California Air Resources Board

I. Objective

The California Air Resources Board (CARB) is responsible for developing quantification methodologies (QM) for estimating greenhouse gas (GHG) emission reductions and other benefits from projects funded by the Greenhouse Gas Reduction Fund (GGRF). The goal of this project is to update and improve the input data used in the CARB’s QM for Forest Restoration & Management. The contractor will explore potential new data sources, assess and validate them, and recommend QM improvements and/or calibrations using novel data. These steps will improve the accuracy of greenhouse gas estimates from forest restoration California Climate Investments projects.

II. Background

Under California’s Cap-and-Trade program, the State’s portion of the proceeds from Cap-and-Trade auctions is deposited in the GGRF. The Legislature and Governor enact budget appropriations from the GGRF for State agencies to invest in projects that help achieve the State’s climate goals. These investments are collectively called California Climate Investments. Senate Bill (SB) 862 (Committee on Budget and Fiscal Review, Chapter 36, Statutes of 2014) requires CARB to develop guidance on reporting and QMs for all State agencies that receive appropriations from the GGRF. These QMs are designed to estimate the greenhouse gas benefits of projects using information available at the time of project selection.

One of California Climate Investments major investment areas is nature-based solutions to accelerate climate smart management of California’s landscapes. This is implemented by a suite of California Department of Forestry and Fire Protection’s (CAL FIRE) programs, which funds restoration and fuel reduction treatments in California forests.

 Due to the complexity and variability of forest management, the QM for Forest Restoration & Management is among the most sophisticated of the California Climate Investments QMs. It includes methods for modeling the short-term carbon emissions associated with multiple forms of forest fuel reduction, as well as the expected value of avoided greenhouse gas emissions from the dampening effect of fuel reduction on wildfire emissions, determined probabilistically based on the likelihood that a wildfire will occur.

One of the most important input data types in the QM is maps of forest structure, which are used to determine fuelbeds (live and dead vegetation that drive fire behavior) as well as forest carbon density. The QM allows for several potential data sources to be used for forest structure map inputs, but in practice they most often come from Landscape Ecology, Modeling, Mapping, and Analysis (LEMMA) data, which is produced by the U.S. Forest Service and Oregon State University.1  The LEMMA data represent forest conditions as they were in 2012. CARB is currently exploring more recent data sources to bring the QM up to date. This project will explore and test potential new data sources for their use in two aspects of the QM: fuelbed (Task 1) and forest carbon estimation (Task 2). The contractor will review available data sources, compare them with each other and with validation data, and recommend improvements to the QM based on these findings.   

III. Scope of Work

Task 1 - Fuel Map Assessment

The QM relies on wildfire modeling of both project scenarios and baseline (counterfactual) scenarios to estimate the benefits of treatments on potential wildfire emissions. The QM calls for wildfire simulations that hold all inputs constant across scenarios except for fuelbed, which is a categorical variable describing the forest fuels. Project scenarios reduce simulated wildfire emissions through their effects on reducing modeled fuel load. Thus, accurate parameterization of fuelbed in both baseline and project scenarios is critical to estimating the greenhouse gas benefits of fuel treatment projects.

Fuel models in the QM are determined using the Fire and Fuels Extension of the Forest Vegetation Simulator (FFE-FVS). Starting vegetation for initialization of FFE-FVS model runs currently comes from LEMMA. This initialization data strongly influences wildfire emissions estimates, with a particularly strong influence on the baseline scenario, which is not influenced by the fuel modification from treatments.

Currently, CARB is in the process of working with scientists on a separate contract to migrate the FFE-FVS model runs to a cloud platform. This migration will allow for increased volume of model runs for each proposed project. As part of this migration, CARB is exploring alternative data sources for initializing FFE-FVS, such as TreeMap2  or synthetic treelists generated from imputed Forest Inventory and Analysis (FIA) data.3  These modeled forest structure maps rely on “spreading” field measurements across large land areas using satellite data. The fuel classes that primarily drive wildfire behavior (surface fuels and ladder fuels) are difficult to model using this methodological approach because satellites often cannot detect them under forest canopy cover. However, the accuracy of fuel models derived from modeled forest structure maps has not been systematically tested or calibrated.

This task will assess the quality and reliability of fuel models derived from imputed forest structure maps by comparing them to each other, to the Fuel Characteristic Classification System (FCCS) Fuelbeds map developed by LANDFIRE, and to available field data and sub-canopy remote sensing measurements. This assessment will enhance the QM by improving fuel model accuracy and uncertainty estimates in wildfire simulations.

