Grilled Cheese

ExploreLog inSign up
Terms of UsePrivacy PolicyCommunity StandardsHelpGet the app

Grilled Cheese is a product of Village Compute

Version devBuilt at: 2026-10-12 00:11:13 EDT

Explore

PostsPeople
LatestRanked
@ijwildlandfire.bsky.socialOct 10, 2026, 4:14 AM

🔥 New in IJWF:

Bhatta et al. developed a smartphone-based AI framework using YOLOv11, object tracking and depth estimation to detect and approximately map fuels around homes in 3D, supporting wildfire risk assessment and mitigation.

🔗 doi.org/10.1071/WF25...

#IJWildlandFire

Mapping of object detection results using the B-FiDSY-3D framework: (a) Open Street Map visualization; (b) Google Earth visualization; (c) closer view of detected objects in Google Earth.Application of the B-FiDSY-3D framework to a street at the Eleven13 apartments: (a) mapping object detection results in Google Earth; (b) 3D object detection model for enhanced visualization.Demonstrating home ignition zone attributes comprising building features and defensible space fuels.
@ijwildlandfire.bsky.socialOct 2, 2026, 2:22 PM

🔥 New in IJWF:
Ochoa et al. introduce ESFire30, a 30 m burned-area dataset mapping nearly four decades of wildfire across peninsular Spain. The dataset reveals strong spatial variability, frequent reburning in the northwest.
🔗 doi.org/10.1071/WF26...
#IJWildlandFire

Fig. 4. Pixels that burned multiple times between 1985 and 2023: (a) burned area (BA) in large fires; (b) burned area > 500 ha for the same period. wf26035Fig. 3. Annual burned areas classified by fire type (a), proportion of patches by size and number of fires; and (b) the total number of patches per year with frequency in logarithmic scale.
@ijwildlandfire.bsky.socialSep 21, 2026, 1:25 AM

🔥 New in IJWF:

Morrison et al. show bushfire evacuation is rarely a simple A-to-B journey. Interviews with 52 residents across three Australian fires reveal stops for children, pets, neighbours, belongings and information, shaped by social connections.

🔗 doi.org/10.1071/WF25...
#IJWildlandFire

Conceptual model explaining intermediate stop locations for evacuees during three Australian bushfires; the types of stop sites are shown in bold (surrounded by dotted-lined boxes).
@ijwildlandfire.bsky.socialSep 16, 2026, 1:08 AM

🔥 New in IJWF:

Van Dusen et al. show fuel treatments can reduce high-severity fire, especially when recent and combining thinning with prescribed fire, while extreme fire spread can overwhelm treatment effects in complex landscapes.

🔗 doi.org/10.1071/WF25...
#IJWildlandFire

A semi-automated, empirical workflow to evaluate the effects of fuel treatments across biophysical conditions. Our workflow is built with flexibly to support the evaluation of treatment effects across multiple measures effectiveness. In this study, we focus on one response variable: vegetation burn severity (low-moderate vs high severity). The workflow includes three broad phases: Data preparation, Fire-level Matching, Landscape-level Modeling. (1) Geospatial data preparation includes the selection of previous fires and historical treatments, gathering geospatial datasets, selecting a response variable (burn severity) and data extraction, (2) fire-level matching aims to pair treated and untreated points based on fire-specific drivers of the response variable, and (3) landscape-level models use model-based inference to evaluate treatment effectiveness across multiple biophysical conditions.The mean predicted probability of high severity among treatment classes across the Klamath-Trinity Landscape for the east and west random forest classification models. The partial dependence mean probability of high severity for each class was calculated by predicting each treatment class across all other predictors values then taking the mean response. The error bars for each treatment class represents the 95th confidence interval of the mean probability using 1000 bootstrapped datasets to fit our model.Partial dependence plots for treatment age the west (a) and east (b) study area of the Klamath-Trinity Landscape separated by each treatment type. The dashed, navy untreated line represents the mean probability of high severity for untreated points. Importance was determined by permutation variable importance from the landscape random forest classification model. Each solid line displays the mean relationship between each treatment type and environmental variable across all other environmental variable combinations. The shaded region represents the 95th percentile across variables by treatment type through 1000 bootstrapped models.