Application of Criminology in Urban Ecology and Evolution: Application of Criminology in Urban Ecology and Evolution: Routine Activity Theory and Field Equipment Disappearance Routine Activity Theory and Field Equipment Disappearance Dynamics Dynamics

Research in


Introduction
Urban areas currently occupy about 3% of the Earth's land and are home to more than half of the human population; both numbers are expected to increase over time (Seto et al., 2012;United Nations, 2019).
In line with this growth, urban ecology and evolution emerged in the past decades as a blooming field of research worldwide (James and Douglas, 2014;Rivkin et al., 2019;Szulkin et al., 2020a).Knowledge gained from these studies can ultimately allow us to build more sustainable cities, and lead toward a more harmonious coexistence of humans and other biological life (Douglas et al., 2011).As such, both the scientific community and land managers working in the urban space need carefully planned data collection protocols to work efficiently.
Field studies are the cornerstone staple for the collection of biological data.To achieve reliable and repeatable results, scientists and other stakeholders must apply efficient data collecting protocols that rely on the use and function of field equipment.Deployable equipment is used to gain information on biological processes across a wide-range of habitat types, from wild, green areas to urbanised neighbourhoods, and often has to be left outdoors.(e.g.Munshi-South and Kharchenko, 2010, Sprau et al., 2017, Corsini et al., 2022a, 2022b).Such equipment differs in terms of costs, quality and scope (Perkin et al., 2014, Zárybnická et al., 2016).Human response to scientific equipment set outdoors may vary widelyfrom indifference and inoffensive curiosity to equipment theft and/or vandalism.The latter risk is likely to be amplified in cities, where densities of humans are increased (Corsini et al., 2019; see Table S1 in Supporting Information).The loss of equipment often results in data loss or, more generally, failure in achieving project aims (Meek et al., 2019).It can also threaten the target species of the study: for example, the removal of nestboxes with eggs or nestlings inside (personal observations) will inevitably lead to the death of these individuals.

Preventing the disappearance of field equipment
Steps taken to minimise the loss of scientific equipment in the field are regularly considered and may include: securing items with chains, locks or boxes (Fiehler et al., 2007, Meek et al., 2013), hiding and camouflaging (Jackson and Hutchison, 1985), placing such items in hard-to-reach places, avoiding the set up in highly-frequented areas, (Rovero andMarshall, 2009, Gil-Sánchez et al., 2011) or shortening deployment periods (Meek et al., 2019).Some of these practices can significantly undermine the quality of obtained results: for example, Meek et al. demonstrated that placing camera traps at three meters height to avoid their potential theft may significantly reduce the detection rates of mammals (Meek et al., 2016).Attaching labels with pictograms and messages to inform city dwellers about the study is another approach taken to minimise equipment loss (Clarin et al., 2014).Clarin et al. (2014) have shown the effectiveness of phrasing and verbal tone used on the scientific equipment labels to prevent vandalism and theft events: 60 identical equipment dummiesconsecutively distributed for a week in four urban parks in Munich, Germanywere assigned with one out of three types of labels differing in tone (neutral, personal, or threatening).The study conclusively showed that using a personal label, written in a kind tone and accompanied by a picture of a juvenile squirrel, was the most effective way to reduce vandalism and theft.Although Clarin et al. (2014) introduced a new method for protecting field equipment through verbal labelling, the study lacked real-life context (thus bringing grater liberty in modulating treatment effects, such as label tone) and did not explore other factors contributing to equipment disappearance, such as: Fig. 1.Graphical representation of the RAT-VIVA framework, including predictions on the effect of external factors on scientific equipment disappearance.The definition of a guardian (c) is not limited to a person, it is a concept: "the physical or symbolic presence of an individual (or group of individuals) that acts either intentionally or unintentionally to deter a potential criminal event" (Hollis-Peel et al., 2011).Therefore we treat labels as potential guardians (analogous to the use of "watching eyes" imagery in studies inferring theft reduction and other socially negative behaviours (Dear et al., 2019)).
• environmental variablesthe characteristics of the immediate environment next to the scientific item of interestsuch as human presence, distance to paths or roads, tree cover (Fig. 1) • equipment attributesthe intrinsic properties of an item, such as approximate value, mass, volume, colour or type of installation (Fig. 1 & Table 1).• temporal aspects of equipment disappearance, and specificallythe effect of passers-by habituation with the items (further referred to as novelty) on disappearance probability (Dinnin, 2009) According to the quite distinct field of crime theory (Felson andClarke, 1998, Felson andClarke, 1998), specifically the Routine Activity Theory, all drivers above can play a crucial role in the probability of object disappearance when driven by theft.Therefore, in this study, the following research questions were addressed: 1. Extrinsic factors.Do environmental variables affect the dynamics of deployable field equipment disappearance in urban areas?2. Intrinsic factors.Are some types of deployable field equipment more prone to disappearance? 3. Novelty.Does novelty, measured as the period the equipment spent in a certain area, play a role in the item's disappearance over time?
Based on Routine Activity Theory (acronym RAT) framework and VIVA attributes (embedded within RAT; see Fig. 1 detailed in Methods; Miró, 2014), we predicted that external factors may considerably affect the dynamics of deployable scientific equipment disappearance by providing more or less suitable settings for its removal (Fig. 1).Likewise, some intrinsic properties of the equipment may simplify and encourage or, conversely, impede and discourage its removal (Fig. 1, Table 1).As part of these intrinsic factors, verbal cues in labelling (such as language and tone) may play a role in the disappearance of scientific equipment, as readily investigated by Clarin et al. (2014).Lastly, we also predicted a decrease in scientific equipment disappearance over time, which can be potentially explained by visitors' habituation to the presence of these items, leading to further reduction in removal rates (Dinnin, 2009).

