2.0 Methods

Field studies were conducted at 36 total bridges including 12 each in California, Florida, and Ohio. The original study plan divided structures into six structures per state used to examine physical exclusion efficacy and six structures per state used to examine acoustic deterrent efficacy. Within each cluster of structures, three received sham treatments (i.e. a control) while three received actual treatments (deterrents or physical exclusion). Total numbers of sham and operational acoustic deterrents were based on the number of expansion joints and width of each structure. Selected structures contained potential bat roosting areas positioned in observable areas, facilitating a structure monitor’s (field biologist’s) ability to identify abandoned bat pups. Targeted structures also featured observable roosts and harbor regionally abundant (common) species such as big brown or Brazilian free-tailed bats.

The research team was unable to access suitable California structures during the 2023 field season. Thus, studies in 2023 focused on bridges in Florida and Ohio. After reviewing results from the 2023 field studies, a 2024 study was completed at 12 bridges in California that focused entirely on effectiveness of small acoustic deterrents.

2.1 2023 Study Design (Florida and Ohio)

The 2023 experimental treatments (exclusions/acoustic deterrents) were applied in specific structures (12 each) in Florida and Ohio chosen because 1) the structures were known bat roosts and 2) the structures were similar to other structures known to contain bats. Within each state, replicates were grouped based on structure similarity and location. Treatments were applied in early spring 2023 (to avoid impacts to most bats) with structures monitored until 15 August 2023.

Field efforts in Florida and Ohio demonstrated traditional physical exclusion was effective at removing bats from structures and remained effective as long as the exclusion materials remained intact. Conversely, acoustic deterrent efficacy used in the study was inconclusive.

Thus, the research panel recommended studies in California focus entirely on the effectiveness of acoustic deterrents and subsequently approved a plan dividing treatments at the 12 California structures into six with actual acoustic deterrents and six with sham deterrents.

2.2 2024 Study Design (California)

The 2024 experimental treatment focused on deterrents applied in the Central Valley and Gold Country regions of California. Specific structures (12 total) in California were included in the sample based on 1) known bat presence and 2) similarity with other

structures. Twelve California structures were identified, and replicates grouped based on structure similarity, location, and bat species present.

2.3 Initial Structure Selection

Identification of structures included in the study was an iterative process. First, a series of bridges suitable for use were identified in all three states and these were transmitted to the research panel as a Structure Selection Memorandum. Following panel review, structures were assigned to treatment and control groups and a Site Selection Memorandum (covering all three states) was submitted to NCHRP on 30 November 2022. The panel reviewed the document through 20 December 2022 and requested additional information on selected structures. Responses with additional information were provided to the panel and discussed during an in-person meeting on 13 March 2023. Approval was granted to proceed with material acquisition and installation of experiments in Florida and Ohio following the in-person meeting. All assigned structures in Florida and Ohio received treatment in 2023, and a report on findings was submitted to the panel on 27 February 2024. California field efforts were delayed until spring 2024 resulting from challenges associated with contracting, permitting, and weather.

On 23 March 2023, California Department of Fish and Wildlife (CDFW) issued Dr. Doty a permit to complete the study. California Department of Transportation (Caltrans) required exact structure identifications for final Right-of-Entry permit applications, requisite prior to implementing the experiment in California. Final California structure selection was approved by the panel on 5 January 2024 and the Caltrans permit application was submitted on 30 January. Caltrans granted preliminary approval on 21 February 2024 and the permit was formally granted on 6 March.

As outlined in the approved study plan, implementation of field control and treatment applications in California concluded prior to 15 April 2024. During the implementation process, alterations were made to California structure selections outlined in the Task 4b California Site Selection Memorandum and are detailed below.

2.4 Florida Structure Selection Modifications

The panel requested additional structure surveys in Florida confirming each study structure was occupied by bats. Based on field survey results, structures without confirmed bat presence were replaced with alternate structures known to contain bats. Two additional pre-selected structure pairs in Florida became active construction sites prior to installation of experiments and these structures were also replaced with alternatives. An additional structure on the pre-selection list was identified as containing numerous bats and was approved by Florida DOT District 7 during initial vetting. However, subsequent investigation revealed the structure is within a designated natural area and used for environmental education programs. Thus, the final list of Florida sites includes six structures from the initial proposed study site list presented to the Panel on 30 November 2023 and six structures either drawn from the alternates list or identified during subsequent field studies. All structure pairs selected in Florida were co-located pairs (i.e., geographically adjacent).

