Back

Reconstructing Ambient Temperature in Forensic Death Time Estimation

Shanmugam, J. S.; Hubig, M.; Schenkl, S.; Muggenthaler, H.; Springer, S.; Weiser, M.; Sudau, J.; Shah, F.; Mall, G.

2025-11-14 biophysics
10.1101/2025.11.14.688394 bioRxiv
Show abstract

Ambient temperature TA has a strong impact on temperature based time since death estimation (TTDE). Frequently TA is lowered instantaneously at a time t0 from a start value TA0 to TA1 < TA0 e.g. by window or door opening. We focus on reconstructing TA0 and t0. TTDE literature suggests temperature measurements in closed compartments as e.g. cupboards or neighboring rooms, where TA0 could have been preserved after t0. We aim to estimate t0 and TA0 from temperature measurements TZ(t) in closed compartments Z at times t > t0. We obtained promising results assuming Newtonian cooling for boxes filled with air, with heaps of clothes or even with books in two different experimental scenarios. Two different parameter estimators, (t0^, TA0^) forsingle quadruple temperature measurement and (t0*, TA0*) for N quadruple measurements were tested. Our results in a climate chamber were partially appropriate for TTDE input. A decline at time t0 from TA0 = 22.5{degrees}C {downarrow} TA1 = 14{degrees}C was reconstructed at t = t0 + 95min with relative deviations {rho}t0^ = 27% and {rho}TA0^ = 19% relative to t - t0 and TA0 - TA1 respectively, for N = 1 quadruple measurement with span {Delta}t = 50min. For N = 200 quadruple measurements in [t = t0 + 95min, t = t0 + 295min] we found {rho}t0^ = 5% and {rho}TA0^ = 11% with the same {Delta}t. Further research is necessary to guarantee applicability in routine casework. We will investigate more elaborate cooling models, estimation algorithms, and evaluation localization.

Matching journals

The top 11 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.