Meteorological data source comparison-a case study in geospatial modeling of potential environmental exposure to abandoned uranium mine sites in the Navajo Nation

Abstract

Meteorological (MET) data is a crucial input for environmental exposure models. While modeling exposure potential using geospatial technology is a common practice, existing studies infrequently evaluate the impact of input MET data on the level of uncertainty on output results. The objective of this study is to determine the effect of various MET data sources on the potential exposure susceptibility predictions. Three sources of wind data are compared: The North American Regional Reanalysis (NARR) database, meteorological aerodrome reports (METARs) from regional airports, and data from local MET weather stations. These data sources are used as inputs into a machine learning (ML) driven GIS Multi-Criteria Decision Analysis (GIS-MCDA) geospatial model to predict potential exposure to abandoned uranium mine sites in the Navajo Nation. Results indicate significant variations in results derived from different wind data sources. After validating the results from each source using the National Uranium Resource Evaluation (NURE) database in a geographically weighted regression (GWR), METARs data combined with the local MET weather station data showed the highest accuracy, with an average R² of 0.74. We conclude that local direct measurement-based data (METARs and MET data) produce a more accurate prediction than the other sources evaluated in the study. This study has the potential to inform future data collection methods, leading to more accurate predictions and better-informed policy decisions surrounding environmental exposure susceptibility and risk assessment.

Keywords: Abandoned uranium mines; GIS multi-criteria decision analysis; Meteorological; Navajo nation; Particulate matter; Radom forest; Spatial analysis and modeling.

Fig. 1

Overview map of the Navajo Nation

Fig. 2

Flowchart of overall GIS-MCDA modeling approach including weight determination, and validation process with illustrating the k-fold cross validation method

Fig. 3

MET station wind rose example for Mexican Hat, UT

Fig. 4

Illustration of data inputs, main processing steps, and output criteria layers (30 × 30 m resolution) employed in the present analysis. Notes.¹point; ²line; ^avector spatial data format; ^braster spatial data format, 30 × 30 m resolution; ^cinput meteorological data including local airport METARs, Uranium Mill Tailings Remedial Action (UMTRA) MET stations, and gridded reanalysis data from the North American Regional Reanalysis (NARR); ^dsum of inverse distance from each cell to all AUMs within 50 km, weighted by the surface area of each AUM site; ^esum of the difference in the angle to AUM source and prevailing wind direction on a scale of zero to one; ^fcalculated for groundwater arsenic and uranium concentrations (MCL) in more than 467 local wells; results were interpolated using inverse distance weighting interpolation; Abbreviations. AUM abandoned uranium mine, DEM digital elevation model, NDVI Normalized Difference Vegetation Index

Fig. 5

Wind direction data from different meterological data sources in the Navajo Nation

Fig. 6

Potential environmental exposure to AUM predicted by the GIS-MCDA model with different version of meteorological data: METARs (a); MET stations (b); NARR (c); METARs and MET stations (d); NARR and METARs (e); NARR and MET stations (f); NARR, METARs, and MET stations (g); Overview of the Navajo Nation—with the area represented in the other figures outlined in red (h)

Fig. 7

Frequency distribution of 500 random NURE samples from the modeling results derived from different meterological data sources (Note: the X axis represents the log (base2) uranium exposure potential value; the Y axis represent the frequency of each value; the red line represents the mean value for each result version

Fig. 8

GWR validation results for each model version X-axis: sample number (0–999) Y-axis: GWR global R² value

Fig. 9

Predicted exposure surfaces across fold runs. Note high consistency in predicted surfaces

Fig. 10

R² value from the spatial scale tests using independent subset samples based on 99 iterations