Source: UNIV OF HAWAII submitted to
AGRIWATCH: INNOVATING AGRICULTURAL DISASTER RESPONSE WITH AI-EMPOWERED REAL-TIME MONITORING
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
Reporting Frequency
Annual
Accession No.
1031678
Grant No.
2024-68016-41561
Cumulative Award Amt.
$268,472.00
Proposal No.
2023-09956
Multistate No.
(N/A)
Project Start Date
Dec 1, 2023
Project End Date
Nov 30, 2025
Grant Year
2024
Program Code
[A1712]- Rapid Response to Extreme Weather Events Across Food and Agricultural Systems
Recipient Organization
UNIV OF HAWAII
3190 MAILE WAY
HONOLULU,HI 96822
Performing Department
(N/A)
Non Technical Summary
In early August 2023, a devastating fire disaster struck Maui, Hawaii, fueled by severe drought and Hurricane Dora. The serious destruction, loss of life, and widespread damage to agricultural areas, prompted a state of emergency declaration by Hawaii and federal major disaster declaration by President Biden. The National Agricultural Statistics Service (NASS), as a USDA data arm, is obligated to provide timely and accurate disaster impact assessments on agriculture. However, real-time monitoring is not available due to a lack of current crop data and up-to-date disaster information. To address these challenges, this project aims to develop high-resolution crop data layers for 2023 and 2024 using AI (Artificial Intelligence)/ML(machine learning) models, create high-resolution fire and vegetation condition maps, and build a Google Earth Engine based online disaster monitoring and impact assessment platform. The platform will integrate multi-sensor Earth observation and geophysical data, enabling near real-time disaster monitoring and impact assessment. The project will collaborate with farmer-supporting organizations to provide outreach and train stakeholders to use the data and tools. Graduates and high-school or undergraduate students from the Hawaii State 4-H program will also actively participate in research activities. The data and tools developed will be integrated into university teaching and research. This project will enhance stakeholders' disaster response capabilities by providing analysis-ready data and real-time monitoring tools. It will aid in understanding vulnerabilities, mitigating risks, and fostering resilient agricultural practices. NASS and USDA will directly benefit from the improved disaster monitoring operation and decision support in disaster relief, and mitigation amid extreme weather and disasters.
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
8077210303080%
1220120206020%
Goals / Objectives
The overarching goal of this project is to develop rapid response and impact monitoring and assessment capabilities for disasters related to fire and agriculture in Hawaii. This will be achieved by) producing latest high spatialresolution maps of cropland and near real-time high-resolution maps of fires and vegetation conditions for the state of Hawaii and 2) developing web-based tools to easily and quickly disseminate the relevant information to farmers, resource managers, scientists, and decision makers in Hawaii and USDA NASS. Specifically, we have the following specific objectives:Objective 1: Develop 10-m-resolution Hawaii Cropland Data Layers (HCDL) for 2023 and 2024 using convolutional neural network (CNN)-based deep learning models and enable crop specific vegetation conditions and disaster impact monitoring and assessment.Objective 2: Create high spatial-temporal resolution fire and vegetation condition map products by fusing Sentinel-2 and VIIRS data using machine learning approach and enable near real time high resolution agricultural disaster monitoring and assessment.Objective 3: Build a comprehensive disaster monitoring and impact assessment platform using Google Earth Engine (GEE) and integrate multi-sensor Earth observation data and geophysical data to facilitate agricultural disaster monitoring and assessment.
Project Methods
We will use the following methods to achieve our goals and generate our expected products and outcomes:1) develop and incorporate latest technologies in AI, satellite remote sensing, cloud-computing, and web applications to quickly and accurately map crop types, fire, and vegetation conditions. For high resolution crop type mapping, we will use convolutional neural netowork (CNN)-based deep learning methods that have demonstrated successes in California and Texas and adapt these technologies in Hawaii. For high spatial-temporal mapping of fire and vegetation conditions, we will fuse satellite imagery of different saptial and temporal resolutions (Sentinel-2 and VIIRS) via an enhanced super-resolution generative-adversarial network that also has been demonstrated in previous studies. For the web app development, we will integrate Google Earth Engine (GEE) and other multi-sensor Earth observation and geophysical data to make our platform comprehensive, accessible, and user-friendly.2) maximize efficiency and quality in reference data collection by leveraging resources and networks from farmer-supporting organizations and coordinating efforts of a diverse team consisting of scientists, technicians, and students. The project directors, field technician, high-school and undergraduate students are from different islands, which allow us to harness knowledge from local experts and reduce travel costs.3) disseminate our products, platform (i.e., web app), technologies, and promote their wide use by organizating training workshops and field trips that are open to a variety of organizations,farmers, managers, students, stakeholders, and other potential users, sharing ourdata products and algorithms widely while following security policies and federal regulations, and presentating our technologies in national conferences and eventually publishing them in peer-reveiwed journals.

Progress 12/01/23 to 11/30/24

Outputs
Target Audience:We have reached the potential audiences through several activiities: 1) the organization of the AgriWatch workshop and field trip, 2) the participation of the Hawaii Agriculture Conference, and 3) the involvement of staff and students in field work. Through these activities, we reached the following audiences: 1) farmer-supporting organizations such as Hawaii Farmers' Union United, Hawaii Farm Bureau, the Hawaii Food Hub Hui, and the Cooperative Extension Services of the University of Hawaii, 2) growers, 3) extension staff, 4) students, 5) USDA/NASS scientists and managers, 6) staff and managers from state governments, scientists and managers from NGO and universities. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?PD Q. Chen served as the mentorto a graduate student who worked on the project for field data collection and the automatic mapping of the basil crop.Co-PD H. Chen served as the mentor to two graduate students who worked on the super-resolution mapping of vegetation indices and fire. We organized a workshop on November 6, 9am-3:30pm, to share our findings, methodologies, and preliminary products. The workshop was well attended with close to 50 participantsfrom federal and state agencies, universities, and NGOs. We also participated in the Hawaii Agriculture Conference on November 7-8, 2024 and held an exhibition booth to share our crop maps and methdologies with the conference attendees. How have the results been disseminated to communities of interest?Our workshop news was widely circulated among local and federal agencies, classrooms, and professional organizations andit is the first time for many of them to be exposed to the topics that invovle the use of remote sensing and AI for agriculture and fire mapping. What do you plan to do during the next reporting period to accomplish the goals?We will continue to make a high resolution map of crops using imagery from 2025. We will finish the super-resolution mapping of fire. And we will deliver the final version of the GEE platforms that disseminate our maps.

Impacts
What was accomplished under these goals? During the first year, we developed the Hawaii Crop Data Layer of 10-m resolutions using Lansdat and Sentinel imagery for the year of 2024. We also developed the high spatial resolution maps of vegetation index maps by combining VIIRS and Landsat imagery. We also developed a prototype GEE-based platform to share our preliminary maps in the AgriWatch workshop and the Hawaii AgricultureConference.

Publications