Search

There is growing interest in using deep learning models to automate wildlife detection in aerial imaging surveys to increase efficiency, but human-generated annotations remain necessary for model training. However, even skilled observers may diverge in interpreting aerial imagery of complex environments, which may result in downstream instability of models. In this study, we present a framework for assessing annotation reliability by calculating agreement metrics for individual observers against an aggregated set of annotations generated by clustering multiple observers' observations and selecting the mode classification. We also examined how image attributes like spatial resolution and texture influence observer agreement. To demonstrate the framework, we analyzed expert and volunteer annotations of twelve drone images of migratory waterfowl in New Mexico. Neither group reliably identified duck species: experts showed low agreement (43–56%) for several common species, and volunteers opted out of the task. When simplified into broad morphological categories, there was high agreement for cranes (99% among experts, 95% among volunteers) and ducks (93% among experts, 92% among volunteers), though agreement among volunteers was notably lower for classifying geese (75%) than among experts (94%). The aggregated annotation sets from the two groups were similar: the volunteer count of birds across all images was 91% of the expert count, with no statistically significant difference per image (t = 1.27, df = 338, p = 0.20). Bird locations matched 81% between groups and classifications matched 99.4%. Tiling images to reduce search area and maintaining a constant scale to keep size differences between classes consistent may increase observer agreement. Although our sample was limited, these findings indicate potential taxonomic limitations to aerial wildlife surveys and show that, in aggregate, volunteers can produce data comparable to experts'. This framework may assist other wildlife practitioners in evaluating the reliability of their input data for deep learning models. [ABSTRACT FROM AUTHOR]

Northern bobwhite (Colinus virginianus) populations have declined across much of their range. In response to these declines, wildlife biologists and managers have increased survey efforts and tried to optimize detection and capitalize on technological advances to improve population estimates and cost‐effectiveness. Our objective was to determine how environmental conditions influence detection of the reproduction call, or whistle, of masked bobwhite (C. v. ridgwayi), an endangered subspecies of northern bobwhite, using autonomous recording units (ARUs). We estimated the call intensity of the masked bobwhite reproduction call as 112 ± 0.5 decibels (mean ± SE) at 10 cm. We then broadcasted 16,284 calls during 17 trials to compare manual and automated call detection in recordings collected with ARUs. We used these data to model detectability of a bobwhite reproduction call, for when the bird is present and available, as a function of distance and weather conditions using generalized linear mixed models with trial as a random effect. Regardless of detection type, one model structure was competitive and suggested detection probability was a function of distance, wind speed, and wind direction. Detectability decreased with increased distance and wind speed and was influenced by wind direction. We demonstrate the use of our results to predict the probability of detecting a reproduction call during ARU‐based monitoring efforts. By understanding the effects of environmental factors on the detection of a bobwhite reproductive call, bobwhite surveys can be improved. [ABSTRACT FROM AUTHOR]

Results In the original article [[1]], there was a coding error that resulted in an error in the values reported in the first and second paragraph of Results. These errors were the result of the same coding error identified for the Results Section 3 and corrected values are increased by a factor of two. Graph: Figure 16 Fine-fuels and fuel-type data compared within principal fire management units (n = 59) on the Buenos Aires National Wildlife Refuge (BANWR) showed a strong positive relationship between non-native grass cover and average fine-fuel biomass. [Extracted from the article]

The spread of flammable invasive grasses, woody plant encroachment, and enhanced aridity have interacted in many grasslands globally to increase wildfire activity and risk to valued assets. Annual variation in the abundance and distribution of fine-fuel present challenges to land managers implementing prescribed burns and mitigating wildfire, although methods to produce high-resolution fuel estimates are still under development. To further understand how prescribed fire and wildfire influence fine-fuels in a semi-arid grassland invaded by non-native perennial grasses, we combined high-resolution Sentinel-2A imagery with in situ vegetation data and machine learning to estimate yearly fine-fuel loads from 2015 to 2020. The resulting model of fine-fuel corresponded to field-based validation measurements taken in the first (R² = 0.52, RMSE = 218 kg/ha) and last year (R² = 0.63, RMSE = 196 kg/ha) of this 6-year study. Serial prediction of the fine-fuel model allowed for an assessment of the effect of prescribed fire (average reduction of -80 kg/ha 1-year post fire) and wildfire (-260 kg/ha 1-year post fire) on fuel conditions. Post-fire fine-fuel loads were significantly lower than in unburned control areas sampled just outside fire perimeters from 2015 to 2020 across all fires (t = 1.67, p < 0.0001); however, fine-fuel recovery occurred within 3--5 years, depending upon burn and climate conditions. When coupled with detailed fuels data from field measurements, Sentinel-2A imagery provided a means for evaluating grassland fine-fuels at yearly time steps and shows high potential for extended monitoring of dryland fuels. Our approach provides land managers with a systematic analysis of the effects of fire management treatments on fine-fuel conditions and provides an accurate, updateable, and expandable solution for mapping fine-fuels over yearly time steps across drylands throughout the world. [ABSTRACT FROM AUTHOR]

