1 of 13

The 2025 Young Scholar Symposium on Spatiotemporal Data Science

Speaker: Zongrong Li, zongrong@usc.edu

Mentor: Siqin Wang

University of Southern California

2 of 13

01 Personal Introduction

​

​

​

​

​

​

Master: University of Southern California

Master of Science, Spatial data science (Expected May 2025)

Bachelor: Nanjing Audit University

Bachelor of Economics, Finance(Sep 2018 – June 2022)

​

​

​

Personal Web

3 of 13

4 of 13

01 Background

02 Study Area & Data

03 Methodology

04 Results

05 Discussion & Future Work

00 Contents

5 of 13

Wildfires have become more frequent and severe worldwide due to climate change and land-use changes, causing significant ecological and economic damage. In the summer of 2023, Maui Island, Hawaii, experienced a devastating wildfire, highlighting the risks to island ecosystems.

Aims to address three key questions:

  1. How can multi-source satellite data improve wildfire boundary delineation and vegetation assessment?
  2. What are the effects of wildfires on different land cover types?
  3. How do wildfires impact different demographic groups, and how can dasymetric mapping refine population distribution analysis?

01 Background

​

6 of 13

02 Study Area & Data

7 of 13

Suspected Burn Area

Maui Landcover

Maui Block

Open Street Map Data

Overlap

FIRMS Data

Planet Data

Natural Environment

Add Field:

“Total_Pixel”

= Sum([{Landcover Category}])

“P_{Landcover Category}”

= [{Landcover Category}] / [Total]

“Expected_Population”

= Sum([P_{Landcover Category}] * RA)

Feature to Raster

Field: Total_Pixel

Field: Expected_Population

Field based on: Block_ID

Field: Total_Population

Wildfire-affected Area

RA

Block

ID

Land

cover

Total Population

Landcover Table

Expected

Population

Total Pixel

Raster Calcul

-ation

RA

Total Population

/

×

20

=

×

×

20

×

Expected

Population

Total Pixel

Built Environment

Step1: Wildfire area identification

Step2: Downscaling population through dasymetric mapping

Step3: Tri-environmental fire impact analysis

Data Fusion

Calculate the

differential NDVI (dNDVI)

threshold: 0.35

NDVI=(NIR−R)/(NIR+R)

Converted into

polygon layers

Add Field:

“RA(Relative Weighted Value)”

High Intensity Developed: 46

Open Space Developed: 26

Palustrine Aquatic Bed: 0

Pasture or Hay: 5

Grassland: 4

Evergreen Forest: 3

Scrub Shrub: 3

Palustrine Forested Wetland: 1

Palustrine Scrub Shrub Wetland: 1

Palustrine Emergent Wetland: 1

Unconsolidated Shore: 0

Bare Land: 0

Open Water: 0

Cultivated Land: 10

Cell Size: 20m

Feature to Raster

Field: RA

Field: Land_cover

Tabulate Area

Fine-grained Population Density

Cell Size: 20m

Class field: Value

Zone field: Block_ID

Cell Size: 20m

Feature to Raster

Field based on: Block_ID

Join Data

Social Environment

Overlap

03 Methodology

8 of 13

Suspected Burn Area

FIRMS Data

Planet Data

Wildfire-affected Area

Step1: Wildfire area identification

Data Fusion

Calculate the

differential NDVI (dNDVI)

threshold: 0.35

NDVI=(NIR−R)/(NIR+R)

Converted into

polygon layers

03.1 Wildfire Area Identification

9 of 13

03.2 Population through Dasymetric Mapping

10 of 13

Open Street Map Data

Overlap

Natural Environment

Wildfire-affected Area

Built Environment

Fine-grained Population Density

Social Environment

Overlap

03.3 Tri-environmental Analysis

11 of 13

04 Results

12 of 13

05 Further Work: Application in LA

  1. Data Integration: government datasets & VIIRS nighttime light data

  1. Temporal Burned Area Analysis

​

  • Tri-environmental framework (natural, built, social)

13 of 13

Thanks for your attention!