2.5 Ohio Structure Selection Modifications

Preferred Ohio structure options remained the same throughout the initial selection memo and presentation of the detailed field test plan. However, on 16 February 2023, in response to an Ohio-specific study plan, ODOT indicated some selected structures were not under the state agency’s direct control, thus triggering coordination with appropriate township and county engineer offices. Efforts to contact county engineering offices for two different structures were unsuccessful. Flood-water levels at a third Ohio structure precluded safe installation of equipment. Thus, three alternate structures identified during subsequent field studies as housing big brown bats served as replacements for the initially identified Ohio sites. No structure pairs selected in Ohio were co-located pairs (i.e., structure pairs possessed similar but not identical traits and were not geographically adjacent).

2.6 California Structure Selection Modifications

Alterations made to the selection of California structures outlined in the Task 4b California Site Selection Memorandum are as follows:

• A field visit to CA-03 treatment structure (SFN 46-0056L) revealed high water conditions and no external features adequately accommodating solar panel installation. As this combination rendered the structure unsuitable for application of the treatment, SFN 46-0056L was designated as a control structure for CA-06;
• Structure SFN 50-0433, the originally designated control for structure pair CA-06, was instead assigned as replacement treatment structure for CA-03. Within one hour of installation, treatment equipment was stolen from SFN 50-0433. Subsequent safety and financial concerns led to exclusion of SFN 50-0433 for either treatment or control;
• Thus, CA-03 treatment structure was subsequently appointed to SFN 50-0206, originally listed as an alternate structure in the Task 4b California Site Selection Memorandum.

2.7 Treatment/Control Applications

2.7.1 Acoustic Deterrents

Three replicates in Florida and Ohio, and six replicates in California focused on testing acoustic deterrent effectiveness. Acoustic deterrents were installed in Florida and Ohio from 1 March through 13 April 2023 and in California from 18 March through 13 April 2024 when bat population expectations are low and well before any pup presence thus precluding a need to decommission acoustic devices to prevent bats from abandoning their young.

To reduce potential sources of variation, a single brand and model of acoustic deterrent was used for the study. The BD100 Ultrasonic Bat Deterrent by Binary Acoustic Technology (binaryacoustictech.com) is marketed as a residential device applying a proprietary wideband ultrasonic masking technology. The manufacturer asserts the device interferes with bat sonar, complicating navigation, and thus discouraging bat

presence near the deterrent. These devices have previously been used to exclude bats from structures in Arizona. The manufacturer describes the effective radius per device as 12 to 15 feet (3.66 to 4.57 m) and deterrent dimensions are approximately 6 X 2.6 X 2.6 inches (15.24 X 6.60 X 6.60 cm). Each device requires 2 Watts or less to power and has a maximum transmittal power of 96 dB SPL @ 50 KHz. Unlike other commercially available deterrents, these devices emit broad-spectrum ultrasound (thus not targeting any species of bat) within a limited range to minimize risk of impacts beyond the target structure. The BD100 is also relatively inexpensive (approximately $495.00 per unit) and potentially powered by a variety of power sources ranging from packs of “D-cell” batteries for short-term deployments in remote areas to 120 V alternating power processed through a power inverter for permanent applications.

Sham deterrents were made of sections of 4- X 4-inch and 2- X 4-inch (10.2- X 10.2-cm and 5.1- X 10.2-cm) lumber or empty water bottles connected by a wire and painted black to mimic the appearance of detectors and battery containers. Sham deterrents installed on an individual structure were made of the same material (lumber or empty water bottles). Similar to actual deterrents, mounting techniques varied between structure types.

Because bats in Florida structures primarily roosted in expansion joints and abutments, acoustic deterrents were mounted to each structure in two ways. Acoustic deterrents were mounted along the side of the structure with sound focused down the joint or hung vertically below joints running perpendicular to the direction of the structure and both abutments. Sham and operational acoustic deterrents hung from spring clips wedged into expansion joints and were supported (as needed) by quick-dry silicone. Each acoustic deterrent was wired to a battery box containing eight D-cell batteries and the battery box was fixed to the top of structure support pillars using plastic zip ties. Sham deterrents were attached to sham battery containers using electrical wire used on operational acoustic deterrents.

In Ohio, bats primarily roost in gaps between box beams and in abutments; however, gaps typically ran parallel to the structure’s direction. Most operational acoustic deterrents and battery boxes were suspended from Ohio structures using rope tied to either the guard rail of the bridge or to structures (usually trees) under the bridge. Total numbers of sham and operational acoustic deterrents were based on the number of expansion joints and width of each structure. Numbers and placement of sham deterrents on control structures followed the same process as treatment structures.

