Recipient Organization
BLUE FOREST CONSERVATION
171 5th St
Lake Oswego,OR 97034-3029
Performing Department
(N/A)
Non Technical Summary
Forest restoration, or reducing the vegetation density of overgrown forestland, is effective at reducing high-severity wildfire risk across the western US. Restoration can also increase the water yield from forests in many different geographic regions, but the effects are highly variable depending on the individual watershed location, vegetation, and climate characteristics.Accurate, affordable, and scalable measurement of water yield enhancement following ecologically-based forest restoration would enable a full accounting of the water volume benefits from specific forest restoration projects. This approach could allow the costof restoration to be shared among multiple downstream beneficiaries, including hydroelectric power and water utilities.Several methods are commonly used to measure changes in water yield following forest restoration. The traditional paired-watershed approach is a robust and proven method, but expensive and time-consuming, resulting in a poor fit for the requirement of a quick and mobile application to assess the impacts of scalable forest restoration. Using sophisticated hydrologic modeling is more suitable to a low-cost scalable approach, but the accuracy of models can vary widely depending on model selection and calibration data availability. Model simulations can be calibrated exclusively using remote-sensing and other existing data (e.g. temperature records and streamflow on major rivers) however additional ground-based data can dramatically improve model accuracy. Installing and maintaining ground-based monitoring equipment, however, is also expensive and time-consuming. For scalable application across large restoration watersheds, understanding the relative modeling accuracy value of each measurement component would allow prioritization of limited measurement resources.As of yet, no comprehensive analysis hasbeen performed in the California Sierras to evaluate the opportunity to use remote sensing data on quarterly and annual time steps to measure changes in water yield due to vegetation change. While sophisticated hydrologic modeling has been completed showing significant changes in water yield, we propose evaluating how well remote sensing approaches would have correlated with these past modeling results.We propose to develop a remote-sensing-based watershed scale tool for determining water yield changes following forest restoration in the California Sierra Nevada.To create this framework, we will evaluate the accuracy of existing physically-based hydrologic models compared to results we will predict using different remote sensing data for assessing water yield changes following forest restoration.
Animal Health Component
40%
Research Effort Categories
Basic
20%
Applied
40%
Developmental
40%
Goals / Objectives
Forest restoration, or reducing the vegetation density of overgrown forestland, can be effective at reducing high-severity wildfire risk across the western US (Agee and Skinner, 2005; Finney et al. 2008). Restoration may also increase the water yield from forests in some cases in many different geographic regions (Stednick, 1996; Brown et al., 2005), but the effects are highly variable depending on the individual watershed location, vegetation, and climate characteristics. Accurate, affordable, and scalable measurement of water yield enhancement following ecologically-based forest restoration would enable a full accounting of the water volume benefits from specific forest restoration projects. More sophisticated measurement and tools could allow the costs of restoration to be shared among multiple downstream beneficiaries, including hydroelectric power and water utilities.Several methods are commonly used to measure changes in water yield following forest restoration. The traditional paired-watershed approach is a robust and proven method, but expensive and time-consuming, resulting in a poor fit for the requirement of a quick and mobile application to assess the impacts of scalable forest restoration. Using sophisticated hydrologic modeling is more suitable to a low-cost scalable approach, but the accuracy of models can vary widely depending on model selection and calibration data availability. Model simulations can be calibrated exclusively using remote-sensing and other existing data (e.g. temperature records and streamflow on major rivers) however additional ground-based data can dramatically improve model accuracy. Installing and maintaining ground-based monitoring equipment, however, is also expensive and time-consuming. For scalable application across large restoration watersheds, understanding the relative modeling accuracy value of each measurement component would allow prioritization of limited measurement resources. To our knowledge, no comprehensive analysis has yet been performed in the California Sierras to evaluate the opportunity to use remote sensing data on quarterly and annual time steps to measure