Forest Carbon Yield Modeling (Cloud Computing / AI)
Description
This project involves using cloud computing services such as virtual machines to conduct data processing, analysis, and visualization of large data sets that are publicly available from the USDA Forest Service, FIA program. Such outputs include the summarization of forest types, figures made on demand from simulation outputs and database queries for communication to better service stakeholders and the public while reaching department strategic goals and prioritizes like the NJ State Forest Action Plan. The project will require the use of cloud services by four NJ Forest Service personnel, including myself. Services to be used include temporary storage to load large data sets into models, elastic computing to provide virtual machines configured to conduct machine learning and simulation modeling, and potentially ready-made interfaces designed specifically for machine learning.
Problem Statement
The NJ Forest Service is tasked with developing a forest carbon yield model for the state using publicly available data from the USDA Forest Service, Forest Inventory and Analysis (FIA) program. These datasets have become relatively large (approaching 20 GB) and simulation and analysis using domain specific models, such as the Forest VegetationSimulator (FVS), and machine learning models, such as those in SciKitLearn, have become impractical to run on conventional desktop systems, often taking several hours to complete a single iteration of analysis, and are better suited to cloud based architecture such as Amazon Web Services (AWS) or Microsoft Azure, for example.
Project Justification
Stakeholders have expressed interests in experimenting with results while discussing of technical material with NJFS staff being able to discuss results while in an active meeting. NJFS is hoping to capture efficiencies by being able to conduct detailed analyses in shorter amounts of time. NJFS currently uses contractors that use this technology for data analysis and simulation work on specific projects, that is similar to how NJFS staff currently conducts these activities for larger planning efforts without the benefits of distributed computing. State Forest Service counterparts like Washington State and the USDA Forest Service currently utilize such services to support data analysis needs.
Estimated Transactions
Multiple iterations of models to be run.
Components
Large Language Model (LLM)
Target Rollout Date
1 July 2024
Target Rollout Date Reason
July 2024; This start would coincide with the beginning of the stakeholder process for updating the USDA Forest Service regarding the NJ State Forest Action Plan; stakeholders have expressed interest in forest carbon yield
Attachments
FILE
NJFS_CloudComputing_it-project-sheet-10f.pdf
2024-10-02 20:12
239 KiB
Activity
Bill Zipse and team have successfully leveraged and continue to use the cloud resources to execute and iterate with the modelling they do.
DOIT has secured access to AWS and Azure AI services as of December. Work is needed to learn the platform.