In California, bats roosted in various structural components, most commonly expansion joints, crevices and cracks within the structure, weep holes, and closed box girders. Depending on the structure’s physical properties, acoustic deterrents were mounted in various ways. Attachment methods included a) mounting deterrents along the side of the structure and aiming the deterrent into the roosting space, b) using binder clips to fasten deterrents beneath weep holes or cracks known to serve as roosts, c) fastening deterrents to the top of long PVC pipes mounted on the ground to project the deterrent sound upward and into a roost, or d) hanging deterrents on the sides of structure walls using paracord. Total numbers of operational and sham deterrents were based on the

number of expansion joints and width of each structure. Number and placement of sham deterrents on control structures followed the same process as treatment structures.

Initial calculations indicated D-cells discharged in 10 to 20 days. However, under field conditions during 2023 in Florida and Ohio, acoustic deterrents functioned for only five days. Thus, battery packs were replaced by high-capacity batteries that were either charged with a solar panel or regularly replaced. A key question raised by the 2023 study was whether bats returned to structures once deterrents lost power. Thus, in California, higher capacity batteries recharged with a solar panel were used for all sites, which greatly reduced efforts required to maintain deterrent functionality between monitoring visits.

2.7.2 Physical Exclusions

Three replicates within Florida and Ohio focused on testing physical exclusion effectiveness. Physical exclusions were applied 1 March to 13 April 2023 when most bats were away from the site ensuring application did not cause bats to abandon young. Physical exclusion placement targeted areas with openings of 0.5 inch (1.27 cm) or greater in size. Unoccupied potential roosting areas were sealed prior to areas containing bats to prevent displaced bats from relocating within the same structure. Field crews accessed targeted gaps and crevices on foot or by ladder. Physical exclusion materials comprised a combination of tubes of foam pipe insulation, quick-set expanding spray foam, quick-set caulk, and one-way exits created with PVC pipe. Benefits of using foam pipe insulation include the ability to be compressed or trimmed to fit a variety of gap sizes, especially linear crevices. The downside to using foam pipe insulation is the lack of adhesive can allow insulation to dislodge in cases of heavy water flow or from manual disturbance. Foam pipe insulation also is less useful in large gap areas.

Expanding foam sealant was often used to secure foam pipe insulation and to seal hollow areas where foam pipe insulation was inappropriate. Spray foam used for the current project is tack-free within 3 to 10 minutes and solid enough to cut in one hour. Expanding foam could entangle any bats that contacted the foam before it dried. In the absence of escape routes, both pipe insulation and expanding foam could potentially entomb bats. Structures where expanding foam was used as the primary exclusion tool were checked for bats by a qualified bat biologist. Installation of the expanding foam only occurred if no bats were present at the time of exclusion.

If bats were present at the time of the exclusion, expanding foam and silicone were only used to secure pipe insulation and one-way exits. Pipe foam was pushed into the gaps where bats entered/exited roosts and then cemented into place with expanding foam or silicone.

Physical exclusion controls were installed using activities comparable to treatment exclusion installation. Specifically, a similar amount of time was spent under the structure. Wearing gloves, biologists probed cracks prior to installing exclusion materials and blew canned air into unreachable spaces.

2.8 Monitoring

Bat activity at structures was monitored using emergence counts, direct observation of roosting bats (referenced as monitoring counts), and sampled with acoustic bat detectors. Monitoring efforts were planned around three periods: initial, weekly, and monthly monitoring periods. Initial monitoring occurred for the four days immediately following installation of control/experimental treatments. Weekly monitoring was conducted once a week for four consecutive weeks following the initial monitoring period. Monthly monitoring was conducted once a month following the weekly monitoring period until 15 August. Thus, the planned monitoring included 12 assessments of bat activity during the study. During maintenance activities including replacement of batteries on acoustic deterrents and repairs to physical exclusions, biologists would also complete and record direct observations of roosting bats. Monitoring and maintenance were discontinued when inclement weather, floods, or wildfire made it unsafe to access the structures.

2.8.1 Emergence Counts

Emergence counts were completed following guidance issued by the U. S. Fish and Wildlife Service (USFWS 2023; 2024) and afforded safe and accurate visual counts of emerging bats during periods when bats are reliably active. Per the guidance, emergence counts were suspended during poor weather conditions, including temperatures below 50°Fahrenheit (10°C), wind speeds greater than 9 miles per hour (14.48 kph), and during rain (USFWS 2023; 2024).

As noted above, emergence counts were implemented on a planned schedule that included initial (once a day for four consecutive days following implementation), weekly (once a week for four consecutive weeks following initial monitoring period), and monthly (minimum of once a month until 15 August) monitoring periods. Supplemental emergence counts were also completed whenever possible. During each count biologists recorded the number and behavior of emerging bats.

