Dating the death of skeletonized human remains has been one of the significant challenges in forensic anthropology, as bones are subjected to a wide range of environmental and individual factors that influence their decomposition and diagenesis. As a result, estimating time since death with accuracy is particularly difficult, even when the objective is simply to differentiate between recent remains and archaeological ones. Although several methods are currently available for PMI estimation, many of them lack precision, are technically complex, costly, and not widely accessible [20, 24]. In this context, the present study aimed to explore a complementary quantitative approach for PMI estimation that is both low-cost and simple to apply, yet effective.

This work represents a pilot study designed to evaluate the feasibility of quantitatively measuring luminol chemiluminescence in human bone to estimate the postmortem interval. As such, it involved a limited but as representative as possible sample, providing preliminary insights into the potential and constraints of this quantitative approach while establishing a foundation for future, larger-scale research.

The results obtained in this study demonstrate a clear decrease in luminol chemiluminescence intensity (Ipeak) with increasing time since death. This decline, most prominent in the first 30 years after death, reflects the progressive degradation of hemoglobin derived iron compounds in the clavicle sample. For this reason, as the PMI increases, the luminol reaction intensity diminishes, resulting in lower Ipeak values. Overall, Ipeak values decrease markedly in recent PMIs, whereas older samples exhibit much lower and more uniform intensities.

The initial attempt to model the entire dataset using a single exponential decay function (Fig. 3a) revealed significant structural limitations. Although the overall trend showed a decline in peak intensity with increasing PMI, the residual analysis (Fig. 3b) demonstrated clear heteroscedasticity: samples with PMIs below 30 years exhibited large, systematic deviations from the fitted curve, whereas samples with higher PMIs clustered tightly around very low values. This pattern indicated that the dataset did not follow a single continuous exponential trajectory but rather reflected two distinct phases of degradation.

To address these structural limitations, a segmented regression approach was adopted. This model allowed the dataset to be partitioned into two distinct segments, separated by a changepoint supported both by visual inspection (Fig. 3c) and by statistical analysis (Fig. 3d), at approximately 30 years postmortem. Below this threshold, Ipeak values followed a steep exponential decline, whereas above it the data were best described by a low-variation trend. This segmentation provides a more accurate representation of the underlying degradation process than a single continuous model and is consistent with both the residual structure and the exploratory analyses. In fact, the identification of a statistically supported changepoint at approximately 30 years postmortem, which represents a practical threshold in the behaviour of luminol chemiluminescence intensity, constitutes a central outcome of this study. This threshold reflects a shift from a highly variable early postmortem phase to a long-term phase characterized by consistently low and minimally variable Ipeak values. From a forensic perspective, this distinction is particularly relevant, as it aligns with the practical need to differentiate remains that may fall within a forensic timeframe from those more likely to be of archaeological origin. Importantly, this threshold emerges from the segmented modelling approach and is supported by the residuals structure, the changepoint analysis, and the cross-validation.

The exponential behavior observed in the first segment reflects the rapid and irregular degradation of hemoglobin-derived compounds during the early postmortem period. In contrast, the second segment represents a long-term phase characterized by markedly reduced variation and low Ipeak values (below 300 A.U.), suggesting that hemoglobin components have largely degraded at minimal levels. The linear function used to model this second segment was selected for statistical robustness, namely to avoid overfitting, given the small number of available samples (n = 6). However, this choice does not rule out the possibility that the second segment follows a non-linear trend; rather, it reflects a statistical decision based on the current dataset. Future studies, including a larger number of older samples, should further investigate the underlying pattern of this segment.

The performance of the segmented model is further supported by the cross-validation results (LOOCV), which indicate a low predictive error in the second segment (6.16%) and confirm that separating the early highly variable phase from the later low-variation phase improves the overall performance of the model. Together, these findings demonstrate that the relationship between luminol chemiluminescence intensity and PMI is best interpreted as a biphasic process and support its value as a presumptive screening tool for distinguishing skeletonized human remains of forensic interest from archeological ones.

Despite the overall consistency of the segmented model, variability remained particularly pronounced among the clavicles with the lowest PMIs. This variability was reflected not only in the residual structure but also in the outlier diagnostics, with three samples with low PMIs (samples 1, 2 and 5) exceeding Cook’s distance thresholds and therefore being excluded. Such heterogeneity in the results obtained from the samples with early PMIs is likely related to the influence of taphonomic factors, including intrinsic factors (such as the individual’s physical condition, age or bone structure) and extrinsic variables (such as burial environment, temperature, humidity and oxygen exposure), which can affect the initial rate and trajectory of hemoglobin degradation. These elements tend to exert a profound influence during the initial stages of decomposition, when molecular components such as hemoglobin are more abundant and susceptible to environmental conditions.

In a previous study conducted by our team [25], the intensity of the luminol chemiluminescence reaction was qualitatively evaluated in a human bone sample buried under different environmental conditions. It was concluded that results obtained through the application of the luminol technique for estimating time since death could be influenced by taphonomic factors such as pH and type of soil, temperature and humidity, reinforcing the importance of considering the context in which a body is found when applying this method. However, in the present study, only one of the three clavicles excluded as statistical outliers had a known and controlled postmortem history, having been collected during an autopsy conducted at the Portuguese National Institute of Legal Medicine and Forensic Sciences and subsequently kept under monitored conditions. In contrast, the remaining two, although also of recent origin, were sourced from dry skeletal remains from the 21 st Century Identified Skeletal Collection, and underwent decomposition under environmental conditions that are not fully documented or controlled (although the characteristics of the graves and the burial ritual are similar, as they originated in the same cemetery). For this reason, the exclusion of these outliers cannot be explained based on the decomposition conditions of this subsample.