Labelling and Routine Activity Theory implementation
Based on Clarin's et al. (2014) work, we conducted a follow-up study in the city of Warsaw, Poland, a capital city of c. 1.76 million inhabitants in Central Eastern Europe (Statistics Poland, 2018).We implemented the method of equipment labelling while setting up study sites for a longterm research project on the ecology and evolution of two urban birds: the great tit (Parus major) and the blue tit (Cyanistes caeruleus) (e. g.Corsini et al., 2020).We monitored the disappearance of two types of equipment deployed in the field (Fig. 3, Table 1): specifically, nestboxes, where hole-nesting passerines rear their young, and frassboxes, that is collectors of caterpillar faeces, allowing to assess caterpillar abundance (e.g. a key food item for great tit and blue tit offspring) in the immediate vicinity of the collector (Perrins, 1991).
The RAT-VIVA framework, further described in Methods, allowed us to formulate concise and testable predictions about the role of distinct external variables characterising the urban space on the dynamics of scientific equipment disappearance (Fig. 1).Moreover, we also implemented VIVA to compare the attributes of both types of equipment to predict which of these were more prone to disappear (Table 1).To the best of our knowledge, it is the first time that the Routine Activity Theory was empirically tested in the context of urban ecology and evolution research.

Materials and methods
In this study, we refer to all cases of equipment removal by passersby with the neutral word disappearance, which includes both theft or removal for other purposes.

