PARADISE ON FIRE

Annually, about 0.5% of Hawaii's land is affected by wildfires, with human activities being the cause of more than 98% of these fires.

Over the past ten years, on average, wildfires have ravaged around 8,000 hectares each year, with close to 1,000 incidents reported annually, as noted by Rezaie and colleagues in 2023.

Click below to explore Hawaii Hazard Map examined by the Hawaii Wildfire Management Organization (HWMO)

The wildfires that struck Hawaii in 2023, especially on the island of Maui, marked a catastrophic occurrence. Starting on August 8, the flames severely impacted the venerable resort community of Lahaina located on Maui's west coast. The disaster resulted in the loss of 98 lives and the decimation of close to 3,000 buildings.

The fire near Lahaina began as a small brush fire but rapidly intensified due to strong winds, spreading quickly through the dry landscape and wooden buildings of the town.

High winds also hampered emergency communications by knocking down telephone and power poles, further complicating evacuation efforts and firefighting due to decreased water pressure caused by the destruction of water pipes.

Damage Assessment Data Dashboard below.

This StoryMap offers a preliminary analysis aimed at understanding the potential impacts of the 2023 wildfires in Hawaii. Utilizing fire incident data, our goal is to analyze the repercussions of these events and propose relevant recommendations for future wildfire management and mitigation strategies. Through this analysis, we seek to inform stakeholders and aid in the development of more resilient communities against the backdrop of changing climatic conditions.

The island of Maui is the second-largest island of the state of Hawaii at 727.2 square miles.

Most of the land  (51.9% of the total land area) is assigned to agricultural activities including the raising of livestock, cultivation of crops, wind energy generation, timber cultivation, and aquaculture. Conservation districts consist primarily of upper elevation lands in the existing forest and water reserve zones, covering 43.3% of the total land. Rural and urban districts account for 1.1% and 3.8% of the study area, respectively. Rural areas include small farms intermixed with low-density residential lots with a minimum size of one-half acre. The urban district includes areas with concentrated structures, people, services, and vacant areas for future development (Hawai‘i Emergency Management Agency, 2018).

This project incorporates three distinct datasets to examine the dynamics of combustion areas, particularly on Maui Island, using a multi-tiered approach. Each dataset serves a strategic role in the methodology, enabling a detailed analysis from a broad geographic scale down to the micro-level impacts of fire events.

•  FIRM Data for Initial Detection: The first dataset comprises the Fire Information for Resource Management System (FIRM) data, which spans the entirety of the target region and includes thermal anomalies indicative of active fires. This dataset, derived from satellite observations, is pivotal in recognizing extensive fire events and is characterized by a resolution of approximately 1 kilometer, apt for identifying broad areas affected by fire but requiring further refinement for granular scrutiny.

• Sentinel-2 for Temporal and Spatial Refinement: The second dataset is procured from the Sentinel-2 mission, offering multispectral imagery that encompasses the entire area of interest. This dataset is instrumental in narrowing down the temporal window of the fire events and enhancing the spatial resolution of the affected areas. Although the Sentinel-2 data has a moderate resolution of 20 meters and a revisit cycle of five days, it provides vital insights into the scale and the progression of the combustion areas.

•  Planet Data for Detailed Analysis: The third dataset is sourced from Planet Labs, which delivers high-resolution imagery with a spatial resolution of 3 meters, capable of daily updates. This dataset is especially valuable for conducting in-depth analysis of the precise areas impacted by fire, focusing on vegetation cover changes through indices such as the Normalized Difference Vegetation Index (NDVI). The high cadence and fine resolution of Planet data facilitate a meticulous assessment of the immediate and longitudinal effects of fire on the landscape.

The flowchart outlines a three-step process for analyzing fire-affected areas: confirming the fire area with FIRM data, estimating burning time with Sentinel-2 data through NDVI calculation, and analyzing burn severity with high-resolution Planet data, concluding with the calculation of both NDVI and NBR (Normalized Burn Ratio) for detailed assessment.

