Because lead, nickel, and cadmium were the factory's principal raw materials, they were the contaminants most expected in the soil. The site's twentieth-century production of submarine batteries also made it a target of repeated wartime bombing, which may have added oil, heavy metals, or nitroaromatic explosives to the mix of pollutants. Geologically, the site is challenging.

Beneath a surface sealing up to 1.1 meters thick lies a heterogeneous artificial backfill reaching 4.1 meters deep, composed of sand, rubble, slag, broken bricks, and occasional wood, plastic, or metal. The soil is coarse and poorly sorted, with grain sizes ranging mostly from medium sand to fine gravel. Previous explorations suggested the heavy metal contamination does not extend beyond this backfill into the underlying natural geology, which consists of Quaternary river deposits of sandy or silty gravel up to 3.5 meters thick, partially overlain by flood sediments, all resting on Upper Carboniferous claystone bedrock.

Samples were taken at least once per drilled meter, or whenever the soil changed noticeably. Eight of the probing points were measured directly on site with a Bruker S1 Titan 800 pXRF instrument, yielding 75 measurements taken immediately on the cut surface of the drill cores without any pretreatment. The remaining samples were air-dried for at least 48 hours, broken up, and sieved to 2 millimeters before being measured under laboratory conditions through low-density polyethylene sample bags.

One of the study's key technical contributions concerns those plastic bags. Measuring through any plastic layer can attenuate, amplify, or scatter the fluorescence signal, and previous studies had given conflicting advice on whether to avoid barriers altogether. The team compared 0.05-millimeter-thick LDPE bags against standard 0.0015-millimeter XRF thin film using certified reference materials, measuring each ten times with and without the bag.

The lightest elements suffered dramatically: between 50 and 90% of the signal for magnesium oxide, aluminum oxide, silicon dioxide, phosphorus, sulfur, chlorine, and potassium oxide was lost through the bag. Elements heavier than potassium deviated by less than 30%, and some, including manganese, strontium, yttrium, and barium, were actually amplified. From these differences the researchers calculated empirical correction factors for the heavy metals of interest, making bag measurements a practical option for rapid screening.

The on-site measurements, taken on unprepared drill cores, proved useful for immediate field decisions and for selecting samples for further analysis, but their reliability was limited by matrix effects, grain size, and moisture content. Water is a known enemy of X-ray fluorescence: previous research recorded a 37% signal loss for arsenic in samples containing 20% moisture by weight, and the effect is stronger for elements with lower-energy fluorescence lines. At several depths the water-saturated zone prevented on-site measurement altogether.

The team noted that a minimal preparation protocol, involving removal of debris, shaking, sieving to 2 millimeters, and air-drying, would reduce the technique's speed advantage but substantially increase the significance of each reading. The heart of the study lies in the comparison between laboratory pXRF and conventional analysis. A subset of 113 samples underwent microwave-assisted aqua regia digestion followed by inductively coupled plasma optical emission spectroscopy (ICP-OES), the standardized wet-chemistry route for heavy metals in soil.

Mercury was quantified separately by cold vapor atomic absorption spectrometry, which offers a detection limit as low as 13.2 nanograms per gram. When the two methods were judged against German regulatory limits for the soil-to-human exposure pathway in industrial and commercial areas, they agreed in 548 consensus negatives and 46 consensus positives out of 633 comparisons, against 12 false negatives and 27 false positives, a concordance of nearly 94%. Yet the agreement was not uniform across elements, and the systematic biases the team uncovered matter for remediation practice.

Copper, nickel, and zinc results clustered close to the ideal one-to-one line, with pXRF marginally overestimating nickel. Arsenic and chromium were systematically overestimated by pXRF: ICP-OES recorded arsenic above 100 milligrams per kilogram only three times, while pXRF recorded twenty such values. Conversely, lead, the site's most abundant pollutant, was systematically underestimated, with pXRF logging only six readings above 15,000 milligrams per kilogram against fourteen for ICP-OES.

The highest single-element concentration measured by pXRF was roughly 6,000 milligrams per kilogram of lead in a mixed sample, while ICP-OES recorded 20,000 milligrams per kilogram at one depth interval at sampling point 34, a location where the soil contained almost pure slag between 2.1 and 3.6 meters and where inconsistencies among all three methods pointed to extreme heterogeneity. The authors caution that incomplete aqua regia digestion of arsenic- and chromium-rich components may partly explain the overestimation, since this partial extraction does not fully dissolve metal silicates. Beyond classification accuracy, the density of pXRF data proved transformative.

While only about half the samples could be analyzed conventionally, every sample was measured by pXRF, and each measurement captured 43 elements rather than the eight heavy metals routinely tested by ICP-OES. This broader geochemical lens revealed antimony contamination in certain project areas that would likely have gone undetected, as antimony is not part of the standard remediation analysis suite. The dense grid also fed a three-dimensional interpolation of lead concentrations built with ordinary kriging in Earth Volumetric Studio, using 227 pXRF measurements from 59 drill cores and 224 ICP-OES measurements from 129 cores including historical data from 2007 and 2017.

The two models showed comparable large-scale pollution patterns, though the pXRF-based model displayed the expected lower lead values and some small low-concentration zones absent from the laboratory model. The authors are careful not to overstate the case: pXRF is not a stand-alone substitute for laboratory analysis, and its high detection limits for cadmium and chromium sit close to regulatory thresholds. Sampling error remains the dominant source of uncertainty in contaminated-site work.

But their conclusion is that the advantages outweigh the limitations when the technology is deployed as part of a hybrid strategy, with pXRF supplying the bulk of low-cost, high-density data and laboratory methods providing confirmatory checks on a selected subset. Future work, they suggest, should quantify moisture effects more precisely, test simple field preparation methods, and explore applications such as measuring heavy metals in surrounding vegetation to assess ecotoxicological impacts.

Subject of Research: Portable X-ray fluorescence assessment of heavy metal contamination in soils at a former battery factory site in Hagen, Germany

Article Title: Portable X-ray fluorescence (pXRF) analysis of heavy metals in contaminated soils: a case study from Varta Insel, Hagen

Article References: Kaulingfrecks, J., De Vleeschouwer, D., Michalski, A., & Achten, C. (2026). Portable X-ray fluorescence (pXRF) analysis of heavy metals in contaminated soils: a case study from Varta Insel, Hagen. Environmental Geochemistry and Health, 48(16), Article 667. https://doi.org/10.1007/s10653-026-03538-5

Image Credits: AI Generated

DOI: 10.1007/s10653-026-03538-5

Keywords: pXRF, heavy metals, soil contamination, remediation, ICP-OES, lead, arsenic, chromium, Varta Insel, Hagen, X-ray fluorescence, contaminated sites