In summary, the luminol chemiluminescence results may be influenced by multiple taphonomic variables, such as burial conditions, climate, and postmortem handling, which can affect the survival of hemoglobin in bone and, consequently, the reaction’s intensity. For instance, unusually dry and hot environments may preserve hemoglobin longer, resulting in unexpectedly high intensities even in older remains. Therefore, low luminol intensity generally suggests an older PMI but cannot be interpreted as definitive without considering environmental context. These nuances reinforce the role of luminol as a presumptive test rather than a standalone determinant of postmortem interval. In applied forensic contexts, a positive luminol reaction may justify further analysis, while negative or borderline results should prompt careful consideration of burial and postmortem factors. The integration of contextual information remains essential for accurate interpretation. However, according to our results, the influence of such variability appears to diminish over time, with older samples converging towards consistently low Ipeak values, below 300 A.U.

Despite the variability observed at lower PMIs, the overall trend identified in the present study remains consistent. Samples with PMI below 20 years consistently yielded higher Ipeak values, while those with PMI exceeding 30 years remained below 300 A.U. (Table 2). These findings are in line with our previous research using visual classification systems [26], which reported that “strong positive” luminol reactions were only seen in remains up to 20 years PMI, while any reaction in remains beyond 30 years is mainly “Barely visible” or “Negative”. Our quantitative results not only reinforce those patterns but lend them greater accuracy by replacing categorical visual scoring with continuous numerical measurements. Compared to our previous studies [11, 21, 26], in which luminol chemiluminescence intensity was evaluated using both a binary and a five-level visual scale, the quantitative approach adopted in the present work allows for a more objective and reproducible assessment of the luminol technique. While the prior studies demonstrated the potential of luminol as a tool for discriminating between forensic and non-forensic remains, the method was inherently limited by its reliance on visual evaluation, although it yielded excellent inter and intraobserver agreement. The present findings not only confirm the trends observed in those earlier works but also enable a more refined statistical treatment of the data.

Our findings also align with previous research that employed a quantitative approach to estimate PMI in skeletonized remains through the luminol technique. Sarabia et al. (2018), using a luminometer to quantify luminescence in femora with PMIs ranging from 15 to 64 years, observed a significant inverse correlation between chemiluminescent intensity and PMI, particularly evident in the initial 10–20 s of the reaction. Moreover, their classification model achieved 88.4% accuracy in identifying remains older than 20 years, corresponding to the forensic relevance threshold applied in Spain, where the study was conducted. In contrast, the marked decrease in chemiluminescence intensity observed in the present study occurred beyond the ~ 30-year threshold identified through segmented modelling. Overall, the referred study confirmed the possibility of distinguishing human skeletonized remains of forensic interest through this quantitative and objective instrumental technique, reinforcing this approach as a more precise complement to the qualitative analysis previously studied, which is supported in the present investigation.

Nonetheless, as mentioned before, the luminol chemiluminescence technique should be interpreted as a presumptive screening tool rather than a standalone method for estimating PMI. The observed false positives (older remains exhibiting residual chemiluminescence) and potential false negatives (observed in previous studies) demonstrate the limitations of relying solely on luminol intensity. Accordingly, integrating contextual information is essential for accurate interpretation. In forensic casework, this technique can be particularly useful for the initial triage of human skeletonized remains, since a strong luminol reaction may prompt further forensic investigation, while a weak or absent one may suggest archaeological interest. Yet, a combination of methods, such as other screening techniques or confirmatory analysis, like radiocarbon dating, should be employed to strengthen the estimation of time since death.

Despite its potential, there are some limitations to this study, and to the luminol technique in general, that warrant discussion. Firstly, the small sample size, which consists of 24 individuals (23 used for statistical purposes), represents an inherent limitation of this study. As previously mentioned, this study was conceived as a pilot investigation introducing a new quantitative approach to measure luminol reaction intensity. For this reason, a smaller sample was intentionally used to explore methodological feasibility before expanding to larger datasets. Future studies on this matter should involve a larger sample to strengthen the robustness of the results. Nevertheless, despite the reduced sample size, PMIs are well distributed across the dataset, albeit with a notable temporal gap between 53 and 500 years postmortem. Secondly, as previously referred, and as in real forensic cases scenarios, the absence of detailed taphonomic contexts introduces uncontrolled variability, as factors such as burial environment, temperature and humidity exposure, and some individual characteristics were largely unknown. Thirdly, the analysis was restricted to clavicles, which may not fully capture inter-bone variability in the technique’s responsiveness. Finally, as a presumptive test, luminol is sensitive to iron-containing compounds but not entirely specific to human hemoglobin, leaving room for potential interference by other substances. These constraints highlight the need for cautious interpretation and underscore the importance of complementary data in forensic applications.

Overall, these results reinforce the utility of the luminol chemiluminescence technique as a rapid, cost-effective, and minimally destructive screening method for dating death, requiring only a small quantity of bone powder and producing results within minutes. The identification of a robust low-variation second segment and a statistically validated changepoint at approximately 30 years represents a significant step towards establishing luminol chemiluminescence as a presumptive screening tool for broad PMI estimation, determining whether remains are likely to fall within a forensic timeframe or are of archaeological interest. While accurate PMI estimation remains challenging, the results provide a conceptual and quantitative framework for building more refined models. However, the influence of taphonomic factors should be considered when interpreting individual results. Future research should aim to further investigate the influence of intrinsic and extrinsic factors and expand the dataset to include additional bone types and broader PMIs, improving the accuracy and applicability of this technique. Comparative validation with other emerging PMI estimation techniques should also be considered in future studies. With further refinement, the luminol technique can evolve from a promising presumptive test toward a more reliable, standardized technique in forensic anthropology.