Crime theory-based RAT-VIVA framework and resulting predictions for the selective disappearance of scientific equipment
To better understand the dynamics of equipment disappearance, we took an interdisciplinary approach and applied theory from criminal studies to our research.Tools developed in that field often focus on the issue of infringement of private property, such as theft, and thus, were found useful for this study.Cohen and Felson (1979) formulated the Routine Activity Theory (RAT) -one of the most influential theoretical constructs in the field of criminology (Fig. 1).Put succinctly, instead of focusing on the figure of the criminal and the psychological, biological, or social factors that motivate the criminal act, RAT studies crime as an event (Miró, 2014).The theory highlights the crime's relation to space and time, and emphasizes its ecological nature and the implications thereof.Cohen and Felson (1979) specified three essential elements needed for a criminal event to occur and to explain it: (1) a potential offender with the capacity to commit a crime; (2) a suitable target or victim; (3) the absence of guardians capable of protecting targets (Fig. 1).
An important aspect of the Routine Activity Theory in terms of identifying the causes of scientific equipment disappearance relates to the suitable target concept ((2); Fig. 1).The probability of an item becoming a target is influenced by four attributes: Value, Inertia, Visibility, and Access (acronym VIVA; Fig. 1; Table 1) -described from the point of view of the offender (Cohen and Felson, 1979, Felson and Clarke, 1998, Miró, 2014).All of these four attributes are predicted to translate into object disappearance: • Value, real or symbolic, from the perspective of the offender; • Inertia, referring to size, weight, and shape, or to other physical aspects of the good; • Visibility of target to the offender; • Access, referring to object positioning in space.

Field data 2.2.1. Study sites
We used field data collected as part of a long-term project addressing the impact of urbanisation on the biology of two wild passerines in Warsaw, Poland (Fig. 2).Poland is a democratic, post-communist country in Central Eastern Europe and its capital city, Warsaw, hosts c. 1.76 million people, making it the largest city of the country (Statistics Poland, 2018) and the 11th largest city in the European Union ('Eurostat/Regions and Cities Illustrated (RCI)' 2017).The project started in 2016 and currently consists of eight study sites presented in Fig. 2. For more information on the study sites, please see Table S1, S2 & Text S1, as well as Corsini et al., 2019Corsini et al., , 2022a;;Szulkin et al., 2020b.

Scientific equipment in the field
In this study, we compared the disappearance of two types of scientific equipment: nestboxes, used as breeding cavities in the project, and frassboxesused in a two-year study to quantify caterpillar abundance, assessed by collecting frass (i.e., caterpillar faeces; Fischbacher et al., 1998).
Frassboxes (Fig. 3b & S1) were built using thick A4 white paper (at least 200 g/m 2 ), folded into containers of caterpillar frass by bending the walls and stapling them together (Sudyka et al., 2022).They measured c.W 24 × H 2.5 × D 15.5 cm, weighed c. 15.5 g., and were set in groups of three items (based on nearby nestbox occupancy), fixed to the ground with a long metal nail under the most common tree species of a given location.The frassboxes were used in all but one study site (site B, National Park) where frass assessment was not possible due to permit restrictions.Frassboxes were placed at the exact same locations in both years of the study.Frassbox disappearance data covers two consecutive breeding seasons (from May to mid-June in 2018 and 2019).

Equipment disappearance and its drivers
For the empirical analysis of RAT-VIVA framework-based predictions (Fig. 1, Table 1), we used equipment disappearance data and analysed it in the context of equipment labelling, environmental variables contributing to fine-scale urban heterogeneity, equipment attributes and the time the equipment spent in the field.

Equipment disappearance
Nestboxes were regularly checked each year (2016-2019) during cleanings in October and in February and during the entire avian breeding season (March -July).Frassboxes were placed in the field around the 1st of May (2018May ( , 2019) ) and checked every four days until mid-June (12 checks in 2018 and 11 checks in 2019).Each disappearance of either type of scientific equipment (nestbox or frassbox) was recorded and the missing item was replaced.At each replacement, new Fig. 4. Labels placed at the bottom of 474 nestboxes across 8 study sites: 1 (Informative), 2 (Neutral), 3 (Harsh).Similar labels, printed on paper, were fixed to 121 frassboxes (Fig. 3).
items always had a different type of labelling tone (see below).Items that were damaged were not included in our analyses, as we could not distinguish vandalism from other causes, such as storms or animal interactions.