Task 1a. Review available data sources.

The contractor will perform a review of available forest structure maps suitable for use in the QM. These maps should be “wall-to-wall” in that they cover all forests in California rather than only a specific region of the state. The contractor will also review and gather validation data suitable for assessing fuel models derived from these maps. Validation data could include: publicly available field data; field data obtained through scientific collaborations; remote sensing measurements of sub-canopy fuels (e.g., Terrestrial Laser Scanning); or others. Validation data will require precise plot locations to compare it with fuel models from forest structure maps.

Task 1b. Generate fuel models.

The contractor will use FFE-FVS as described in the QM to generate fuel models using each of the forest structure maps identified in Task 1a. These fuel models do not need to be generated for every pixel of every forest structure map. Rather, they should be generated for a randomly selected sample of pixels covering forested area of California, as well as all pixels overlapping the source locations of validation data. The sample size of randomly selected pixels will be determined based on power analysis and in consultation with CARB to identify the smallest sample size needed to ensure statistically significant results.

Task 1c. Compare generated fuel models to FCCS and validation data.

The contractor will compare data from Tasks 1a and 1b. The contractor will use statistical analysis to assess the similarity of QM-generated fuel models to validation data. Based on this analysis, the contractor will generate uncertainty estimates for use of each forest structure map in the QM. For regions of California that lack validation data, FCCS fuel models will be used for validation.

Task 1d. Identify calibration opportunities.

The contractor will make a recommendation for the best forest structure map to use for fuel model initialization in the QM and suggest potential calibration that could be performed to improve QM fuel model accuracy. For example, if the contractor identifies a systemic bias in the fuel model estimates, they may recommend a customization of forest structure map inputs or of the translation of FFE-FVS outputs to fuel model classifications.

Task 2 - Forest Carbon Assessment

Forest structure maps are used in the QM not only to determine fuelbed, but also to estimate carbon stored in live vegetation. The majority of stored carbon in forests exists in standing live trees. Accurate forest structure maps are important for quantifying the carbon stored in forest stands and therefore the carbon benefits of treatments. Forests with higher total carbon will experience greater emissions under the baseline scenario because there is more carbon to lose, which may tip the balance in favor of fuel reduction treatments in the QM. Stand-scale field measurements are the most accurate method for estimating forest carbon, but these measurements are infeasible for many project applicants. Thus, the QM relies on modeled forest structure maps.

Recent developments in satellite imagery technology have introduced new fine-resolution forest mapping products that capture variables relevant to forest carbon density. These products may provide a level of accuracy that exceeds modeled forest structure maps, without the high cost and delays inherent to field measurements. Proprietary satellite products, such as those from Planet, are expensive to obtain for wall-to-wall analyses of California forests. However, these data products may be affordable on the scale of individual projects.

Task 2a. Summarize relevant high-resolution remote sensing products.

The contractor will research high-resolution derived satellite products (e.g., maps of forest carbon density, tree density) that are currently available freely or for purchase. The contractor will compile a table showing each relevant product and its attributes, such as accuracy, resolution, time lag, thematic resolution (i.e., what forest components it captures), pricing structure, price for the average-sized forest project, pros and cons for use in the QM, and data format.

Task 2b. Compare remote sensing products to forest structure maps.

The contractor will sample the forest structure maps described in Task 1 for areas within currently funded forest projects. For those areas, the contractor will obtain samples of the remote sensing products described in Task 2a. The contractor will then compare each map and create summary statistics and supporting graphics representing the deviation between the information contained in each high-resolution remote sensing product and each forest structure map.

Task 2c. Develop a workflow that calibrates the QM forest structure inputs using remote sensing products.

Based on the results of Tasks 1 and Task 2b, the contractor will recommend a forest structure map for use in the QM. The contractor will then develop a methodology to calibrate the QM input data using project-specific remote sensing data. For example, trees per acre may be adjusted if remote sensing data estimate a higher tree density than existing forest structure maps. Note that remote sensing data products cannot fully replace modeled forest structure maps in the QM because forest structure maps have tree-level data (e.g., diameter, species) that is required for FFE-FVS simulation runs.

IV. Deliverables

The project pre-proposal must include but is not limited to the following deliverables. Proposals that respond to only Task 1 or Task 2 will be considered for a lower funding amount.

During Active Contract Period

• Work with CARB staff at the beginning of the project to create a 1-page plain-language outreach deliverable for the public describing the project’s goals, process, and planned deliverables (available in multiple languages, template will be provided).
• Quarterly Progress Reports
• Monthly progress update meetings with CARB contract manager.

