A furnished studio apartment of approximately 10 m2 located on the 2nd floor of a 4-storey apartment building in an urban residential area (Frankfurt/M., Germany) was used, comparable to crime scenes involving social housing, student dormitories, prostitute workplaces, small hotel rooms and the like. This apartment has a split window area consisting of a small tilt and turn window (in the following referred to as “window a”) and a large casement window (in the following referred to as “window b”). Four different experimental scenarios (termed a1, a2, b1, and b2) were tested. Tilt and turn windows, commonly found in some European countries, can be opened in two different ways: For scenario a1 the small tilt and turn window was only partially opened with the upper part of the window tilting into the room and the window base still attached to hinges (Fig. 1, b). For scenario a2, opening of the window much like a door, inwards into the room (Fig. 1, c). Based on this concept, the large, horizontally opening casement window was only partially opened in a simulated tilt function for scenario b1 and completely opened for scenario b2. Accordingly, the following opening surfaces were calculated: 0.18 m2 for scenario a1, 0.81 m2 for scenario a2, 1.35 m2 for scenario b1, and 1.51 m2 for scenario b2. In total, the number of experimental window openings amounted to 81 across all four scenarios with window opening periods of 5 (n = 4), 10 (n = 4), 15 (n = 8), 30 (n = 8), 60 (n = 20), 120 (n = 17), 240 (n = 19), and 360 (n = 1) minutes. During the window openings, the ambient temperature was measured using a total of 20 data loggers (iButton® DS1922L, Maxim Integrated Inc., San Jose, CA, USA [45]), attached to a metal rack in pairs of two data loggers per location, one at a height of 1.0 m and one at a height of 0.1 m (Fig. 1, a). The data logger pairs were placed in 10 different locations (A – J) within the apartment furnished with a bed, a desk mounted on the wall, 5 closets and storage racks, one refrigerator (which remained disabled throughout all measurements), and a lavatory (Fig. 2). During the whole measurement period, there were no people in the apartment and the lighting and electronic devices were switched off. The radiator was situated beneath the sill under the desk (Fig. 2, bottom right). The regulator (height above the ground floor approximately 0.7 m) was adjusted to a normal room temperature of approximately 21 °C to resemble the corridor temperature avoiding additional artefacts due to the door opening procedure. The sampling rate of all data loggers was set to one measurement per minute with a resolution of 11-bits (0.0625 °C). The measuring accuracy is specified by the manufacturer as ± 0.5 °C. All data loggers were set with a time delay for the start of temperature data collection after initial synchronization to ensure time-synchronicity in every logger-position. The temperature loggers where delivered in a calibrated state. A loss of calibration due to battery loss did not occur.

(a): Metal rack construction and data logger positioning at 0.1 m and 1.0 m height (black arrow 1.0 m, dotted black arrow 0.1 m). (b): Tilt-function exemplary for window a (experimental scenario a1). (c): Turn-function exemplary for window a (experimental scenario a2)

Top left: Ground plan indicating the locations of all indoor data logger pairs (A—J, squares mark logger pairs included in further statistical analysis). Top right: 3D side views of the apartment. Bottom left: Photograph showing logger pairs B, C, E, and G. Bottom right: Photograph showing logger pairs F, I, and J (loggers at a height of 0.1 m positioned under and loggers at a height of 1.0 m positioned above the desk), the radiator location and the south/southeast oriented wall containing windows a and b. Notice the weather station rack attached by vacuum-handles on window b (exemplary setup for testing scenarios a1 and a2)

Since studies comparing temperature measurements at the site of discovery with official data from nearby weather stations showed that the temperature from meteorological weather stations and the actual data can consistently differ significantly regardless of the TSD and the body discovery site [46,47,48,49], a dedicated weather station (Gill Multi-plate Radiation Shield [50, 51]) was chosen to yield precise location-based temperature information. It was self-constructed and mounted on a rack attached to the outside surface of the window not used in the respective experimental scenario (Fig. 2, bottom right). Measurements of the outside ambient temperature were executed at the level of the window (> 2 m above the outside ground level) with a distance between weather station and window plane of approximately 60 cm. To avoid malfunction problems, 2 data loggers (resulting in one pair) were placed inside the station. Furthermore, the data of temperature profiles for the complete experimentation phase of the three closest public weather stations (1420, 1424, and 7341 [52]) were acquired from the DWD (Deutscher Wetterdienst, Offenbach, Germany) and a mean temperature of the hourly temperature values was calculated and compared to our measured data. To achieve a high temperature difference and thus evoke maximum effects, experiments were carried out in the winter months of January and February.

Pretests documenting indoor and outdoor temperatures during the course of the day were carried out for 6 days, after three of which shutters were used to provide for a temperature course without relevant solar irradiation artefacts. The door could be opened towards the apartment and was reached through a heated indoor corridor (with an average temperature of approximately 21 °C).

The measurement data collected by the loggers were transferred to, as well as cropped and plotted with the software program Excel (Microsoft Office Professional Plus 2016, Microsoft Corp., Redmond, WA, USA) after completion of the experiments. Statistical analysis was carried out using the statistical software packages SPSS (SPSS Statistics 29, IBM Corp., Armonk, NY, USA) and MedCalc (MedCalc Software Ltd., Ostende, Belgium).

For in total 22 data loggers (10 indoor pairs, location A – J, and 1 outdoor pair) per 81 scenarios (a1, a2, b1, b2) resulting in n = 1,782 recorded temperature data sets (one data set consisting of a varying amount of data points depending on the window opening period). All recorded measurements were reviewed to identify and characterize outliers and common characteristics. Based on these observations and analyses, three optimal locations of data logger pairs for evaluation of temperature profiles of the room for each window were identified. Further analyses were based on data originating from data logger pairs of these three selected locations C, G, and J for scenario a1 and a2 and C, F, and I for scenario b1 and b2 (Fig. 2, top left, bottom left, bottom right).

To determine the expected minimum and maximum ambient temperature alterations and the resulting errors possibly modifying temperature-based TSD estimations, the mean temperature loss within the first 15 min after opening the window and until the lowest recorded temperature for all scenarios were calculated. Furthermore, the respective temperature losses were compared to the window opening surfaces.

The observation period of ambient temperature after closing the windows was assessed. The time span necessary for reversion to pre-alteration temperatures was analyzed for individual temperature data sets with follow-up periods long enough to reverse. Furthermore, we aimed at detecting possible overcompensation phases after reaching reversion levels due to the radiator thermostat response to alteration-induced cooling of the room.

To evaluate the possible influence of the measured temperature alterations on the temperature-based TSD estimation we calculated differences of mean TSD estimations for all tested scenarios. At first, assuming a relatively stable room temperature of 21 °C, the TSD of an unclothed fictional corpse (Henssge empiric correction factor of 1.0) with a weight of 70 kg (serving as the cooling weight) was calculated [53] with presumed rectal body temperature measurements between 36.2 °C and 22.2 °C in intervals of 1.0 °C (resembling a decrease of body temperature over time for typical detection times nearly reaching the initial ambient temperature of 21 °C, bearing in mind, that the permissible critical Q-value of 0.2, was calculated with 24.24 °C and thus the lowest values would not be used for TSD estimation, whereby such problematic cases are already addressed by an previously published proposed solution [10]) using the Henssge Nomogram method [49]. The calculations were then repeated, using the mean values of the measured minimum room temperatures for scenarios a1, a2, b1, and b2 instead of 21 °C and the results were compared.