Labels
To assess the impact of labelling tone (phrasing), we applied the experimental design of Clarin et al. (2014), who tested the effects of contrasting labelling tone on the vandalism and theft of scientific equipment dummies left unattended in 4 urban parks in Munich, for a total of 23 days.In this study, the original approach was applied to an on-going research project set in a gradient of urbanisation in a Central Eastern European capital city (Warsaw).Both nestboxes and frassboxes were labelled with one of three types of labels (8 x 5 cm, metal for nestboxes and paper for frassboxes; Fig. 4).The message on each type of label was reported as follows (text translated from Polish): An equal number of each type of label (a, b, c) was assigned to all nestboxes and frassboxes, which were further evenly spread across space in the study sites.

Environmental axes of urbanisation
We characterised urban heterogeneity by measuring, for every nestbox and frassbox, levels of human presence, tree cover and distance to closest paths and roads.
Human presence.At each study site, fieldworkers counted for 30 s all humans and dogs within a 15 m radius from each nestbox.Counts were repeated 20 times for every nestbox location (for more details, see Corsini et al., 2019).As we did not quantify humans and dogs at each frassbox location, we extrapolated human presence estimates by using the open-source software QGIS (v.3.10.4) and calculated the Inverse Distance Interpolation (IDW) on the human presence data collected at each nestbox (see Corsini et al., 2019).Specifically, the IDW was derived from the geoprocessing toolbox: nestbox locations were fitted as vector layers, while the human presence index was fitted as the interpolation attribute (z).Each IDW was calculated by selecting manually the extent on canvas for each study site separately.Although the IDW is a simple and intuitive spatial interpolation method, it assumes that the relationship between distance and influence between sampling points is constant within each study site, which may not be always valid for human and dog occurrence.
Tree cover.The percentage of tree cover was measured at the nestbox and frassbox level following Szulkin et al., 2020b.After downloading a 20-m-pixel resolution raster-layer Tree Cover Density from Copernicus Land Monitoring Services (https://land.copernicus.eu/sitemap; Forests/Tree Cover Density) -defined as the vertical projection of tree crowns to a horizontal Earth's surfacethe data were processed in QGIS.Averaged value of tree cover (in %) was calculated in a 100 m radius around each nestbox using the Zonal Statistics function.
Distances to paths and roads.The spatial location of each nestbox was recorded using a GPS Garmin Map 64s and all the coordinates were downloaded using the open-source Software DNRGPS Minnesota.Distances to the closest road and to the closest path were both measured in meters and taken from the middle of the road/path and the equipment location (i.e., frassbox or nestbox) using the Measure line tool in QGIS (Szulkin et al., 2020b).

Novelty
To measure the effect of noveltypassers-by habituation to specific items over time (Dinnin, 2009) on the disappearance probability of scientific equipment, we used disappearance data collected over the years (see Methods (3.1)).Frassboxes were left in the field for two months in both 2018 and 2019, and their checks occurred at regular intervalsevery four days from the 1st of May to mid-June.Frassbox novelty in the environment was thus measured as the number of days an item spent in the field before disappearance occurred (starting from the set-up date until the end of frass collection each year).We treated each check separately (consequently, the sample size of frassbox disappearance events was the outcome of the number of frassbox locations multiplied by the number of frassbox checks performed; random effects were fitted to control for repeated sampling of the sample frassbox location (see below)).We were unable to perform an analogous novelty analysis on nestboxes as the study sites were set in different years (see Methods (2.1)) and the checks were performed infrequently over the year and during the field season (see Section 2.2.2)).

Statistical analysis
Statistical analyses were computed in R (v. 3.6.2).Plots and diagrams were built and visualized using ggplot2 (v.3.3.5)(Wickham, 2011) and the open-source software Inkscape (Harrington et al., 2004(Harrington et al., -2005)).R-packages used for each analysis are detailed below.Due to the different experimental setup of nestboxes and frassboxes (see 2.2.2 Scientific equipment in the field, 2.3.1 Equipment disappearance) and their contrasting features (see Table 1), nestbox and frassbox disappearance was always analysed separately.