2.8.2 Monitoring Count Modifications

In addition to emergence counts, biologists (typically 1 or 2 individuals per structure) also used boroscopes and direct visual observations (with or without the aid of a spotlight) to record the number of bats at each roost.

Roosting counts were conducted prior to and after each emergence count and were also opportunistically completed during unsuitable weather conditions and during daylight visits for maintenance activities. Emergence and roost counts were combined to estimate the number of bats present and examine general trends in treatment and control structures over time. Importantly, bats were not always visible in structures despite efforts to locate structures where bats would typically be visible. Bats were sometimes visually observed roosting in structures but did not emerge that night. Where the number of estimated roosting bats before a count was lower than the number of bats observed emerging, the emergence value was used. Where the number of roosting or emerging individuals exceeded 500 bats, estimates of observed bats were rounded and considered an index.

2.8.3 Acoustic Monitoring

Acoustic monitoring stations consisted of a bat detector (Wildlife Acoustics SM4BAT-FS; Wildlife Acoustics SM Mini Bat) positioned at least 328.08 feet (100 m) away from a structure along a travel corridor intersecting the structure. Preferred acoustic monitoring sites had limited clutter to maximize quality of the calls recorded (Britzke 2004, Broders et al. 2004) and regular bat traffic. Sites were 1) borders of riparian corridors running through open landscapes; 2) fencerows adjacent open habitats; 3) utility corridors; 4) water sources including ponds and open stretches of streams; or 5) other open linear corridors, including logging and other woodland roads/trails. Typically, areas with high amounts of acoustic clutter created by wind, vegetation, insects, other bats, open water, sheer rock surfaces, or high-tension lines were avoided. For this study, detectors were placed along potential flyways servicing acoustic treatment and control structures to determine if bat activity was unaffected by acoustic deterrents within flyways.

In general, detectors were positioned at least 15 feet (5 m) in any direction from obstructions and in areas with minimal or no vegetation occurring within 33 feet (10 m) in front of the microphone. Detectors were placed parallel to woodland edges and at least 49 feet (15 m) from known or suitable roosts. Microphones were elevated to a minimum of 10 feet (3 m) above ground level. Where possible, detectors were placed a minimum of 656 feet (200 m) apart. In Florida, a single bat detector was placed along the flyway servicing each structure pair (treatment and control). Ohio and California structures were not geographically paired; thus, each structure received a bat detector to monitor the flyway.

2.9 Analysis

2.9.1 Monitoring Count Data

Statistical analyses on monitoring count data were completed using program R (version 4.3.2; R Core Team; Vienna, Austria). Regression models were used to determine relative effects of measured predictor variables on the response variable of bat presence at structures. The response variable is the difference in the maximum number of bats present (maximum value of emergence and roost counts) between initial and final monitoring counts (up to 12 count event dates or until end of treatment). The response variable was standardized to avoid heteroscedasticity due to higher numbers of bats in Florida bridges.

Potential predictor variables included: 1) treatment type (acoustic or physical installation), 2) treatment versus control installation, , 3) structure length, 4) structure width, 5) structure deck area, 6) structure type (stringer/multibeam, box beam, or beam), 7) service under structure (roadway or waterway), 8) state (Florida, Ohio, or California), and 9) bat species present (Brazilian free-tailed bat, big brown bat, canyon bat, Yuma Myotis, or “other” in the case of CA-01T where no bats were observed either roosting or emerging from the structure during the study period). Variables “deck area” and “structure length” were autocorrelated (p<0.05) indicating variables successively explain similar response variation, thus variable “deck area” was omitted from analysis. A subset of variables (“bat species”, “state”, and “service under structure”) were uniquely standardized (e.g., big

brown bats represent the only species recorded at Ohio structures servicing waterways and Brazilian free-tailed bats at Florida structures servicing only roadways) leading to autocorrelation of all combinations of the three categorical variables (phi-coefficient = 1, p ≤ 0.05). Thus, the term “state” was included in analysis, excluding “service under structure” and “bat species”. Following removal of autocorrelated variables, predictor variables used in analysis included: 1) treatment type (acoustic or physical installation), 2) treatment versus control installation, 3) structure length, 4) structure width, 5) structure type, and 6) state.

Multiple or single linear regression models were fitted using all permutations of the remaining seven predictor variables. The most parsimonious model explaining the difference in number of bats present between initial and final monitoring counts was determined using Akaike’s Information Criterion adjusted for small sample sizes (AICc; Program R Package AICcmodavg; Mazerolle 2023).

