Integrating Remote Sensing and Machine Learning with Ecosystem Services Modeling

Providing Tools for USDA Monitoring and Assessment

CBI is using the latest remote sensing, machine learning, and modeling approaches to assess the quality of grasslands and pollinator habitat. Outputs from this analysis will be delivered to USDA through an online-decision support system, designed and built by CBI, called the USDA Conservation Toolkit. This suite of map-based tools will support the USDA Conservation Reserve Program (CRP) and other land-based conservation and restoration programs operating across the United States. CRP currently covers ~24.8 million acres of working lands nationwide.

Goal: Assess ecosystems services across the CRP portfolio

• Generate products to support USDA policies and program initiatives
• Provide actionable conservation intelligence

Targets

• Quantify foundational supporting ecosystem services across CRP enrollments
  • Grasslands ecosystem quality
  • Pollinator habitat quality
• Incorporate resulting spatial data and metrics into online applications, allowing for program reporting at multiple scales
• A priority for this effort is to quantify the CRP national portfolio return on investment, screening enrolled farms, and assessing at state, regional and national scales

CBI CRP Grassland Quality Model – Overview

Components

• Integrative, comprehensive approach to modeling using the best available spatial data and remotely sensed metrics (e.g. National vegetation surveys, NLCD, Rangeland Analysis Platform (RAP), custom CBI products)
• Leverages outputs from the Google Earth Engine (GEE) cloud-computing platform with CBI’s customized library of 200+ vegetation indices

How is it unique?

• Hierarchical measure of grassland ecosystem vigor, community composition, and resilience
• Expandable framework produces multidimensional metrics to quantify patterns across the landscape
• Summarizes information across CRP farms and practices
• Offers insights on baseline current state and local condition at multiple scales

Pollinator Habitat Quality Model – Overview

What does it offer USDA?

• Multi-dimensional, non-monetary valuation of pollinator habitat
• Foundation for field to landscape-level spatial analytics for CRP assessment and reporting
• Illustrates CRP return on investment at a programmatic scale and quantifies enrolled land performance with nuanced metrics

How is it unique?

• Builds on the benchmark InVest© model framework 
• Informed by 300,000 bee observations from the Global Biodiversity Information Facility (GBIF)
• Innovations for bee flight, nesting, and floral suitability estimates

Abstract:

“Wildfires can be devastating for social and ecological systems, but the recovery period after wildfire presents opportunities to reduce future risk through adaptation. We use a collective case study approach to systematically compare social and ecological recovery following four major fire events in Australia and the United States: the 1998 wildfires in northeastern Florida; the 2003 Cedar fire in southern California; the 2009 Black Saturday bushfires in Victoria, southeastern Australia; and the 2011 Bastrop fires in Texas. Fires spurred similar policy changes, with an emphasis on education, land use planning, suppression/emergency response, and vegetation management. However, there was little information available in peer-reviewed literature about social recovery, ecological recovery was mostly studied short term, and feedbacks between social and ecological outcomes went largely uncon- sidered. Strategic and holistic approaches to wildfire recovery that consider linkages within and between social–ecological systems will be increasingly critical to determine if recovery leads to adaptation or recreates vulnerability.

Abstract:

Ecological connectivity is increasingly acknowledged as crucial for biodiversity conservation. Iverson et al. suggest that increasing stewardship to ensure permeability is a better approach than protecting linkages between protected areas. We argue that the optimal approach depends on the landscape context, conservation goals, and species involved and suggest that linkage plans can prioritize specific places for protection and improved management. However, when using connectivity models as predictive tools, model validation is vital. We commend Iverson et al. for assessing whether modeled linkages were important predictors of species presence. We disagree, though, with the authors’ conclusion that their findings challenge the theory and practice of modeling linkages and explain that the reason may be the misalignment of the validation assumptions with model objectives. We offer our perspective on best practices for conducting validation studies and note factors to consider with respect to data used for model validation and model expectations.