Effect of labels on scientific equipment disappearance
We used Chi-squared tests to examine whether differences in scientific equipment disappearance are driven by label type.The items (nestboxes or frassboxes) were replaced only if they were found missing or damaged (see 2.3.1 Equipment disappearance for details on equipment checks).A new label, containing a different type of message than the phrasing that disappeared (see Fig. 4) was provided at each replacement, and each label was treated separately in the analysis.Because frassboxes were deployed in both 2018 and in 2019 for the duration of the breeding season (while nestboxes remained in the field throughout the entire duration of the study), we summed the number of disappearances and the total number of available frassboxes within each label type from the first and the second year of the study.

Environmental drivers of scientific equipment disappearance across the urban mosaic
Nestbox disappearance was monitored across 4 years (2016)(2017)(2018)(2019).As multiple disappearances of the same nestbox were rare (N = 4), we fitted a binomial response variable as 0 -never disappearedor 1disappeared at least once (in contrast to the label analysis where all disappearance events were analysed).Because the response variable was zero-inflated (i.e., observed zeros in the dataset exceeded predicted zeros), we specified the binomial family and the ziformula as 1, which applied a single-zero inflation parameter at each observation in the glmmTMB package in R (Brooks et al., 2017).Due to the high correlation (r Pearson > 0.6, p < 0.001, N = 474) between the environmental variables Distance to road, Distance to path, and Tree cover in the nestbox dataset (see Fig. S2 and Table S3), we fitted three distinct models, which differed in including one of these variables in substitution of the other, but contained an analogous model structure in terms of the other predictors: specifically, Human presence (fitted as continuous explanatory variable) and Site (fitted as random effect to control for visitor behavioural differences (and approach towards scientific equipment) in the different sites (e.g.national park vs urban park), Zegras, 2004;Teixeira, 2021, N = 8 levels).
Frassbox disappearance was analysed using Generalised Linear Mixed Effects Models (GLMMs), where the total number of disappeared frassboxes was fitted as a Poisson-distributed response variable (glmer function in lme4, v.1.1-21)(Bates et al., 2015).Because of the strong correlation between Distance to the road and Tree cover (r Pearson = -0.7,p < 0.001, N = 141, see Fig. S2 and Table S3), analyses were performed in two distinct models, which included one or the other environmental variable (as described earlier).The other predictors included in both models were Human presence, Distance to paths (both as continuous), and Year (as a two-level factor).Frassbox ID, nested within Site (N = 7 levels), were fitted as random effects to avoid pseudo replication in both models.Multicollinearity was checked using the R-package performance (v.0.8.0) (Lüdecke, D., 2021) and always resulted in vif < 2. All model fits and assumptions were verified using the DHARMa package (v.0.4.4) (see Fig. S3, Hartig, 2020).Continuous predictors in all models described above were scaled and mean-centred for estimation clarity.All interactions between predictors were tested and retained in the final models only if significant.

Novelty and scientific equipment disappearance
To test whether novelty was associated with disappearance probability, we fitted a Generalised Linear Mixed Effect Model (GLMM) where the binomial-variable (disappeared or not disappeared, coded as 1 or 0, respectively) was set as response, and Novelty (starting with the value of 1 on the day of equipment setup at the start of each breeding season; see Table 4 & Table S5a-b) was included as continuous predictor.Frassboxes were distributed on the ground across the study sites starting when the majority of great tit and blue tit nestlings were in the nest (from 1st of May until mid-June in 2018 and in 2019).Year was included as a twolevel factor (2018 or 2019) and FrassboxID was fitted as a random effect, as the same frassbox was regularly checked (every four days) at multiple times each year (N of checks = 12 and 11 in 2018 and 2019, respectively).Multicollinearity issues, as well as model fit and assumptions, were checked as described in Methods 3.2.