The Normalized Difference Vegetation Index (NDVI) is a quantifiable index typically derived from space-based observations, utilized in the analysis of data acquired through remote sensing techniques. NDVI serves as a critical metric for gauging vegetation greenness, thus facilitating an understanding of vegetation density and enabling the assessment of variations in plant health. Traditionally, NDVI is computed based on the ratio of near-infrared (NIR) and red (R) spectral values：

NDVI= (NIR- R) / (NIR + R)

The three-step combustion area analysis method used in this study is a linear progression process that moves from a macroscopic perspective to a microscopic perspective in space. Essentially, after screening a large area of burning land, a small-scale vegetation cover analysis can be conducted to determine the extent affected by the fire. It combines three specific steps:

1. Obtaining approximate burn area information and determining the occurrence of a fire event through the Fire Information for Resource Management System (FIRM) database. Due to data limitations, the resolution of the burn area obtained at this stage is around 1 kilometer.
2. Using Sentinel 2 data over a large area to make rough estimates of the burning time and burn area. Sentinel 2 data has a wide range of spectral bands, allowing for a rough determination of the burn area, and it is open data, making it easily accessible. However, due to the five-day revisit cycle of Sentinel 2 and a data resolution of 20 meters, more precise data are needed for impact analysis of the burn area.
3. Utilizing commercial satellite data from Planet to conduct a detailed analysis of the precise burn area's impact, specifically through vegetation cover analysis based on the NDVI index. Planet's database has a resolution of only 3 meters, and it provides more abundant imagery from various dates, allowing for accurate analysis, often with daily image availability.

The attributes inherent in the Fire Information for Resource Management System (FIRMS) layer provide a rapid and approximate delineation of the fire's extent.

In the NDVI images, regions depicted in purple typically exhibit lower NDVI values, signifying a scarcity or absence of vegetative cover. Conversely, areas represented in green are indicative of higher NDVI values, denoting the presence of substantial vegetation cover.

Below The Interactive Map of 2023 Muai Wildfire

The research successfully utilized a fire analysis method that spanned from macro to micro levels to analyze the process and impact of a wildfire on Maui Island, Hawaii. The findings included the location of the fire-affected area and changes in vegetation coverage. The innovation in this study is mainly reflected in two aspects:

• Multi-Sources Images: Images of varying resolutions were used in this study, including FIRM data with a resolution of 1km, Sentinel 2 remote sensing data at 20m resolution, and Planet commercial database with a resolution of 3m.

• Top-Down Method: The method of analysis progressed from event-level to analysis-level, advancing from the confirmation of large-scale fire-affected areas to the detailed analysis of fire impacts within smaller regions. This approach allowed for rapid and precise analysis while also saving on data costs.

Future research could further improve and expand upon these findings through additional data and methods:

• Short-wave infrared Band Loss: For detailed analysis, Planet data lacks the short-wave infrared bands commonly used in fire analysis, which hinders the ability to extract fire-affected areas. Using high-resolution images with short-wave infrared and more bands could precisely locate fire areas and allow for a multifaceted impact analysis.

• Multifaceted Fire Impact Analysis: While the study focused on the impact of fires on vegetation coverage, it could also consider additional aspects such as urban destruction, air pollution levels, and wildlife distribution.

Ka'anapu, Kemana. "PARADISE ON FIRE: HOW HAWAI ‘I DEVELOPED ONE OF THE WORST WILDFIRE PROBLEMS IN AMERICA." Intersections: Journal of Asia Pacific Undergraduate Research 1, no. 1 (2023).

Nazeer, Malik Muhammad. "Tragic Aftermath of Californian Jungle Fire and Hawaiian Volcano Outburst, a Warning about Persistently Rising Global Warming, Resulting into High Rated Disaster’s Chain." Open Access Library Journal 6, no. 01 (2019): 1.

CNN. 2023. “‘Everything Was on Fire’: The Hours That Brought Lahaina to Ruins.” Www.cnn.com. August 18, 2023. https://www.cnn.com/interactive/2023/08/hawaii-wildfires-timeline-maui-lahaina-dg/index.html.

Leon, Jose Raul Romo, Willem J.D. Van Leeuwen, and Grant M. Casady. 2012. "Using MODIS-NDVI for the Modeling of Post-Wildfire Vegetation Response as a Function of Environmental Conditions and Pre-Fire Restoration Treatments." Remote Sensing 4 (3): 598-621.

Hislop, Samuel, Simon Jones, Mariela Soto-Berelov, Andrew Skidmore, Andrew Haywood, and Trung H. Nguyen. 2018. "Using Landsat Spectral Indices in Time-Series to Assess Wildfire Disturbance and Recovery." Remote Sensing 10 (3): 460.

Banerjee, Polash. 2021. MODIS-FIRMS and ground-truthing based wildfire likelihood mapping of Sikkim Himalaya using machine learning algorithms.

Bolouk Heidari, Farzad, and Ramin Arfania. 2022. "Wildfire Susceptibility Mapping using NBR Index and Frequency Ratio Model." Geoconservation Research 5 (1): 240-260.