changes in water yield due to vegetation change. While sophisticated hydrologic modeling using the RHESSys model has been completed showing significant changes in water yield, we propose evaluating how well remote sensing approaches would have correlated with these past modeling results.We propose to develop a remote-sensing-based watershed scale tool for determining water yield changes following forest restoration in the California Sierra Nevada. To create this framework, we will evaluate the accuracy of existing physically-based hydrologic models compared to results we will predict using different remote sensing data for assessing water yield changes following forest restoration.Our overall R&D goal is to demonstrate the feasibility of a customizable and scalable landscape scale water quantity evaluation framework that can be broadly applied to quantify changes in watershed yield following forest restoration across the western US. At the core of this framework will be a rigorous decision-making matrix based on technical, logistical, and environmental feasibility, along with cost and water yield prediction accuracy associated with different hydrologic measurement scenarios.We will initially assess the feasibility of this framework using data from an existing research project in the northern Sierra at the Last Chance study site (Figure 2, pg 10). Last Chance site has been intensively instrumented, studied, and modeled by Dr. Roger Bales and Dr. Martha Conklin at the Sierra Nevada Research Institute (SNRI - researchers affiliated with SNRI are a subaward recipient). This completed analysis using past complex hydrologic models will be used as a "best-case" hydrologic measurement scenario. Our feasibility assessment will focus on comparing the water yield accuracy, cost, and feasibility of broadly applicable and low-cost hydrologic quantification scenarios based on remote sensing data. This will be done by acquiring the MODIS satellite NDVI data for the Last Chance site that overlaps the period of measurements and treatments. We will investigate the data requirements, accuracy, and application of the remote sensing approach for comparison to existing hydrologic model outputs for this watershed. The measurement scenarios will consist of:Scenario 1: Quantification of water yield response to forest treatments comparing past RHESSys model outputs to remote sensing-based evapotranspiration estimates (both MODIS and NDVI-ET relationship) using a water balance over varying time steps.Scenario 2: Quantification of water yield response to forest treatments using the MODIS evapotranspiration estimates in both treated and adjacent untreated watersheds.Specific technical questions that will be addressed during this feasibility phase to quantify changes in water yield will allow us to determine the relative accuracy of each of the scenarios:Is there a statistically significant difference in quantifying the water balance using the remote sensing approach compared to the physical modeling at the 95% confidence level?Are the remote sensing approaches significantly correlated with the direction and magnitude of water yield changes in response to forest restoration?What is the estimated cost for predicting changes in streamflow following forest restoration using data-intensive modeling versus remote sensing?What is the feasibility to scale-up a remote sensing water quantity framework across the American River watershed and other Sierra Nevada catchments?
Project Methods
Phase 1 consists of five tasks.Task 1: Compile data from the Last Chance study siteTask 1 will focus on compiling and processing results from the Last Chance study site Regional Hydro-Ecologic Simulation System (RHESSys; Tague and Band, 2004) results for comparison to the remote sensing analysis conducted in Tasks 2 and 3. These data will include ground-based measurements from the Last Chance site as well as widely available spatial data including soil, topography, and remote sensing.Forest treatments were completed at the Last Chance site in 2011-12, with adjacent watersheds left as controls. Restoration entailed partial thinning, mastication, and prescribed burning in accordance with ecologically based practices adapted from GTR-220 (North et al. 2009), which reflects the current US Forest Service approach to California Sierra Nevada vegetation management. Briefly, this consisted of strategically placed thinning and removal of 8% of the total vegetation across the treated portions of the Last Chance site, including one of the test catchments. The test catchments are approximately 400 acres with outlet elevations of approximately 5100 feet. Both test catchments are heavily monitored to allow precise measurement of the hydrologic response to forest restoration. The sites have been monitored continuously since 2008 with treatments completed in 2012 to allow for comparisons before and after restoration. At a minimum, it is anticipated that the following data will be compiled in Task 1:Daily streamflow: Ground-based. Daily streamflow measurements