Geology Reports⌕ Search

Geology topics

Bradley A. Pickens

Publications and source records attributed to Bradley A. Pickens.

9 recordsLinked to original sources
USGS2023

Challenges and solutions for automated avian recognition in aerial imagery

Remote aerial sensing provides a non-invasive, large geographical-scale technology for avian monitoring, but the manual processing of images limits its development and applications. Artificial Intelligence (AI) methods can be used to mitigate this manual image processing requirement. The implementation of AI methods, however, has several challenges: (1) imbalanced (i.e., long-tailed) data distribution, (2) annotation uncertainty in categorization, and (3) dataset discrepancies across different study sites. Here we use aerial imagery data of waterbirds around Cape Cod and Lake Michigan in the United States to examine how these challenges limit avian recognition performance. We review existing solutions and demonstrate as use cases how methods like Label Distribution Aware Marginal Loss with Deferred Re-Weighting, hierarchical classification, and FixMatch address the three challenges. We also present a new approach to tackle the annotation uncertainty challenge using a Soft-fine Pseudo-Label methodology. Finally, we aim with this paper to increase awareness in the ecological remote sensing community of these challenges and bridge the gap between ecological applications and state-of-the-art computer science, thereby opening new doors to future research.

Remote Sensing in Ecology and Conservation↗
USGS2022

Automating sandhill crane counts from nocturnal thermal aerial imagery using deep learning

Population monitoring is essential to management and conservation efforts for migratory birds, but traditional low-altitude aerial surveys with human observers are plagued by individual observer bias and risk to flight crews. Aerial surveys that use remote sensing can reduce bias and risk, but manual counting of wildlife in imagery is laborious and may be cost-prohibitive. Therefore, automated methods for counting are critical to cost-efficient application of remote sensing for wildlife surveys covering large areas. We conducted nocturnal surveys of sandhill cranes ( Antigone canadensis ) during spring migration in the Central Platte River Valley of Nebraska, USA, using midwave thermal infrared sensors. We developed a framework for automated counting of sandhill cranes from thermal imagery using deep learning, assessed and compared the performance of two automated counting models, and quantified the effect of spatial resolution on counting accuracy. Aerial thermal imagery data were collected in March 2018 and 2021; 40 images were analyzed. We applied two deep learning models: an object detection approach, Faster R-CNN and a recently developed pixel-density estimation approach, ASPDNet. Model performance was determined using data independent of the training imagery. The effect of spatial resolution was quantified with a beta regression on relative error. Our results showed model accuracy of 9% mean percent error for ASPDNet and 18% for Faster R-CNN. Most error was related to the undercounting of sandhill cranes. ASPDNet had < 50% of the error of Faster R-CNN as measured by mean percent error, root-mean-squared error and mean absolute error. Spatial resolution affected accuracy of both models, with error rate increasing with coarser resolution, particularly with Faster R-CNN. Deep learning models, particularly pixel-density estimators, can accurately automate counting of migratory birds in a dense, aggregate setting such as nocturnal roosting sites.

Nebraska↗
USGS2017

Indicator-driven conservation planning across terrestrial, freshwater aquatic, and marine ecosystems of the south Atlantic, USA

Systematic conservation planning, a widely used approach to identify priority lands and waters, uses efficient, defensible, and transparent methods aimed at conserving biodiversity and ecological systems. Limited financial resources and competing land uses can be major impediments to conservation; therefore, participation of diverse stakeholders in the planning process is advantageous to help address broad-scale threats and challenges of the 21st century. Although a broad extent is needed to identify core areas and corridors for fish and wildlife populations, a fine-scale resolution is needed to manage for multiple, interconnected ecosystems. Here, we developed a conservation plan using a systematic approach to promote landscape-level conservation within the extent of the South Atlantic Landscape Conservation Cooperative. Our objective was to identify the highest-ranked 30% of lands and waters within the South Atlantic deemed necessary to conserve ecological and cultural integrity for the 10 primary ecosystems of the southeastern United States. These environments varied from terrestrial, freshwater aquatic, and marine. The planning process was driven by indicators of ecosystem integrity at a 4-ha resolution. We used the program Zonation and 28 indicators to optimize the identification of lands and waters to meet the stated objective. A novel part of our study was the prioritization of multiple ecosystems, and we discuss the advantages and disadvantages of this approach. The evaluation of indicator representation within prioritizations was a useful method to show where improvements could be made; some indicators dictated hotspots, some had a limited extent and were well represented, and others had a limited effect. Overall, we demonstrate that a broad-scale (408,276 km 2 of terrestrial and 411,239 km 2 of marine environments) conservation plan can be realized at a fine-scale resolution, which will allow implementation of the regional plan at a local level relevant to decision making.

