Search for: All records

Multiple landscape factors, including slope, vegetation density, and surface roughness, work together to affect pedestrian travel rates in off-path environments. Most previous pedestrian travel rate models have only quantified the effects of these factors at relatively coarse spatiotemporal resolutions, ignoring the fine-scaled information contained within LiDAR data and instantaneous travel trajectories. Previous studies have also relied on parametric function fitting to model off-path travel rates. This paper presents the first known examination of instantaneous travel rates using random forests—a machine learning algorithm well-suited to the presence of non-linear relationships and the ability to robustly evaluate variable importance. Principal components of airborne LiDAR-derived slope, vegetation density, and surface roughness were used as predictor variables in a random forest model. The model explained over 77% of the variance in observed travel rates from an independent test dataset (R² = 0.778). An analysis of permutation importance indicated that LiDAR-derived slope was the most important predictor of travel rate, followed by vegetation density, then surface roughness. Examining the importance of slope, vegetation density, and surface roughness together provides new insight beyond their individual effects, revealing how these landscape factors interact to shape pedestrian travel rates. While slope is the most important single predictor, our results show that vegetation density and surface roughness exert distinct and nonlinear influences that become clearer only when evaluated jointly. Notably, vegetation density reduces travel rate more sharply than surface roughness, and high vegetation density remains strongly limiting even when roughness is low, whereas the reverse is not true. These combined effects also exceed the explanatory power of individual median speed, underscoring the importance of environmental conditions over individual-level differences in exertion or fitness. Potential applications of this work include modeling instantaneous travel rates for wildland firefighter safety, search and rescue, and migration applications. 

Accurate mapping of understory vegetation is critical for applications ranging from wildlife habitat and biodiversity monitoring to fire behavior modeling and ecosystem carbon accounting. Lidar offers unique potential for quantifying subcanopy vegetation structure, yet the relative performance of airborne laser scanning (ALS) and uncrewed aerial system (UAS) laser scanning (ULS) for this task remains poorly quantified across different forest types. We compared ALS and ULS estimates of low-lying vegetation density (0.5–2.0 m aboveground), using high-resolution mobile laser scanning (MLS) as reference data, across 26 field plots spanning seven common forest and woodland types in northern Utah, USA. We examined the effects of platform, spatial scale, overstory structure, and vegetation type on model performance, and tested whether predictive accuracy could be improved by incorporating vegetation type and overstory metrics into linear mixed-effects models. Our results show that ALS consistently outperformed ULS despite its lower point density and older acquisition dates, achieving stronger correspondence with MLS-based density across most scales and forest types. ULS was advantageous at very fine spatial resolutions (<2.5 m subplot radii). Across platforms, predictive accuracy was maximized at 4.5 m subplot radii, balancing spatial detail with sufficient point sampling. Vegetation type exerted a strong influence, as those with greater spatial variation in observed density yielded higher predictive performance. Incorporating vegetation type as a random effect and adding overstory density as predictors substantially improved predictive performance, especially for ALS. These findings highlight ALS as a robust and transferable platform for broad-scale mapping of low-lying vegetation density, while ULS provides complementary value for fine-scale, site-specific, and temporally flexible applications. Our results underscore the importance of accounting for overstory conditions and vegetation type when modeling understory structure, and demonstrate the benefits of mixed-effects modeling frameworks. Future research should extend these analyses to additional ecosystems, disturbance contexts, and temporal scales, and explore machine learning approaches that more fully integrate structural and ecological predictors for operational understory mapping. 

Accurate prediction of walking travel rates is central to wide-ranging applications, including modeling historical travel networks, simulating evacuation from hazards, evaluating military ground troop movements, and assessing risk to wildland firefighters. Most of the existing functions for estimating travel rates have focused on slope as the sole landscape impediment, while some have gone a step further in applying a limited set of multiplicative factors to account for broadly defined surface types (e.g., “on-path” vs. “off-path”). In this study, we introduce the Simulating Travel Rates In Diverse Environments (STRIDE) model, which accurately predicts travel rates using a suite of airborne lidar-derived metrics (slope, vegetation density, and surface roughness) that encompass a continuous spectrum of landscape structure. STRIDE enables the accurate prediction of both on- and off-path travel rates using a single function that can be applied across wide-ranging environmental settings. The model explained more than 80% of the variance in the mean travel rates from three separate field experiments, with an average predictive error less than 16%. We demonstrate the use of STRIDE to map least-cost paths, highlighting its propensity for selecting logically consistent routes and producing more accurate yet considerably greater total travel time estimates than a slope-only model. 

Background Situational awareness is an essential component of wildland firefighter safety. In the US, crew lookouts provide situational awareness by proxy from ground-level locations with visibility of both fire and crew members. Aims To use machine learning to predict potential lookout locations based on incident data, mapped visibility, topography, vegetation, and roads. Methods Lidar-derived topographic and fuel structural variables were used to generate maps of visibility across 30 study areas that possessed lookout location data. Visibility at multiple viewing distances, distance to roads, topographic position index, canopy height, and canopy cover served as predictors in presence-only maximum entropy modelling to predict lookout suitability based on 66 known lookout locations from recent fires. Key results and conclusions The model yielded a receiver-operating characteristic area under the curve of 0.929 with 67% of lookouts correctly identified by the model using a 0.5 probability threshold. Spatially explicit model prediction resulted in a map of the probability a location would be suitable for a lookout; when combined with a map of dominant view direction these tools could provide meaningful support to fire crews. Implications This approach could be applied to produce maps summarising potential lookout suitability and dominant view direction across wildland environments for use in pre-fire planning. 

Wildland firefighters must be able to maintain situational awareness to ensure their safety. Crew members, including lookouts and crew building handlines, rely on visibility to assess risk and communicate changing conditions. Geographic information systems and remote sensing offer potential solutions for characterizing visibility using models incorporating terrain and vegetation height. Visibility can be assessed using viewshed algorithms, and while previous research has demonstrated the utility of these algorithms across multiple fields, their use in wildland firefighter safety has yet to be explored. The goals of this study were to develop an approach for assessing visibility at the handline level, quantify the effects of spatial resolution on a lidar-driven visibility analysis, and demonstrate a set of spatial metrics that can be used to inform handline safety. Comparisons were made between elevation models derived from airborne lidar at varying spatial resolutions and those derived from LANDFIRE, a US-wide 30 m product. Coarser resolution inputs overestimated visibility by as much as 223%, while the finest-scale resolution input was not practical due to extreme processing times. Canopy cover and slope had strong linear relationships with visibility, with R2 values of 0.806 and 0.718, respectively. Visibility analyses, when conducted at an appropriate spatial resolution, can provide useful information to inform situational awareness in a wildland fire context. Evaluating situational awareness at the handline level prior to engaging a fire may help firefighters evaluate potential safety risks and more effectively plan handlines.