Accurate estimations of animal populations are necessary for management, conservation, and policy decisions. However, methods for surveying animal communities disproportionately represent specific groups or guilds. For example, transect surveys can provide robust data for large arboreal species but underestimate cryptic or small‐bodied terrestrial species, whereas camera traps have the inverse tendency. The integration of information from multiple methodologies would provide the most complete inference on population size or responses to putative covariates, yet a simple, robust framework that allows integration and comparison of multiple data sources has been lacking. We use 27,813 counts of 35 species or species groups derived from concurrent visual transects, dung transects, and camera trap surveys in tropical forests and compare them within a generalized joint attribute modeling framework (GJAM) that both compares and integrates field‐collected dung, visual, and camera trap data to quantify the species‐ and trait‐specific differences in detection for each method. The effectiveness of survey method was strongly dependent on species, as well as animal traits. These differences in effectiveness contributed to meaningful differences in the reported strength of a known important covariate for animal communities (distance to nearest village). Data fusion through GJAM allows clear and unambiguous comparisons of the counts provided from each different methodology, the incorporation of trait information, and fusion of all three data streams to generate a more complete estimate of the effects of an anthropogenic disturbance covariate. Research and conservation resources are extremely limited, which often means that field campaigns attempt to maximize the amount of information gathered especially in remote, hard‐to‐access areas. Advances in these understudied areas will be accelerated by analytical methods that can fully leverage the total body of diverse biodiversity field data, even when they are collected using different methods. We demonstrate that survey methods vary in their effectiveness for counting species based on biological traits, but more importantly that generative models like GJAM can integrate data from multiple sources in one cohesive statistical framework to make improved inference in understudied environments.
- Award ID(s):
- 1903972
- NSF-PAR ID:
- 10147565
- Date Published:
- Journal Name:
- IEEE Homeland Security Technologies (HST 2019)
- Page Range / eLocation ID:
- 1 to 5
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
Effective transportation performance measurement (TPM) benefits from ubiquitous transportation system monitoring both spatially and temporally. In the context of freight-oriented TPM, traditional devices such as inductive loops, cameras, manual counts, and so forth, may fail to provide comprehensive and high-resolution coverage, providing, for example, only volume counts for a small subset of links across a large network with no indication of trip linkages. New sources of big data from mobile sensors including on-board global positioning system (GPS) devices allow more comprehensive network coverage and insights into trip chaining behaviors. However, to gain actionable insights into system performance from large and noisy streams of mobile sensor data, it is necessary to mine it for relevant operational characteristics of the trucks it represents. Such characteristics include stop locations, stop duration, stop time of day, trip length, and trip duration. To address this methodological need, this paper presents three heuristic algorithms: “stop identification,”“path identification,” and “trip identification.” To address the issue of determining relevant operational characteristics, a multinomial logit (MNL) model approach is applied to determine the commodity carried based on the outputs of the heuristic algorithms. The MNL model is novel in that it relates operational characteristics to commodity carried thus filling a critical data gap that currently limits the development of advanced freight forecasting models. The set of models developed in this paper allow large-scale GPS data to be used for freight planning while maintaining levels of data anonymity that allow such data to be shared with public agencies.more » « less
-
more » « less
Holographic cloud probes provide unprecedented information on cloud particle density, size and position. Each laser shot captures particles within a large volume, where images can be computationally refocused to determine particle size and location. However, processing these holograms with standard methods or machine learning (ML) models requires considerable computational resources, time and occasional human intervention. ML models are trained on simulated holograms obtained from the physical model of the probe since real holograms have no absolute truth labels. Using another processing method to produce labels would be subject to errors that the ML model would subsequently inherit. Models perform well on real holograms only when image corruption is performed on the simulated images during training, thereby mimicking non-ideal conditions in the actual probe. Optimizing image corruption requires a cumbersome manual labeling effort. Here we demonstrate the application of the neural style translation approach to the simulated holograms. With a pre-trained convolutional neural network, the simulated holograms are “stylized” to resemble the real ones obtained from the probe, while at the same time preserving the simulated image “content” (e.g. the particle locations and sizes). With an ML model trained to predict particle locations and shapes on the stylized data sets, we observed comparable performance on both simulated and real holograms, obviating the need to perform manual labeling. The described approach is not specific to holograms and could be applied in other domains for capturing noise and imperfections in observational instruments to make simulated data more like real world observations.
-
more » « less
Synopsis Although gigantic body size and obligate filter feeding mechanisms have evolved in multiple vertebrate lineages (mammals and fishes), intermittent ram (lunge) filter feeding is unique to a specific family of baleen whales: rorquals. Lunge feeding is a high cost, high benefit feeding mechanism that requires the integration of unsteady locomotion (i.e., accelerations and maneuvers); the impact of scale on the biomechanics and energetics of this foraging mode continues to be the subject of intense study. The goal of our investigation was to use a combination of multi-sensor tags paired with UAS footage to determine the impact of morphometrics such as body size on kinematic lunging parameters such as fluking timing, maximum lunging speed, and deceleration during the engulfment period for a range of species from minke to blue whales. Our results show that, in the case of krill-feeding lunges and regardless of size, animals exhibit a skewed gradient between powered and fully unpowered engulfment, with fluking generally ending at the point of both the maximum lunging speed and mouth opening. In all cases, the small amounts of propulsive thrust generated by the tail were unable to overcome the high drag forces experienced during engulfment. Assuming this thrust to be minimal, we predicted the minimum speed of lunging across scale. To minimize the energetic cost of lunge feeding, hydrodynamic theory predicts slower lunge feeding speeds regardless of body size, with a lower boundary set by the ability of the prey to avoid capture. We used empirical data to test this theory and instead found that maximum foraging speeds remain constant and high (∼4 m s–1) across body size, even as higher speeds result in lower foraging efficiency. Regardless, we found an increasing relationship between body size and this foraging efficiency, estimated as the ratio of energetic gain from prey to energetic cost. This trend held across timescales ranging from a single lunge to a single day and suggests that larger whales are capturing more prey—and more energy—at a lower cost.
-
more » « less
Abstract This study uses cloud and radiative properties collected from in situ and remote sensing instruments during two coordinated campaigns over the Southern Ocean between Tasmania and Antarctica in January–February 2018 to evaluate the simulations of clouds and precipitation in nudged‐meteorology simulations with the CAM6 and AM4 global climate models sampled at the times and locations of the observations. Fifteen SOCRATES research flights sampled cloud water content, cloud droplet number concentration, and particle size distributions in mixed‐phase boundary layer clouds at temperatures down to −25°C. The 6‐week CAPRICORN2 research cruise encountered all cloud regimes across the region. Data from vertically pointing 94 GHz radars deployed was compared with radar simulator output from both models. Satellite data were compared with simulated top‐of‐atmosphere (TOA) radiative fluxes. Both models simulate observed cloud properties fairly well within the variability of observations. Cloud base and top in both models are generally biased low. CAM6 overestimates cloud occurrence and optical thickness while cloud droplet number concentrations are biased low, leading to excessive TOA reflected shortwave radiation. In general, low clouds in CAM6 precipitate at the same frequency but are more homogeneous compared to observations. Deep clouds are better simulated but produce snow too frequently. AM4 underestimates cloud occurrence but overestimates cloud optical thickness even more than CAM6, causing excessive outgoing longwave radiation fluxes but comparable reflected shortwave radiation. AM4 cloud droplet number concentrations match observations better than CAM6. Precipitating low and deep clouds in AM4 have too little snow. Further investigation of these microphysical biases is needed for both models.
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/HST47167.2019.9032986