Journal of Fish and Wildlife Management↗
USGS2017

The effect of urban growth on landscape-scale restoration for a fire-dependent songbird

A landscape-scale perspective on restoration ecology has been advocated, but few studies have informed restoration with landscape metrics or addressed broad-scale threats. Threats such as urban growth may affect restoration effectiveness in a landscape context. Here, we studied longleaf pine savanna in the rapidly urbanizing southeastern United States where a habitat-specialist bird, Bachman's sparrow ( Peucaea aestivalis ), is closely associated with savanna vegetation structure and frequent fire. Our objectives were to construct a species distribution model for Bachman's sparrow, determine the relationship between fire and urbanization, quantify the urban growth effect (2010–2090), identify potential restoration areas, and determine the interaction between restoration potential and urban growth by 2050. Number of patches, patch size, and isolation metrics were used to evaluate scenarios. The species distribution model was 88% accurate and emphasized multiscale canopy cover characteristics, fire, and percent habitat. Fires were less common <600 m from urban areas, and this fire suppression effect exacerbated urban growth effects. For restoration scenarios, canopy cover reduction by 30% resulted in nearly double the amount of habitat compared to the prescribed fire scenario; canopy cover reduction resulted in larger patch sizes and less patch isolation compared to current conditions. The effect of urban growth on restoration scenarios was unequal. Seventy-four percent of restoration areas from the prescribed fire scenario overlapped with projected urban growth, whereas the canopy cover reduction scenario only overlapped by 9%. We emphasize the benefits of simultaneously considering the effects of urban growth and landscape-scale restoration potential to promote a landscape with greater patch sizes and less isolation.

North Carolina↗
USGS2017

Seasonal movements and multiscale habitat selection of Whooping Crane (Grus americana) in natural and agricultural wetlands

Eleven of 15 species of cranes (family: Gruidae) are considered vulnerable or endangered, and the increase of agriculture and aquaculture at the expense of natural wetlands and grasslands is a threat to Gruidae worldwide. A reintroduced population of Whooping Crane ( Grus americana ) was studied in coastal and agricultural wetlands of Louisiana and Texas, USA. The objectives were to compare Whooping Crane movements across seasons, quantify multiscale habitat selection, and identify seasonal shifts in selection. Whooping Cranes ( n = 53) were tracked with satellite transmitters to estimate seasonal core-use areas (50% home range contours) via Brownian bridge movement models and assess habitat selection. Whooping Crane core-use areas ( n = 283) ranged from 4.7 to 438.0 km 2 , and habitat selection changed seasonally as shallow water availability varied. Whooping Crane core-use areas were composed of more fresh marsh in spring/summer, but shifted towards rice and crawfish ( Procambarus spp.) aquaculture in the fall/winter. Within core-use areas, aquaculture was most strongly selected, particularly in fall when fresh marsh became unsuitable. Overall, the shifting of Whooping Crane habitat selection over seasons is likely to require large, heterogeneous areas. Whooping Crane use of agricultural and natural wetlands may depend on spatio-temporal dynamics of water depth.

