No abstract

The use of unmanned aerial vehicles (UAVs) to map and monitor the environment has increased sharply in the last few years. Many individuals and organizations have purchased consumer-grade UAVs, and commonly acquire aerial photographs to map land cover. The resulting ultra-high-resolution (subdecimeter-resolution) imagery has high information content, but automating the extraction of this information to create accurate, wall-to-wall land-cover maps is quite difficult. We introduce image-processing workflows that are based on open-source software and can be used to create land-cover maps from ultra-high-resolution aerial imagery. We compared four machine-learning workflows for classifying images. Two workflows were based on random forest algorithms. Of these, one used a pixel-by-pixel approach available in ilastik, and the other used image segments and was implemented with R and the Orfeo ToolBox. The other two workflows used fully connected neural networks and convolutional neural networks implemented with Nenetic. We applied the four workflows to aerial photographs acquired in the Great Basin (western USA) at flying heights of 10 m, 45 m and 90 m above ground level. Our focal cover type was cheatgrass (Bromus tectorum), a non-native invasive grass that changes regional fire dynamics. The most accurate workflow for classifying ultra-highresolution imagery depends on diverse factors that are influenced by image resolution and land-cover characteristics, such as contrast, landscape patterns and the spectral texture of the land-cover types being classified. For our application, the ilastik workflow yielded the highest overall accuracy (0.82-0.89) as assessed by pixel-based accuracy.

No abstract

Cover-based allometric estimate of aboveground biomass of a non-native, invasive annual grass (Bromus tectorum L.

Occurrence models are frequently used to infer characteristics of habitat and inform management plans for large areas, but the extent to which these models predict reproduction, or the environmental characteristics associated with reproduction, is uncertain. We examined whether occurrence models based on point‐count data, and vegetation attributes within 50 m of presences, predicted the locations where birds were nesting in the Great Basin, USA. This work has practical relevance given that occurrence data require less time and effort to collect than direct measures of reproduction. We hypothesized that occurrence models would better characterize where two nest stratum generalists, Brewer's Sparrow Spizella breweri and Green‐tailed Towhee Pipilo chlorurus, nest than where a nest stratum specialist, American Dusky Flycatcher Empidonax oberholseri, nests. First, we compared which environmental covariates were associated with occurrence, via a model fitted to point‐count data, and nest locations, via a model fitted to data from confirmed nests. Second, we used data on the vegetation within 50 m of known nests to evaluate whether predictions from our occurrence model accurately classified the presence of nests. Third, we examined whether vegetation within 5 m of a known nest differed from vegetation at randomly chosen locations of equal size within 20–80 m of the nest. Occurrence models classified nest locations with relatively high accuracy for Brewer's Sparrow and American Dusky Flycatcher, and with moderate accuracy for Green‐tailed Towhee. Vegetation within 5 m of nests differed from that of random locations within 20–80 m, suggesting that even the nest stratum generalists based nest selection on fine‐grained attributes. Models that rely on point‐count data to examine species–habitat relationships assume that occupancy indicates breeding activity. Our results suggest that for some species, this is a valid assumption, although it is probably conditional on a well‐fitted occurrence model.