To better understand acoustic deterrent effects, only the best predictors of bat presence in California were used. For the California-only analysis, predictor variables “treatment vs. control”, “species”, “structure length”, “structure width”, “service type”, and “structure type” were used. Neither state nor service type under structure were included as 11 out of 12 structures in California were over water.

A generalized linear model (GLM) was subsequently used to create a linear equation explaining bat presence on structures with the most parsimonious AICc model.

2.9.2 Acoustic Monitoring Data

All call files were downloaded and processed through Kaleidoscope Pro (KPro) software (classifier v5.6.8 Wildlife Acoustics, Concord, Massachusetts). The software extracted parameters including the frequency, time, and slope components of each recorded call “pulse”. Each pulse was then assigned a species-level identification, with the entire sequence assigned based on the species most frequently identified. In some cases, even very low confidence identifications were of value, including instances where biologists attempted to locate rarer species such as the spotted bat (Euderma maculatum). In other cases, such as academic research or studies aimed at regulatory compliance, a more complete level of identification is required. The software allowed users the option of tightening or loosening the stringency of the rule governing species-level identifications and could also be adjusted to restrict the analysis to only species expected present (to avoid misidentifications).

KPro software made use of maximum likelihood estimators (MLE), a multivariate statistical technique used to test the strength of a proposed relationship based on known or assumed error rates. In this case, the proposed relationship was the presence of protected bats identified by analytical software. The MLE accounted for the number of call sequences identified as a species and compares it to the number of call sequences identified as belonging to a similar species based on the assumed error rates. Post-hoc MLE p-value classifications were either significant (p ≤ 0.05) for likely species presence, or non-significant and less likely present (p > 0.05). A p-value > 0.05 indicates KPro did

not identify pulses consistent with the species on the data provided. As KPro identified more call sequences for a species and/or less calls sequences of species with overlapping characteristics, identification confidence increased. Assumed error rates were obtained by testing software packages against libraries of known calls. The goal was to provide a mechanism to eliminate errors resulting from misclassification.

Species of interest, such as California Species of Special Concern (SSC) as classified by CDFW and identified by KPro software, were also reviewed in SonoBat software (classifier v30 SonoBat, Humboldt, California) for a more thorough representation of the individual call file.

During 2023, nine species were potentially present in the immediate and surrounding areas of structures in Florida including: big brown, eastern red (Lasiurus borealis), northern yellow (Lasiurus intermedius), Seminole (Lasiurus seminolus), Pallas’s Mastiff (Molossus molossus), southeastern (Myotis austroriparus), evening (Nycticeius humeralis), tricolored (Perimyotis subflavus), and Brazilian free-tailed bats. Additionally, nine species were potentially present in the immediate and surrounding areas of structures in Ohio including: big brown, eastern red, hoary (Lasiurus cinereus), silver-haired (Lasionycteris noctivagans), gray, little brown (Myotis lucifugus), northern long-eared, Indiana, evening, and tricolored bats.

During 2024, 19 species were potentially present in the immediate and surrounding areas of selected California structures including: pallid (Antrozous pallidus), Townsend’s big-eared (Corynorhinus townsendii), big brown, spotted, western mastiff (Eumops perotis californicus), silver-haired, eastern red, hoary, western red (Lasiurus frantzii), western yellow (Lasiurus xanthinus), California Myotis (Myotis californicus), western small-footed (Myotis ciliolabrum), western long-eared (Myotis evotis), little brown, fringed (Myotis thysanodes), long-legged (Myotis volans), Yuma (Myotis yumanensis), canyon (Parastrellus hesperus), and Brazilian free-tailed bats. Species deselected from KPro’s California region pack based on understood ranges and occurrence information included California leaf-nosed (Macrotus californicus), Arizona (Myotis occultus), cave (Myotis velifer), pocketed free-tailed (Nyctinomops femorosaccus), and big free-tailed (Nyctinomops macrotis) bats.

The zero (balanced/neutral) sensitivity setting was used for the analysis and classifier package, allowing calls to be classified to the species-level based on the greatest percentage of the call classified as a single species. Acoustic data are provided electronically upon request and are stored by ESI on designated solid-state hard drives for seven years.

2.10 Financial Comparisons

Financial estimates were provided to DOTs with information needed to facilitate a structured decision regarding relative costs of physical exclusions and acoustic deterrents as described in Section 2.7 above, and maintaining treatments, and implementing monitoring efforts described in Section 2.8. Clean comparison requires removing several confounding expenses from amounts provided. Values excluded from calculations consist

of time spent traveling to and between sites, time spent training new staff to complete treatments and controls, and use of higher cost staff. Implementation costs are tracked on a per treatment basis, but maintenance and monitoring efforts for multiple structures are often pooled.