Results
A total of 474 nestboxes and 141 frassboxes were followed over the duration of the study.Because study sites with nestboxes were set up in different years, 45.36% of nestboxes were followed for four years, 40.93% for three years and 13.71% for two years.All frassboxes were followed for two field seasons (see Table S1 and S2).Over the duration of the study, 6.33 % (N = 30) of nestboxes disappeared at least once.In contrast, frassbox disappearance was c. eight times higher, as almost half (49.65 %, N = 70) of all available frassboxes disappeared at least once.

No effect of labelling on scientific equipment disappearance
Label type did not influence the disappearance of either type of equipment.For nestboxes, we had 15 disappearances of informative labelled items (out of 171 available items with this type of label), nine disappearances of neutral items (out of 163) and 10 disappearances of harsh items (out of 170) -χ 2 = 1.48, d.f.= 2, p-value = 0.478.For frassboxes, we observed 91 disappearances of informative items (out of 217), 84 disappearances of neutral items (out of 224) and 95 disappearances of harsh items (out of 212) -χ 2 = 1.80, d.f.= 2, p-value = 0.600.

Nestbox disappearance
When applying the RAT-VIVA framework to scientific equipment disappearance patterns (Fig. 1), nestbox disappearance was associated with the interaction between human presence and tree cover.Thus, nestbox disappearance increased with higher human presence in areas with high tree cover, and decreased with higher human presence in areas with low tree cover (Fig. 5, Table 2).Nestbox disappearance was not associated with nestbox distance to roads or distance to paths (see Table S4).

Frassbox disappearance
As in the case of nestboxes, frassbox disappearance was also positively associated with the interaction between human presence and tree cover (Fig. 5).Overall, frassbox disappearance increased with human presence, and particularly so in areas with high tree cover (Fig. 6).At the same time, frassbox disappearance was also associated with the interaction between distance to roads and distance to paths.Specifically, the disappearance rate increased closer to roads, if those were located further away from paths (Table 3).

Novelty as a predictor of scientific equipment disappearance
Frassbox disappearance was highly sensitive to the novelty effect, as the majority of frassboxes disappeared in the initial stage of the study (at

Table 2
Zero-inflated GLMMs models with binomial distribution testing the association between probabilities of nestbox disappearance (fitted as never disappeared, 0, or disappeared at least once, as 1, in four years) and environmental variables.Models were analysed using the R-package glmmTMB.Continuous predictors were mean-centred and scaled (sc).Interactions in model structures and variables are indicated with an asterisk.Significance levels are indicated in bold: *p < 0.05, **p < 0.01, ***p < 0.001.

Environmental drivers of nestbox disappearances
Family: zero-inflated with binomial (logit) distribution, N nestboxes = 474 Model structure: Disappeared nestbox (0/1) ~ Human presence sc * Tree cover sc .the start of the breeding season), when items were deployed in the field; the frequency of these disappearances subsequently and gradually decreased (Table 4, Fig. 7).Most items (N 2018 = 21, N 2019 = 24) disappeared in the first, four days long period of frass collection in the season (Table S5a&b, Fig. 7).

