Search for: All records

Accurate wildfire prediction in diverse and geographically dispersed areas is crucial for effective wildfire management. However, the limited availability of labeled data in data-challenged regions, along with the unique characteristics of these areas, poses challenges for training robust prediction models. This study investigates the performance of a convolutional neural network (CNN) on datasets comprising Landsat images from Canada and Alaska. Through principal component analysis (PCA), the study uncovers distinct differences in data distribution between the two regions. It is observed that the reduced data size of the Alaskan dataset, along with its distinct data distribution, leads to a decrease in the CNN's accuracy to 75% compared to an impressive 98% achieved on the Canadian dataset. To address this limitation, we propose a teacher-student model approach, transferring knowledge from a CNN trained on the larger Canadian dataset. The results demonstrate a significant accuracy improvement to 88.96% on the Alaskan dataset. Our findings highlight the effectiveness of the teacherstudent model in mitigating data scarcity challenges, enhancing wildfire prediction capabilities in regions with limited training data. This research contributes to improved wildfire monitoring and prevention strategies in challenging geographical locations. 

Wildfires, which are a natural part of the boreal ecosystem in Alaska, have recently increased in frequency and size. Environmental conditions (high temperature, low precipitation, and frequent lightning events) are becoming favorable for severe fire events. Fire releases greenhouse gasses such as carbon dioxide into the environment, creating a positive feedback loop for warming. Needleleaf species are the dominant vegetation in boreal Alaska and are highly flammable. They burn much faster due to the presence of resin, and their low-lying canopy structure facilitates the spread of fire from the ground to the canopy. Knowing the needleleaf vegetation distribution is crucial for better forest and wildfire management practices. Our study focuses on needleleaf fraction mapping using a well-documented spectral unmixing approach: multiple endmember spectral mixture analysis (MESMA). We used an AVIRIS-NG image (5 m), upscaled it to 10 m and 30 m spatial resolutions, and applied MESMA to all three images to assess the impact of spatial resolution on sub-pixel needleleaf fraction estimates. We tested a novel method to validate the fraction maps using field data and a high-resolution classified hyperspectral image. Our validation method produced needleleaf cover fraction estimates with accuracies of 73%, 79%, and 78% for 5 m, 10 m, and 30 m image data, respectively. To determine whether these accuracies varied significantly across different spatial scales, we used the McNemar statistical test and found no significant differences between the accuracies. The findings of this study enhance the toolset available to fire managers to manage wildfire and for understanding changes in forest demography in the boreal region of Alaska across the high-to-moderate resolution scale. 

Abstract Some of the largest climatic changes in the Arctic have been observed in Alaska and the surrounding marginal seas. Near-surface air temperature (T2m), precipitation ( P ), snowfall, and sea ice changes have been previously documented, often in disparate studies. Here, we provide an updated, long-term trend analysis (1957–2021; n = 65 years) of such parameters in ERA5, NOAA U.S. Climate Gridded Dataset (NClimGrid), NOAA National Centers for Environmental Information (NCEI) Alaska climate division, and composite sea ice products preceding the upcoming Fifth National Climate Assessment (NCA5) and other near-future climate reports. In the past half century, annual T2m has broadly increased across Alaska, and during winter, spring, and autumn on the North Slope and North Panhandle (T2m > 0.50°C decade −1 ). Precipitation has also increased across climate divisions and appears strongly interrelated with temperature–sea ice feedbacks on the North Slope, specifically with increased (decreased) open water (sea ice extent). Snowfall equivalent (SFE) has decreased in autumn and spring, perhaps aligned with a regime transition of snow to rain, while winter SFE has broadly increased across the state. Sea ice decline and melt-season lengthening also have a pronounced signal around Alaska, with the largest trends in these parameters found in the Beaufort Sea. Alaska’s climatic changes are also placed in context against regional and contiguous U.S. air temperature trends and show ∼50% greater warming in Alaska relative to the lower-48 states. Alaska T2m increases also exceed those of any contiguous U.S. subregion, positioning Alaska at the forefront of U.S. climate warming. Significance Statement This study produces an updated, long-term trend analysis (1957–2021) of key Alaska climate parameters, including air temperature, precipitation (including snowfall equivalent), and sea ice, to inform upcoming climate assessment reports, including the Fifth National Climate Assessment (NCA5) scheduled for publication in 2023. Key findings include widespread annual and seasonal warming with increased precipitation across much of the state. Winter snowfall has broadly increased, but spring and autumn snowfalls have decreased as rainfall increased. Autumn warming and precipitation increases over the North Slope, in particular, appear related to decreased sea ice coverage in the Beaufort Sea and Chukchi Seas. These trends may result from interrelated processes that accelerate Alaska climate changes relative to those of the contiguous United States. 

Alaska has witnessed a significant increase in wildfire events in recent decades that have been linked to drier and warmer summers. Forest fuel maps play a vital role in wildfire management and risk assessment. Freely available multispectral datasets are widely used for land use and land cover mapping, but they have limited utility for fuel mapping due to their coarse spectral resolution. Hyperspectral datasets have a high spectral resolution, ideal for detailed fuel mapping, but they are limited and expensive to acquire. This study simulates hyperspectral data from Sentinel-2 multispectral data using the spectral response function of the Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) sensor, and normalized ground spectra of gravel, birch, and spruce. We used the Uniform Pattern Decomposition Method (UPDM) for spectral unmixing, which is a sensor-independent method, where each pixel is expressed as the linear sum of standard reference spectra. The simulated hyperspectral data have spectral characteristics of AVIRIS-NG and the reflectance properties of Sentinel-2 data. We validated the simulated spectra by visually and statistically comparing it with real AVIRIS-NG data. We observed a high correlation between the spectra of tree classes collected from AVIRIS-NG and simulated hyperspectral data. Upon performing species level classification, we achieved a classification accuracy of 89% for the simulated hyperspectral data, which is better than the accuracy of Sentinel-2 data (77.8%). We generated a fuel map from the simulated hyperspectral image using the Random Forest classifier. Our study demonstrated that low-cost and high-quality hyperspectral data can be generated from Sentinel-2 data using UPDM for improved land cover and vegetation mapping in the boreal forest. 

The late-season extreme fire activity in Southcentral Alaska during 2019 was highly unusual and consequential. Firefighting operations had to be extended by a month in 2019 due to the extreme conditions of hot summer temperature and prolonged drought. The ongoing fires created poor air quality in the region containing most of Alaska’s population, leading to substantial impacts to public health. Suppression costs totaled over $70 million for Southcentral Alaska. This study’s main goals are to place the 2019 season into historical context, provide an attribution analysis, and assess future changes in wildfire risk in the region. The primary tools are meteorological observations and climate model simulations from the NCAR CESM Large Ensemble (LENS). The 2019 fire season in Southcentral Alaska included the hottest and driest June–August season over the 1979–2019 period. The LENS simulation analysis suggests that the anthropogenic signal of increased fire risk had not yet emerged in 2019 because of the CESM’s internal variability, but that the anthropogenic signal will emerge by the 2040–2080 period. The effect of warming temperatures dominates the effect of enhanced precipitation in the trend towards increased fire risk.