A set of tools to optimize forest management for carbon, presently being developed to assess practices in the state of New Jersey.
The ForMOM project is possible due to the combined effort of many NJFS employees. This project is organized by Bill Zipse, and has been the consistent work of Lauren Gazerwitz, Courtney Willitts, Michael Gorbunov, Ben Pisano, Bernie Isaacson & Justin Gimmillaro, with participation from Mike Hart, Rosa Yoo & Jason Grabowsky (Rutgers University).
The FIADB re-organization work was primarily accomplished by Lauren Gazerwitz, the FVS work was primarily accomplished by Courtney Willitts, Bernie Isaacson, and Justin Gimmillaro, and the data processing and building of visuals was primarily completed by Bernie Isaacson and Ben Pisano. The development of the database automation process, the constraint builder, and the .csv to .dat converter were completed by Michael Gorbunov.
If you have questions specifically about the software here, please reach out at [email protected].
Roughly speaking there are two steps in the process of forest management optimization.
- Step 1: Simulate different management scenarios for different forest types
- Step 2: Add constraints and optimize to find the optimal management
The simulation of forest management is done primarily with FVS. We take the state inventory from FIA, re-organize it into distinct forest types, and then run it through FVS.
Most of the code in this repository is to assist with step 2. After getting values from FVS, we write constraints using a custom program for constraint building. The constraints are then aggregated and the linear optimization problem is solved using Pyomo, a python package.
The optimization software is for optimization, not carbon optimization. It can be used for any optimization problem.
Before optimizing managements, we simulate them. We take state data from FIA, re-organize it by forest type and year, and run it through FVS. The FVS runs are not in this repository. but the tool we use to re-organize the database are located in src/fvs/input. For information on running it, look to the wiki page about it.
The final output tells us how much carbon is sequestered by forest type, by year, by management. These values give us the objective function.
With an objective function (here's an example old one) from FVS simulations, we need to build constriants and run the model. Constraint building is done with the custom written Constraint Builder. It allows for easier setting up of problems with 100s of variables and constraints. Once constraints are built, the model can be run.
The process of running the model is currently... not ideal. There are a lot of intermediate data formats, and over or under constrained setups are difficult to debug. The constraint builder program will give you a csv containing the actual matrix of the linear optimization problem. This file, along with the objective file get turned into a .dat file by the csvToDat program. Then you can feed this .dat straight into the pyomo runner.
Doing that once would be ok, but to debug the model it's easier to modify the constraint csv, which means re-converting to .dat. The developers are aware this is an issue.
The ForMOM Project maintains three pieces of software
- ForMOM DB Reformatter - for simplifying FIADB data
- ForMOM Builder - for building constraints and models
- ForMOM Runner - for running linear optimization models
Each of them live in their own repo, which are linked in the above bullet points. Additionally, you find info on the Wiki.