Discussion
In this study, we empirically applied for the first time an eventfocused framework developed in the context of criminal studiesthe Routine Activity Theory (RAT; Felson and Clarke, 1998) to urban evolutionary ecology research.By doing so, we were able to successfully predict and statistically verify which external variables and equipment attributes are associated with field equipment disappearance.
The potential offender element of RAT (Fig. 1(a); modelled by human presence), interacted with visibility (Fig. 1(b.3); represented here by tree cover) to influence the disappearance of both nestboxes and frassboxes.Thus, in highly-frequented places, people were more likely to remove the scientific equipment if they felt less seenfor example in cases where trees could act as a cover.Indeed, urban vegetation is reported to have a significant influence on rates of a variety of crimes: a study reporting property crimes over multiple cities from South Africa fully aligns with our findings, confirming that higher tree cover was associated with increased property crime (Venter et al., 2022).
The absence of guardians element of RAT (Fig. 1(c)) can be modulated by label tone.In contrast to Clarin et al. (2014), we did not find a significant difference in the disappearance of either type of equipment based on label types (see Results).A possible explanation is that the labels used in this study needed to be applied to a real-life scenario of efficiency and informativeness, worded specifically for the sake of the long-term study it was designed for.Consequently, the messages were more explicit.For example, labels used in this study did not include any information about the possible consequences of equipment theft ("Every theft will be reported to the police!") or images, such as that of a juvenile squirrel (as performed by Clarin et al., 2014).Research has shown that cuteness can strongly influence human behaviour (Kringelbach et al., 2016): by seeing an image of animals with juvenile features, people may be less likely to steal scientific equipment, especially if that could harm the animals.Our labels were also not as radically different from each other as in the hypothetical scenario assessed by Clarin et al. (2014).Therefore, the overall effect of labels on equipment disappearance in our study system may be weaker.Importantly, relative to Clarin et al. (2014), this study was run on two different types of equipment, considerably larger sample sizes (474 nestboxes and 141 frassboxes vs 60 equipment dummies), and greaterand more diversified in terms of land usenumber of study sites (eight sites set in a gradient of urbanisation vs four urban parks), as well as a longer study period (two to four years [nestboxes] and c. 90 days [frassboxes] vs. 24 days in Clarin et al. (2014)).Consequentlyfollowing recommendations in Clarin et al. (2014) regarding the need of replication in other regions of the worldour findings suggest that labelling may not be a sufficient method to protect the equipment if applied to a realistic, long-term project across different cultural and spatial contexts.
Based on the four VIVA attributes (value, inertia, visibility and access), the RAT framework allowed us to predict that nestboxes are a type of   ).This prediction was confirmed, and reflected in the percentage of disappearances recorded as binary events (nestbox disappearance rate of 6,33 % vs frassbox disappearance rate of 49,65 %).As expected by the framework features, a small, light, eye-catching and easily accessible item such as the frassbox is significantly more prone to negative interactions with people than the unhandy, harder to spot and harder to reach nestbox, despite a considerable difference in valuea woodcrete nestbox is more visually valuable than a paper frassbox, which is also reflected in an items' price (Table 1).Importantly, the VIVA framework allowed us to determine which attributes are influencing the disappearance of items within the same type of equipment (see suitable target, VIVA; Fig. 1).
• Value, determined by novelty, influenced the disappearance of frassboxes (Fig. 7).As anticipated (Fig. 1), most disappearances occurred in the early period of equipment setup, after which the numbers gradually decreased.This strongly suggests that city dwellers lost their original interest in interfering and removing the equipment once it was familiar (Dinnin, 2009).Note that this might not be the case with considerably more valuable equipment, such as camera traps, where the financial cost of the equipment itself can interact with the temporal dimension of novelty (Meek et al., 2019).• Inertia is consistent within equipment type (as the volume and weight of all items, e.g.nestboxes or frassboxes, is the same), thus it could not influence the disappearance dynamics at this level.• Visibilitydiscussed above (see the second paragraph of Discussion).
• Finally, accessmeasured by an item's distance to the closest path or roadwas associated with the disappearance of frassboxes, but not nestboxes.These findings suggest that researchers, site-managers, and stakeholders should consider securing easily reachable field equipment, especially in accessible areas (see Introduction for examples of equipment securing methods).At the same time fixing the equipment beyond the reach of urban dwellers may also decrease the chance of equipment disappearance (as in the case of nestboxes).
Our results thus confirm that the disappearance of field equipment, as tested in the context of a real-life research project run in the urban field, is the result of an interaction between external variables (Fig. 1)such as potential offender (human presence), relative value (novelty), visibility (tree cover), access (distance to paths * distance to roads) -and equipment attributes (Table 1), that is an item's assumed value (here approximated by price), inertia (volume and weight), visibility (colour), access (type of installation).Thus we show that environmental characteristics of specific study site settings may facilitate or impede equipment disappearance across urban areas.Similarly, some types of deployable scientific equipment may be predisposed for negative interactions with people.Moreover, the interactions between the twostudy site environmental features and equipment attributescan influence disappearance dynamics.For example, the interaction between distance to paths and distance to roads (both environmental features) was only associated with frassbox disappearance; this is likely due to the differences in the type of installation -either on the ground on high up in the trees (an equipment attribute).
This study highlights the importance of a multidisciplinary approach while conducting ecology and evolution research in areas occupied by humans (Des Roches et al., 2021).Overall, the successful application of the Routine Activity Theory (RAT), and the VIVA framework embedded within this theory, indicates its considerable potential for cross-cultural validity in the field of urban ecology and evolution research.It thus offers the opportunity to examine equipment and study locations under the RAT-VIVA framework, which is an easily applicable and inexpensive way to better understand the risks of equipment disappearance, thereby strengthening the potential for informed project planning and as a result safer, and more effective studies not only in the urban space but also in other areas where humans abound.