AbstractGeospatial analyses of human-environment interactions are challenged by the multi-scale, multi-dimensional nature of human-environment systems. Research in such contexts must often rely on integrating multiple, independently produced data sources, which presents heterogenous data qualities and interoperability challenges. Understanding data quality and transparency becomes increasingly important in these contexts, and multi‐granularity and context specific spatial data quality indicators are needed. We develop a data pedigree system that accounts for multiple data quality aspects, geospatial ambiguities that may hinder interoperability, and the fitness-for-use of each data source for indicating causal linkages between human activities and environmental change. We demonstrate its application to a particularly challenging and data sparse case study of identifying the location and timing of transnational cocaine trafficking, or ‘narco-trafficking’, in Central America with five spatial and temporal data quality indicators: geographic clarity, geographic interpretation, provenance, temporal specificity, and narco-trafficking certainty. The proposed data pedigree system provides a systematic and coherent analytical framework for interoperability, comparison, and corroboration of fragmented and incomplete data, which are needed to support advanced geospatial analyses, such as causal inference techniques. The study demonstrates the transferability and operationalization of the data pedigree system for examining complex human-environment interactions, especially those influenced by illicit economies. [ABSTRACT FROM AUTHOR]

Fire has historically played an important role in shaping the structure and composition of Sonoran semi-desert grassland vegetation. Yet, human use and land management activities have significantly altered arid grassland ecosystems over the last century, often producing novel fuel conditions. The variety of continuously updated satellite remote sensing systems provide opportunities for efficiently mapping combustible fine-fuels and fuel-types (e.g., grass, shrub, or tree cover) over large landscapes that are helpful for evaluating fire hazard and risk. For this study, we compared field ceptometer leaf area index (LAI) measurements to conventional means for estimating fine-fuel biomass on 20, 50 m × 20 m plots and 431, 0.5 m × 0.5 m quadrats on the Buenos Aires National Wildlife Refuge (BANWR) in southern Arizona. LAI explained 65% of the variance in fine-fuel biomass using simple linear regression. An additional 19% of variance was explained from Random Forest regression tree models that included herbaceous plant height and cover as predictors. Field biomass and vegetation measurements were used to map fine-fuel and vegetation cover (fuel-type) from plots on BANWR comparing outcomes from multi-date (peak green and dormant period) Worldview-3 (WV3) and Landsat Operational Land Imager (OLI) imagery. Fine-fuel biomass predicted from WV3 imagery combined with terrain information from a digital elevation model explained greater variance using regression tree models (65%) as compared to OLI models (58%). Vegetation indices developed using red-edge bands as well as modeled bare ground and herbaceous cover were important to improve WV3 biomass estimates. Land cover classification for 11 cover categories with high spatial resolution WV3 imagery showed 80% overall accuracy and highlighted areas dominated by non-native grasses with 87% user's class accuracy. Mixed native and non-native grass and shrublands showed 59% accuracy and less common areas dominated by native grasses on plots showed low class accuracy (23%). Digital data layers from WV3 models showed a significantly positive relationship (r2 = 0.68, F = 119.2, p < 0.001) between non-native grass cover (e.g., Eragrostis lehmanniana) and average fine-fuel biomass within refuge fire management units. Overall, both WV3 and OLI produced similar fine-fuel biomass estimates although WV3 showed better model performance and helped characterized fine-scale changes in fuel-type and continuity across the study area. [ABSTRACT FROM AUTHOR]