Enhancing Customer Experience, improving assessments, and streamlining workflows 

The pilot version of the toolkit was designed to enhance customer experience, and better manage CRP enrollment and assessments. It initially covers the states of Colorado, Kansas, Nebraska, North Dakota, South Dakota, and Washington. The USDA CRP toolkit (Pilot) was developed with a grant from USDA FSA to Conservation Biology Institute (CBI). It was designed incorporating extensive inputs from USDA staff at the federal, state and county offices, producers, and vendors. More states will be added over time as stakeholders find value and funding is available. 

The version 1.0 of the current set of three tools that are released are publicly available here.

The map-based Plant Finder tool allows you to easily find plants suitable for your location that meet particular growing conditions (e.g., ecological sites, soil conditions, moisture requirements, and specific environmental tolerances) as well as provide specific benefits (e.g., livestock browse, wildlife habitat, and pollinator support). Once a species list is generated, the Vendor Match tool identifies where seeds and/or seedlings from your species list can be purchased. The Vendor Browse tool allows you to easily browse a map view of the registered vendors in your state or across the country where seeds and seedlings can be purchased. 

All data content is from authoritative sources, including NRCS state resources, NRCS Ecological Site program, and the USDA PLANTS database. 

The vendor match and browse tools have been updated in partnership with the Institute for Applied Ecology (IAE)for the Native Seed Network. Note that IAE uses its legacy name `Seed Finder tool’ for what is essentially the same as the USDA vendor match tool.

Tutorial Videos

Southern California’s montane conifer forests are primarily restricted to the “sky islands” of the San Jacinto, San Bernardino, and San Gabriel Mountains. These unique ecosystems protect the upper watersheds of all of the region’s major rivers and provide ecosystem services critical to both human and ecosystem climate resilience. Managers are in a race to restore resilience to these forests, which are threatened with conversion to hardwood and shrub due to severe wildfires and regeneration failure. 

This partnership between the US Forest Service, San Diego State University, and Conservation Biology Institute is applying the latest research on interactions between multiple disturbances specific to this ecoregion to plan effective conservation action. 

The effort expands on research from the Connecting Wildlands and Communities project that developed a landscape-scale framework to map refugia from multiple stressors, and ongoing research projects at CBI developing dynamic wildfire and vegetation succession models for understanding the synergistic impacts of climate change, land use change, and different management scenarios.

The team will work with scientists and managers to build interactive spatial models using CBI’s Environmental Evaluation Modeling System (EEMS) with location-specific data to support the collaborative development of a conservation strategy customized to address the threats faced by southern California’s montane forests.

By disrupting wildfire regimes, climate and land use change transform ecosystems, alter carbon budgets, and drive socio-economic impacts in California. We propose to quantify how projected peri-urban growth in the wildland urban interface, climate change, and local management actions influence wildfire activity and downstream effects on vegetation transitions, carbon release, biodiversity, and vulnerable human communities. 

 We will model wildfire risk as a function of a limited set of stakeholder-guided, realistic future scenarios, using the LANDIS-II simulation model. Past studies have shown the importance of human ignition location and timing on wildfire activity in Southern California, a mechanism we will explore using different land use change scenarios. 

 We will also model and analyze the potential influence of increasing atmospheric aridity on fire size and severity in forests and compare these effects to adjacent shrubland plant communities. 

 Taken together, climate, land use, and other environmental variables can lead to fire-driven vegetation type conversion, which can influence carbon sequestration, biodiversity, and even future fire risk to human communities. 

  Finally, we will explicitly evaluate how different modeled management actions influence wildfire, plant succession, and carbon sequestration. The resulting maps and spatial products will help managers prioritize locations for conservation and management actions. Overlaying maps of vulnerable human communities and biodiversity hotspots with future wildfire change and downstream impacts can better define locations for priority action to facilitate co-benefits to human and natural resources.   This research will contribute to scientific publications and be directly relevant to managers, including the Southern California Montane Forests Project.

The San Diego Zoo Wildlife Alliance (SDZWA) is a world-renowned conservation organization with a 1800-acre zoo, the Safari Park, that houses more than 2600 animal and 3500 plant species. Located in Escondido (northern San Diego County), the Safari Park also includes a 800-acre biodiversity preserve with some of the most well-preserved, California-endemic coastal sage habitat. 

