WEST staff will be in attendance at The Wildlife Society Virtual Conference this week and will be participating in the following on-demand presentation and live Q&A sessions. “Stop” by our virtual booth to say hello!
“Machine Learning and Computer Vision for Wildlife Population Management” – Mikey Tabak, Ph.D., Presenter
Live Q&A Session: November 2nd, 3:00-4:00 PM EST.
The ability to observe wildlife remotely is central to population management. Camera traps, acoustic detectors, and flights of crewless aircraft (drones and fixed-wing aircrafts) are often employed in remote sensing projects. These remote sensing operations collect large amounts of images or sound files that must be processed and analyzed before the data can be used to inform management decisions. Deep learning provides a tool to automatically and rapidly process these data. We used computer vision and deep learning to build models that automatically process images and acoustic recordings of wildlife. Specifically, we evaluated the effectives of using image classification, object detection, and object segmentation computer vision models to analyze remote sensing data. We found that object detection models are able to rapidly classify, detect, and count animal species in camera trap images, as well as images from flights of drones and fixed wing aircraft, with accuracies of 90-98%. We also found that image classification models were able to classify bat species from acoustic recordings with 93% accuracy. We found that object detection and object segmentation models are more effective than classification models at removing empty images from datasets, and they have the advantage of automatically counting and locating animals within images. These models can rapidly (100-1,000 images per minute) process data on laptop computers.
The ability to automatically process images and acoustic recordings allows researchers to expand and more rapidly obtain data for wildlife monitoring programs. Furthermore, these models can be deployed for continuous monitoring and allow for data acquisition in “real time.” In this presentation, I will introduce these different types of deep learning computer vision methods, provide results from our research, and describe applications in wildlife management.
“Curtailment and Acoustic Deterrents Reduce Bat Mortality at Wind Farms” – Rhett Good, Presenter
Live Q&A Session: November 4th, 12:00-1:00 PM EST
Presentation Description: The impacts of wind-energy on bat populations is a growing concern because wind turbine blades can strike and kill bats, and wind turbine development is increasing. We tested the effectiveness of an acoustic bat deterrent developed by NRG Systems combined with curtailing turbine blades when wind speeds were below 5.0 m/sec at 2 adjacent wind-energy facilities in northeast Illinois during the fall migration period (1 Aug–15 Oct) of 2018, compared to turbines that operated at manufacturer cut-in speed and lacked acoustic deterrents. We measured the effectiveness of the acoustic deterrent and curtailment with carcass searches of cleared plots below turbines, and field trials that estimated searcher efficiency and carcass persistence. The effectiveness of each management action was estimated using a generalized linear mixed-effects model with several potential co-variates. Overall bat fatality rates were 66.9% (90% confidence interval [CI]: 54.5–75.9%) lower at curtailed turbines with acoustic deterrents compared to turbines that operated at manufacturer cut-in speed. Curtailment and the deterrent reduced bat mortality to varying degrees between species, ranging from 58.1% (90% CI: 41.1–70.3%) for eastern red bats to 94.4 (68.8–99.0%) for big brown bats. Hoary and silver-haired bat mortality was reduced by 71.4% (90% CI: 55.8–81.5%) and 71.6% (90% CI: 56.1–81.7%), respectively. Curtailment alone reduced overall bat mortality by 42.5% (90% CI: 16.1–60.6%), but also varied by species. Curtailment alone effectively reduced hoary bat and eastern red bat mortality, but not silver-haired bat mortality. We estimated the additional reduction in mortality resulting from the using the acoustic deterrent and curtailment under the assumption that curtailment and the acoustic deterrent would have reduced mortality by the same percentage at adjacent wind-energy facilities. Acoustic deterrents resulted in 31.6%, 17.4%, and 66.7% in additional reductions of bat mortality compared to curtailment alone for eastern red bat, hoary bat, and silver-haired bat, respectively.
“Mixed Evidence of Lake Effect at PV Solar Energy Projects in Southern California” – Karl Kosciuch, Ph.D, Presenter
Live Q&A Session: November 4th, 3:00-4:00 PM EST
Presentation Description: The development of photovoltaic (PV) utility scale solar energy (USSE) in the desert southwest was anticipated to negatively affect birds through impacts to habitat and collision mortality. The discovery of stranded or dead birds that associate with aquatic habitat was unexpected as PV USSE facilities do not contain water and are not tall vertical hazards to migrating birds. Based on early patterns in fatality monitoring data, the lake effect hypothesis (LEH) was developed and suggested that birds misinterpret PV solar panels for water; however, no explicit tests of predictions of the LEH have been completed. We collected data from five PV USSE facilities and reference areas in three habitat types in southern California to determine if general predictions under the LEH were supported for aquatic habitat birds. We did not find that live aquatic habitat birds occurred more frequently at the PV solar sites than in the paired reference areas of similar habitat to the solar sites. Although the bird community (live and dead) contained aquatic habitat species, bird diversity was 3-4 times higher and standardized use was more than an order of magnitude higher at a small regional lake than measured at the PV USSE facilities. Finally, we did not observe aquatic habitat bird fatalities in the desert/scrub and grassland reference areas, but we did observe aquatic habitat bird fatalities in the agricultural reference area. Thus, the idea of ‘lake effect’ in which birds perceive a PV USSE facility as a waterbody and are broadly attracted is likely an overgeneralization of a nuanced process. We suggest there could be species-specific context dependent attraction to PV solar facilities in place of a generalized ‘lake effect’ and that potential causal mechanisms require further investigation.
“Bird and Bat Mortality at Wind Energy Projects in Latin American: A Summary and Review” – M. Sofia Agudelo, Ph.D., Presenter
Live Q&A Session: November 5th, 2:00-3:00 PM EST
Presentation Description: Most post-construction fatality monitoring (PCFM) studies to date have focused on North America and Europe, and this information has been used to assess the impacts of large-scale wind energy on birds and bats. A comprehensive review of wind-wildlife fatality information is still lacking for Latin America; however, given the current installed capacity and the projected increase of wind energy production across this region, it is important to fill in the knowledge gap on impacts to wildlife. To provide a current summary of known impacts to birds and bats in Latin America and to identify gaps on this information, we compiled, reviewed, and synthesized fatality information at wind energy projects in the region. Our literature search resulted in 10 references relevant to the scope of this review, six of which provided number of fatalities by species and the type of PCFM search being conducted, meeting our criteria for inclusion in fatality summaries. Passerines composed the majority of bird fatalities, with no Threatened bird species reported. Molossids composed the majority of bat fatalities, with one Threatened bat species reported. Our review of all studies and focused assessment of only those studies with fatality summaries indicated differences in the amount of information and level of detail related to bird and bat fatalities at wind energy projects in Latin America. Due to the taxon-specific nature of collision risk with wind turbines for birds and bats, it is difficult to make a general impact assessment of wind energy development on birds and bats in Latin America, especially given the limited information available. However, this summary can be used as a starting point to inform conservation efforts aiming at avoiding, minimizing, and mitigating impacts of wind energy development on birds and bats and future, standardized results would enhance our ability to do so.