic acid dehydratase (ALAD), commonly used as a measure of exposure to Pb. The highest hepatic Pb concentration of 61 mg/kg (dry weight) was detected in a song sparrow. Using Al as a marker for soil in songbird ingesta, we estimated average soil ingestion rates as 20% in robins, 17% in song sparrows, and 0.7% in Swainson's thrushes. Soil Pb in ingesta accounted for almost all of the songbirds' exposure to Pb. Based on these results, it is recommended that ecological risk assessments of ground-feeding songbirds at contaminated sites include soil ingestion as a pathway of exposure to Pb.

Evaluation of polychlorinated biphenyl remediation at a superfund site using tree swallows (Tachycineta bicolor) as indicators.

We studied tree swallows (Tachycineta bicolor) at the Sangamo National Priorities List (NPL) site and a reference area (Little Grassy) within Crab Orchard National Wildlife Refuge, Illinois, USA, to evaluate the bioavailability of polychlorinated biphenyls (PCBs) at the NPL site following remediation. Tree swallow eggs and 12- to 17-d-old chicks were collected from April to June 2004 and 2005. Total egg PCB concentrations in eggs did not differ between years at Sangamo; mean concentrations at Sangamo in 2004 to 2005 (4,452 ng/g) were higher than those from Little Grassy in 2004 (351 ng/g; p = 0.002) and 2005 (705 ng/g; p = 0.007). A positive PCB dilution in chicks is equivalent to relatively large amounts of exposure and absorption. Total PCB daily dilution in chicks in 2004 and 2005 averaged 92.4 and -203 ng/g/d at Sangamo and -21.8 and -42.7 ng/g/d at Little Grassy. Dioxin toxic equivalents (TEQs) in eggs in 2004 and 2005 were 1,844 and 676 pg/g at Sangamo and 165 and 128 pg/g at Little Grassy. Mean TEQs in chicks in 2004 and 2005 were 117 and 23.8 pg/g body weight at Sangamo and 0.8 and 0.7 pg/g body weight at Little Grassy. Total PCB concentrations in eggs and chicks at Sangamo and Little Grassy appeared to be comparable to other PCB-contaminated and reference sites. Our data demonstrate that tree swallows using the Sangamo site continue to accumulate significant quantities of PCBs seven years after remedial actions. These findings prompted a reevaluation of remediation at the site and underscore the importance of postremediation monitoring of ecological receptors, particularly in cases involving persistent environmental contaminants.

Blood lead concentrations in waterfowl utilizing Lake Coeur d'Alene, Idaho.

The Coeur d'Alene River Basin, Lake Coeur d'Alene, and the Spokane River contain elevated heavy metal concentrations in sediment and water from historical mining and ore processing operations in the Coeur d'Alene Basin. Lead poisoning has been identified as the cause of death in hundreds of waterfowl utilizing wetlands in the floodplain of the Coeur d'Alene River, but little was known about hazards to waterfowl from heavy metal contamination in shallow bays and wetlands of Lake Coeur d'Alene. We examined lake sediment and blood lead concentrations in waterfowl utilizing Lake Coeur d'Alene, Idaho, to evaluate potential lead contamination of waterfowl utilizing the lake. We collected 56 palustrine and 102 lacustrine sediment samples and 61 mallard and 8 wood duck blood samples. Mean lead concentrations from palustrine and lacustrine sediment samples ranged from 14 to 3508 mg/kg dry weight (dw) and from 19 to 5009 mg/kg (dw), respectively. Lead concentrations in palustrine and lacustrine sediment from several Lake Coeur d'Alene bays were higher than those in lake reference areas and were higher than Bunker Hill Superfund Site target cleanup levels and suggested site-specific toxicity thresholds for swans. Mean blood lead from mallard and wood ducks sampled from Lake Coeur d'Alene bays were within lead toxicity ranges for waterfowl associated with clinical and severe clinical lead poisoning. We also collected 19 Canada goose and 3 mallard fecal samples to evaluate exposure through sediment ingestion. Waterfowl using Lake Coeur d'Alene appear to be exposed to lead by ingesting contaminated lake sediment. Our model predicts a sediment lead effects range of 147-944 mg/kg (dw) and mortality effects level of 1652 mg/kg (dw) for mallards utilizing Lake Coeur d'Alene. The locations of Harrison Slough, Powderhorn Bay, and Cottonwood Bay at the mouth of the Coeur d'Alene River and Blackwell Island and Cougar Bay near the Spokane River outflow of Lake Coeur d'Alene were the areas of greatest concern for waterfowl exposure to lead contaminated sediment.

A cylindrical, collapsible, and economical field cage for mosquito research.

A new cage was designed to study caged mosquitoes in ultra-low volume insecticide spray studies in the field. Insecticide droplets may enter the cage from any direction because all surfaces are mesh. Cages can be collapsed in seconds, and reassembly is quick and simple. A 26-cm-diam cage can be constructed for less than US$6.00.

Bioaccumulation and trophic transfer of polychlorinated biphenyls by aquatic and terrestrial insects to tree swallows (Tachycineta bicolor).

Insectivorous passerines often bioaccumulate polychlorinated biphenyls (PCBs) via trophic transfer processes. Tree swallows (Tachycineta bicolor) frequently are used for estimating PCB bioaccumulation, yet the focus on specific trophic links between contaminated sediment and bird has been limited. Bioaccumulation of PCBs from sediment to tree swallows was examined with focus on trophic pathways by simultaneously examining PCBs in emergent aquatic and terrestrial insects and gut contents of nestlings. Total PCB concentrations increased from sediment (123.65 +/- 15.93 microg/kg) to tree swallow nestlings (2,827.76 +/- 505.67 microg/kg), with emergent aquatic insects, terrestrial insects, and gut content samples having intermediate concentrations. Biota-sediment accumulation factors (BSAFs) varied among congeners for tree swallow nestlings and for male and female Chironomus spp. For nestlings, the highest BSAF was for the mono-ortho-substituted congener 118. Nestling biomagnification values were similar for gut contents and female Chironomus spp., suggesting this diet item may be the main contributor to the overall PCB transfer to nestlings. However, gut content samples were highly variable and, on a PCB congener pattern basis, may have been influenced by other taxa, such as terrestrial insects. Considering dietary plasticity of many insectivorous birds, the present study suggests that a variety of potential food items should be considered when examining PCB accumulation in insectivorous passerines.

A new route to chaos: sequences of topological torus bifurcations.

We consider a sequence of topological torus bifurcations (TTBs) in a nonlinear, quasiperiodic Mathieu equation. The sequence of TTBs and an ensuing transition to chaos are observed by computing the principal Lyapunov exponent over a range of the bifurcation parameter. We also consider the effect of the sequence on the power spectrum before and after the transition to chaos. We then describe the topology of the set of knotted tori that are present before the transition to chaos. Following the transition, solutions evolve on strange attractors that have the topology of fractal braids in Poincare sections. We examine the topology of fractal braids and the dynamics of solutions that evolve on them. We end with a brief discussion of the number of TTBs in the cascade that leads to chaos.