 In 2007, the Witch fire burned through 600 acres of the biodiversity reserve and stopped just outside the Safari Park borders. Since then, the SDZWA has transformed their in-house wildfire preparedness program, including hiring a team of dedicated wildfire suppression staff, fire suppression equipment boxes scattered throughout the park, an infrared and visible spectrum camera array for early ignition detection, and a 8500-gallon water tank for aerial suppression by local first responders. 

 While these investments in wildfire readiness are important for protecting the Safari Park grounds, landscape-scale risk analyses are needed to determine where fires may start and spread outside of park boundaries and which additional management activities could mitigate risk from fires originating beyond Park borders. Conservation Biology Institute (CBI) and Dudek have proposed to co-create the needed wildfire risk analysis for the SDZWA Safari Park.

 At the core of the risk analysis is a fusion and comparison between two commonly employed modeling efforts used in

CBI is a partner in producing a statewide wildlife connectivity ensemble model and associated mapped data in support of a connectivity action plan for the state, which is under development. The modeling will identify priority areas across the state in need of some form of mitigation to promote wildlife connectivity and reduce wildlife/vehicle collisions at the same time. The co-production process includes the close collaboration with an active Technical Advisory Group. The project is being funded by the Washington Department of Fish and Wildlife.

The Giant Sequoia Lands Coalition (GSLC) is a collaboration of public and non-governmental organizations committed to the conservation of giant sequoia grove ecosystems. CBI is supporting the GSLC’s Sequoia Grove Protection and Resiliency Project with tools and assistance that will enable GSLC partners and stakeholders to work collaboratively to protect the species from the impacts of drought, beetles, and wildfire made more intense by climate change.

CBI is using its Data Basin platform (databasin.org) to host the Giant Sequoia Data Portal. The portal will be populated with custom maps of the sequoia groves linked to standard metrics and data for assessing grove condition, fuel treatments, and other range-wide information. To inform this giant sequoia management “dashboard” CBI is working with the GSLC to catalog an extensive collection of assessment, monitoring, and remote sensing datasets created by different agencies and projects over many decades of forest management and research. CBI will train and support the GSLC partners in their use of the Giant Sequoia Data Portal for promoting coordinated scientific research, monitoring, long-term species conservation, and continuing to share new data as it becomes available.

In the 1850s, immigrants seeking gold in California’s Sierra Nevada mountains created a connected network of open channels, wooden flumes, and pipes to direct water to their operations and the rapidly-growing foothill towns of Sonora, Columbia, and Jamestown. 

170 years later, this historic system is an integral part of the water infrastructure that supports residential, agricultural, hydroelectric, ecosystem, and recreational purposes, providing nearly all of the drinking water to the west slope communities of Tuolumne County. Tuolumne Utilities District (TUD), the agency responsible for managing the 70-plus miles of raw water ditches and potable water infrastructure, is contending with a modern-day concern: the threat of severe wildfire.

Conservation Biology Institute is proud to partner with EN2 Resources, Inc. and the Tuolumne Utilities District (TUD) to develop the TUD Wildfire Defense Plan, a roadmap for addressing wildfire risk to the water system and water treatment facilities.

The TUD Wildfire Defense Plan will have two components: CBI is heading up a Wildfire Risk Mitigation Plan to guide projects that achieve fuels reduction, habitat enhancement, and recreation benefits. EN2 is developing a Wildfire Protection Draft-Points Plan for strategically identifying raw water draft-points along the raw water ditches for fire response and preserving precious potable water.

Together these plans will help TUD manage the system as “green infrastructure”, a vision that addresses the integrated needs of people, the communities, and biodiversity under climate change. TUD, together with Pacific Gas & Electric, the US Forest Service, CALFIRE, Tuolumne County, Tuolumne Fire Safe Council and the Tuolumne Band of Me-Wuk Indians are already working intensively to reduce fuels in the region, and this Plan will assist the agency in obtaining the funding needed to continue this important work.

The TUD Wildfire Defense Plan will be completed by the end of the year. Funding for this project comes from the USDA Forest Service Community Wildfire Defense Grant.