Previously the colourimetric decision model based on the L* and B* value proved to be highly accurate for freshly burned human bone sections that were heated in media air or adipose tissue, and even for samples that deviated in exposure duration from the exposure duration used to develop the model. Also, the model proved to be highly accurate for samples that deviated in size from the sections used to develop the model [4]. Now the model has also been tested on embalmed human and non-human bone samples, yielding a high accuracy. Overall, this shows that the colourimetric decision model is robust and broader applicable than originally intended.

The majority of the incorrect temperature estimations were overestimations of the temperature, five of seven incorrect estimates out of 448 samples, of which four from pig. The temperature range of the incorrect estimations was from 220 °C to 250 °C, with just an overestimation of a single cluster, while the exposure temperature of the two underestimations (of also a single cluster) was 650 °C.

The increase in inter- and intrafibrillar cross-links, from the formalin fixation of the collagen in the bone matrix, and the addition of formalin (containing organic components), did not have a measurable effect on the classification accuracy based on HI-changes in colour. This shows that the results from a previous study, on the accuracy and precision of temperature estimations based on embalmed human bone samples, are not influenced by the addition of formalin [11]. However, it remains unclear whether, or to what degree, formalin fixation has an effect on the thermal stability.

Bone mineral density (BMD) might influence the rate of combustion of the organic components, a higher BMD could negatively affect oxygen availability necessary for combusting the organic remnants especially after carbonization. Porcine bone has the highest BMD of the, for this study, used species, based on mid-diaphyseal long bone measurements (femur, tibia, humerus, radius, of which the cow tibia showed a questionable low value but even by using the less questionable distal or proximal value the average BMD for cow remained lower compared to pig), while cow bone has the highest mean volumeBMD (449 mg/cm3) followed by pig (373 mg/cm3) and human bone (178 mg/cm3) [19, 20]. The carbonization of pig bone actually preceded cow and human bone (resulting in overestimation); hence the higher BMD of pig cortical bone did not negatively affect carbonization. The combustion of the organic component of pig bone, at 650 °C, lacked behind for two samples, indicating that BMD may indeed effect combustion. The effect of BMD on HI-changes should be further investigated in the future. Besides BMD, the ratio inorganic versus organic, collagen integrity, and porosity can possibly affect carbonization and combustion of the organic component of bone, interspecies differences of these independent variables can lead to incorrect temperature estimations.

The number of incorrect classifications does not substantiate the exclusion of using the model. However, the chosen species for this study do not reflect all possible species that can be encountered in a fire context. Remains of domestic animals can be expected to be encountered amongst the debris after a fire, also a wider variety of species can be expected to be found during archaeological excavations. It is, therefore, necessary to expand this classification accuracy test, and until then refrain from drawing conclusions on samples of species of which the classification accuracy is unknown. Further, one cannot discriminate between fresh human, embalmed human and non-human bone (Sus scrofa dom. and Bos taurus) with the colourimetric model in cases in which the exposure temperature is known.

Recently a scale, containing a bone colour to exposure temperature chart, was suggested as a tool for temperature estimations [21], which, when used visually, does not deviate from the previously described subjective methods. As previously mentioned, the subjective approach did not uphold to the legal standards [11, 12]. Taking colourimetric measurements is more laborious and time consuming than visually assessing and comparing the colour, with for example reference charts, but the data acquired from colourimetry is objective. We advise to adopt the scale to the colourimetric decision model and incorporate reference colours to test the imaging capturing device (instead of visual comparison).