Louisiana↗
USGS2014

Multiscale habitat selection of wetland birds in the northern Gulf Coast

The spatial scale of habitat selection has become a prominent concept in ecology, but has received less attention in coastal ecology. In coastal marshes, broad-scale marsh types are defined by vegetation composition over thousands of hectares, water-level management is applied over hundreds of hectares, and fine-scale habitat is depicted by tens of meters. Individually, these scales are known to affect wetland fauna, but studies have not examined all three spatial scales simultaneously. We investigated wetland bird habitat selection at the three scales and compared single- and multiscale models. From 2009 to 2011, we surveyed marsh birds (i.e., Rallidae, bitterns, grebes), shorebirds, and wading birds in fresh and intermediate (oligohaline) coastal marsh in Louisiana and Texas, USA. Within each year, six repeated surveys of wintering, resident, and migratory breeding birds were conducted at > 100 points ( n = 304). The results revealed fine-scale factors, primarily water depth, were consistently better predictors than marsh type or management. However, 10 of 11 species had improved models with the three scales combined. Birds with a linear association with water depth were, correspondingly, most abundant with deeper fresh marsh and permanently impounded water. Conversely, intermediate marsh had a greater abundance of shallow water species, such as king rail Rallus elegans, least bittern Ixobrychus exilis, and sora Porzana carolina. These birds had quadratic relationships with water depth or no relationship. Overall, coastal birds were influenced by multiple scales corresponding with hydrological characteristics. The effects suggest the timing of drawdowns and interannual variability in spring water levels can greatly affect wetland bird abundance.

Estuaries and Coasts↗
USGS2014

Linking multi-temporal satellite imagery to coastal wetland dynamics and bird distribution

Ecosystems are characterized by dynamic ecological processes, such as flooding and fires, but spatial models are often limited to a single measurement in time. The characterization of direct, fine-scale processes affecting animals is potentially valuable for management applications, but these are difficult to quantify over broad extents. Direct predictors are also expected to improve transferability of models beyond the area of study. Here, we investigated the ability of non-static and multi-temporal habitat characteristics to predict marsh bird distributions, while testing model generality and transferability between two coastal habitats. Distribution models were developed for king rail ( Rallus elegans ), common gallinule ( Gallinula galeata ), least bittern ( Ixobrychus exilis ), and purple gallinule ( Porphyrio martinica ) in fresh and intermediate marsh types in the northern Gulf Coast of Louisiana and Texas, USA. For model development, repeated point count surveys of marsh birds were conducted from 2009 to 2011. Landsat satellite imagery was used to quantify both annual conditions and cumulative, multi-temporal habitat characteristics. We used multivariate adaptive regression splines to quantify bird-habitat relationships for fresh, intermediate, and combined marsh habitats. Multi-temporal habitat characteristics ranked as more important than single-date characteristics, as temporary water was most influential in six of eight models. Predictive power was greater for marsh type-specific models compared to general models and model transferability was poor. Birds in fresh marsh selected for annual habitat characterizations, while birds in intermediate marsh selected for cumulative wetness and heterogeneity. Our findings emphasize that dynamic ecological processes can affect species distribution and species-habitat relationships may differ with dominant landscape characteristics.

Ecological Modelling↗
USGS2013

Microhabitat selection, demography, and correlates of home range size for the King Rail ( Rallus elegans )

Animal movements and habitat selection within the home range, or microhabitat selection, can provide insights into habitat requirements, such as foraging and area requirements. The King Rail ( Rallus elegans ) is a wetland bird of high conservation concern in the United States, but little is known about its movements, habitats, or demography. King Rails ( n = 34) were captured during the 2010&ndash;2011 breeding seasons in the coastal marshes of southwest Louisiana and southeast Texas. Radio telemetry and direct habitat surveys of King Rail locations were conducted to estimate home ranges and microhabitat selection. Within home ranges, King Rails selected for greater plant species richness and comparatively greater coverage of Phragmites australis , Typha spp., and Schoenoplectus robustus . King Rails were found closer to open water compared to random locations placed 50 m from King Rail locations. Home ranges ( n = 22) varied from 0.8&ndash;32.8 ha and differed greatly among sites. Home range size did not vary by year or sex; however, increased open water, with a maximum of 29% observed in the study, was correlated with smaller home ranges. Breeding season cumulative survivorship was 89% &plusmn; 22% in 2010 and 61% &plusmn; 43% in 2011, which coincided with a drought. With an equal search effort, King Rail chicks and juveniles observed in May-June decreased from 110 in 2010 to only 16 in the drier year of 2011. The findings show King Rail used marsh with &le; 29% open water and had smaller home ranges when open water was more abundant.

Louisiana, Texas↗