Prior to Contract Close

Task 1

• Brief summary (less than two pages) describing available wall-to-wall forest structure maps and validation data suitable for assessing fuel models derived from these maps.
• Database for validation of fuel models derived from the forest structure maps. This database should include the following fields: latitude, longitude, forest structure map source, fuelbed according to the QM, fuel model according to FCCS, fuel model according to validation data.
• Reproducible code, including annotation, documentation, README files, and underlying data, sufficient for CARB staff to repeat fuel model validation and uncertainty analysis as described in Task 1c.
• Final report that describes the methods, results, and recommendations of Task 1.

Task 2

• Summary report describing relevant high-resolution remote sensing products, including a comparison table as described in Task 2a. This will be delivered to CARB prior to the start of Tasks 2b and 2c.
• Reproducible code, including annotation, documentation, README files, and underlying data, sufficient for CARB staff to recreate the comparisons described in Task 2b and the model calibration as described in Task 2c.
• Final report that describes the methods, results, and recommendations of Task 2. This report should include a revised version of the contents of deliverable (e) as well as summary statistics and graphics described in Task 2b and a description of the calibration method recommended in Task 2c.

Overall

• Presentation summarizing findings at a seminar directed to staff at CARB and other relevant state agencies.  

NOTE: For deliverables that will be publicly posted on the web, the contractor will be responsible for ensuring their documents comply with the accessibility standards in the Web Content Accessibility Guidelines 2.0 or subsequent version. CARB will consult with the contractor to identify which deliverables will be publicly posted.

V. Timeline

It is anticipated this project will be completed in 24 months from the start date (start date is estimated to be in Spring 2025). The estimated budget for this project is up to $450,000.

Scoring Criteria

1. RESPONSIVENESS TO THE GOALS AND OBJECTIVES OUTLINED IN THE PROPOSAL SOLICITATION (10 POINTS)

The proposal should explain—in adequate detail and clear, understandable language—how the proposed project satisfies the project objectives. The proposal should demonstrate an understanding of the QM for Forest Restoration & Management.

2. WORK EXPERIENCE AND SUBJECT MATTER EXPERTISE (20 POINTS)

The proposal should demonstrate that the proposers have the work experience and subject matter expertise required to successfully carry out the proposed project as described. Additionally, the proposal should describe how the project will build upon previous relevant work.

3. EXPANDING EXPERTISE (10 POINTS)

The proposal should explain how the project team expands expertise such as by incorporating multidisciplinary expertise or perspectives, including members from various public universities, non-academic institutions, or community-based organizations, or providing opportunities to build skills and expertise for individuals from underrepresented groups. Reviewers will consider if key personnel contributing significantly to the project (i.e., a principal investigator, co-principal investigator or co-investigator, contributing 25 percent or more of their time to the project) are a member of an underrepresented group.

4. EXPLANATION OF TECHNICAL OR METHODOLOGICAL APPROACH (30 POINTS)

The proposal should clearly explain the logic and feasibility of the project’s methodology, spell out the sequence and relationships of major tasks, and explain methods for performing the work. The proposal should include a clear description and plan for how each task will be completed. This should include a detailed methodology for all three tasks, including plans for study design, data collection, field work, statistical analysis, uncertainty analysis, and code development. A timeline for competing each task and subtask should be provided.  

5. LEVEL AND QUALITY OF EFFORT AND COST EFFECTIVENESS (15 POINTS)

The proposal should describe how time and resources will be allocated, how the team organization, work plan, and project management plan will ensure timely completion of tasks, and how the proposer will ensure the project’s success. Proposal reviewers will evaluate, for example: if the objectives of the project can be met given this allocation, if there is adequate supervision and oversight to ensure that the project will remain on schedule, if time and cost are appropriately divvied up across different project tasks and stages.

• 1Landscape Ecology, Modeling, M. and A. (LEMMA) G. GNN Forest Structure Map California (2014). https://lemma.forestry.oregonstate.edu/methods/methods (2014).
• 2Riley, K. L., Grenfell, I. C., Shaw, J. D. & Finney, M. A. TreeMap 2016 Dataset Generates CONUS-Wide Maps of Forest Characteristics Including Live Basal Area, Aboveground Carbon, and Number of Trees per Acre. J. For.120, 607–632 (2022).
• 3Taylor, B. et al. Modeling Individual Tree Attributes at a Landscape-Level using Mesoscale Geospatial Data. in (San Francisco, CA, 2023).