Dewey Dunnington (Ph.D., P.Geo.) is a software engineer and geoscientist based in Winnipeg, Manitoba. As a software engineer he works on all things Apache Arrow at Voltron Data, Inc., including standards for geospatial data connectivity, a C implementation of Apache Arrow, and Arrow Database Connectivity (ADBC). As a geoscientist, he has worked in contaminated site remediation, taught Applied Geomorphology at Acadia University, and has authored more than a dozen articles on lake water and sediment geochemistry. Dewey is an Apache Arrow Project Management Committee member, an RStudio-certified tidyverse instructor, an NSERC Postgraduate Scholarship (Doctoral) recipient, and maintainer of dozens of R, Python, C, and C++ libraries at the intersection of geoscience, geospatial data, and enterprise data connectivity.
PhD Civil & Resource Engineering, 2020
Dalhousie University
MSc Geology, 2015
Acadia University
BScH Environmental Science, 2011
Acadia University
Using an organic carbon removal data set (n = 500), we compared a physically based semiempirical coagulation model (Langmuir sorption-removal) and three ML modeling methods using quantitative (model performance) and qualitative (model interpretability and accessibility) criteria to identify potential barriers to adoption in water treatment. We found that a gradient-boosted tree ensemble and an artificial neural network provided the most accurate predictions of organic carbon removal and that all models provided accurate predictions when test data were well-characterized by the training data and confirmed that the physically based model had the lowest prediction error when extrapolating. As assessed by the ability of model predictions to be reconciled with industry-specific knowledge, the physically based and linear models were the most interpretable. As assessed by the ability for utilities to implement models on an ad hoc basis, the physically based and multiple linear models were deemed to be the most accessible. Collectively, our study suggests that ML-based models offer the best predictive performance when adequate training data are available and that physically based models are best suited when extrapolation is necessary. Potential solutions for limited interpretability of ML-based models include variable importance and sensitivity analysis; a potential solution for limited accessibility of ML-based models is training of stakeholders in modeling techniques.
We evaluated anthropogenic Pb deposition along a west-east transect from the Adirondack Mountains, New York, USA (ADIR) region, the Vermont-New Hampshire-Maine, USA (VT-NH-ME) region, and Nova Scotia, Canada (NS) region using 47 210Pb-dated lake sediment records. We used focus-corrected Pb inventories to evaluate cumulative deposition and breakpoint analysis to evaluate possible differences in timings among regions. Peak Pb concentrations decreased from west to east (ADIR region: 52–378 mg kg−1, VT-NH-ME region: 54–253 mg kg−1, NS: 38–140 mg kg−1). Cumulative deposition of anthropogenic Pb also decreased from west to east (ADIR region: 791–1344 mg m−2, VT-NH-ME region: 209–1206 mg m−2, NS: 52–421 mg m−2). The initiation of anthropogenic Pb deposition occurred progressively later along the same transect (ADIR region: 1869–1900, VT-NH-ME region: 1874–1905, NS region: 1901–1930). Previous lead isotope studies suggest that eastern Canadian Pb deposition over the past ~150 years has originated from a mix of both Canadian and U.S. sources. The results of this study indicate that anthropogenic Pb from sources west of the ADIR region were deposited in lesser amounts from west to east and/or Pb sources reflect less population density from west to east. The timing of the initiation of anthropogenic Pb deposition in the NS region suggests that Pb from gasoline may be an important source in this region.
We assessed factory-calibrated field-portable X-ray fluorescence (pXRF) data quality for use with minimally-prepared aquatic sediments, including the precision of replicate pXRF measurements, accuracy of factory-calibrated pXRF values as compared to total digestion/ICP-OES concentrations, and comparability of calibrated pXRF values to extractable concentrations. Data quality levels for precision, accuracy, and comparability were not equivalent for element/analyzer combinations. All analyses of elements that were assessed for precision and accuracy on a single analyzer were both precise (<10% relative standard deviation) and accurate (r2>0.85) for K, Ca, Ti, Mn, Fe, and Zn. Calibrated pXRF values for Al, K, Ca, Ti, Mn, Fe, Cu, Zn, and Pb were within ∼10% relative difference of total digestion/ICP-OES concentrations. Calibrated pXRF values for Fe, Cu, Zn, As, and Pb were within ∼20% relative difference of extractable concentrations. Some elements had a higher level of data quality using specific analyzers, but in general, no pXRF analyzer had the highest level of data quality in all categories. Collectively, our data indicate that a wide range of factory-calibrated pXRF units are capable of providing high-quality total concentrations for the analysis of aquatic sediments.