Conclusions
This study was conducted in a European capital city (specificallya democratic post-communist Central Eastern European capital city), and as such provides valuable insight into processes (equipment disappearance) that may carry both local features (shaped by specific sociocultural norms), as well as universal ones pertaining to human psychology.While our results align with a study on property crime committed in urban areas from the Global South (Venter et al., 2022), they may differ depending on geographical and cultural contexts.Urban design (e.g.walkability, access to green areas) and social factors (e.g.socio-economic variation, sense of community) may play an important role in shaping equipment disappearance dynamics.However, the RAT-VIVA framework presented here is fully amenable to modifications of original predictions based on such variation, and thus may be applicable to any location of interest.We encourage other researchers to conduct

Fig. 2 .
Fig. 2. Map of study sites located in a gradient of urbanisation in the capital city of Warsaw, Poland.These include: a suburban village (a), a national park (b), an urban forest (c), two residential areas (d and f), an urban woodland (e), an urban park (g) and a university campus (h).Photo and figure credits: Michela Corsini.Map acquired from Google Maps.
(a) Informative: We are studying the effect of cities on the life of birds.Please do not disturb!Thank you.Centre of New Technologies University of Warsaw (b) Neutral: no message (c) Harsh: Please, do not disturb!Centre of New Technologies University of Warsaw

Fig. 5 .
Fig. 5. 95% confidence intervals reporting the probability of nestbox disappearance in relation to human presence and tree cover.To visualise the interaction, we used the lm function in ggplot2; tree cover was categorised based on the median value of the full dataset as high (N = 237, values above 26.4,mean ± SE, 63.2 ± 1.01; green line) or low (N = 237, values between 0 and 26.4,mean ± SE, 7.46 ± 0.45; black line).

Fig. 6 .
Fig. 6. 95 % confidence intervals reporting the predicted number of disappeared frassboxes in relation to human presence and tree cover.To visualise the interaction, we used the glm function in ggplot2, and specified method.args=list('family=poisson') in the function geom_smooth.Tree cover was categorised based on the median value of the full dataset as high (N = 140, values above 27.42,mean ± SE, 51.68 ± 0.9; green line) and low (N = 142, values between 0 and 27.42, mean ± SE, 8.65 ± 0.73; black line).

Fig. 7 .
Fig. 7. Frassbox disappearance was highest when frassboxes were novel in the environment.Barplots indicate the cumulated number of disappeared frassboxes per check in the field (over a total of 141 deployed frassboxes).The trend line was generated using a linear regression (stat_smooth function in ggplot2).Similar trends were found if the models were run separately for both years (2018 & 2019; see Table S5a-b).
Frassbox disappearances (Total) ~ Distance to roads sc * Distance to paths sc + Human presence sc + Year.

Table 4
GLMMs testing the association between novelty and frassbox disappearance.Novelty was calculated as the entire periodin daysthe frass collectors were left in the field (1 = 1st of May, each year, as 2018 or 2019).Continuous predictors were mean-centred and scaled (sc).Model details are reported in Methods (Section 3.3).Interactions in model structures and variables are indicated with an asterisk.Significance levels are indicated in