are collected from both test catchments using the stage-discharge technique.Meteorological data: Ground-based.Two meteorological stations are located in the Last Chance site to record precipitation, temperature, wind speed, and solar radiation.Detailed vegetation surveys: Ground-based.Vegetation surveys were completed by ground-based visual inspection to collect basal area, leaf area index (LAI), percent canopy cover, and canopy density by the UC Berkeley SNAMP Forest Health team.MODIS-Et and NDVI Data: Satellite-based.Remote sensing data from public repositories for the Last Chance site from 2003 to 2016.?Task 2: Measurement Scenario 1: Use MODIS ET and NDVI data for the Last Chance site from 2003 to 2016 and quantify the monthly, quarterly (3-month), and annual water yield based on changes in evapotranspiration using the water balance equation.Task 2 will focus on determining streamflow prediction accuracy using remote sensing data and the simple water balance equation.Changes in soil storage offset to zero over multiple years, and groundwater bypass flow is not expected to be a dominant process in these watersheds. The MODIS satellite evapotranspiration product (MODIS-ET) and spectral reflectance vegetation products (i.e.NDVI) are readily available through public online repositories. New and innovative companies, such as Planet Labs, are developing higher resolution and frequency spatial vegetation products that could eventually be incorporated into this approach. The quarterly to annual estimates are a sufficient first step for utility stakeholders interested in how additional water and power would provide economic value. A simple remote sensing approach would provide a low-cost method for utilities to quantify water yield benefits over large landscape scale restoration sites. By calculating the reduction in evapotranspiration that corresponds to an increase in water yield, this metric has the potential to provide the robust tool utilities need to recognize additional water yield.Task 3: Measurement Scenario 2: Use the NDVI and MODIS-ET data for the Last Chance site from 2003 to 2016 and estimate monthly, quarterly, and annual water yield based on changes in transpiration as compared to adjacent watersheds of restored/unrestored forest.Task 3 will evaluate a new method for estimating changes in water yield by building on the traditional paired catchment approach. Paired catchments require downstream instrumentation to show correlated changes in stream flow before and after restoration as measured through weirs at the catchment outlet. The remote sensing paired region approach uses both NDVI and MODIS-ET data in forest treatment areas and adjacent untreated areas to calculate a difference in evapotranspiration loss following restoration. A comparison between paired regions will show how much evapotranspiration would have been expected had no restoration occurred. This task will use an adjacent area approach, rather than the water balance equation, to predict the change in streamflow water quantity based on remote sensing.Task 4: Compare remote sensing estimates of the water balance to the existing gauged data and RHESSys model output across monthly, quarterly and annual time steps.The goal of this task is to determine the relative accuracy of the remote sensing estimates to a data-intensive model. While any remote-sensing method may yield lower water yield quantification accuracy than ground-based methods, this approach provides several important advantages including lower cost, abundant historical data availability, and ease of scaling across landscapes. To determine whether these advantages could outweigh a reduction in accuracy compared to intensively instrumented sites, we will compare the accuracy of this method. Several remote sensing methods can be used to estimate streamflow by calculating the difference between precipitation and evapotranspiration across a watershed. Measurement scenario 1 compares both the MODIS-ET product and the NDVI-ET relationship (Figure 1) over quarterly and annual time periods. Spatial estimates of precipitation can be estimated from the Parameter-Elevation Regression on Independent Slopes Model (PRISM; Daly et al., 1997) dataset, which provide spatially explicit precipitation and temperature records. Evapotranspiration can be estimated using a physically-based calculation from MODIS satellite data (Mu et al., 2007), however, this method tends to have lower accuracy in mountainous forested watersheds (Glenn et al., 2010). Regression-based remote sensing models that correlateinsituevapotranspiration measurements with remotely sensed vegetation indices can also be employed. SNRI scientists recently used this method to predict streamflow in the southern Sierra (Goulden et al., 2012). We will use both of these remote sensing methods.?Task 5: Collect and compare cost, feasibility, accuracy, and precision data for each approach to estimating water quantity changes following forest restoration across larger landscape areas in the Sierra Nevada.Our final task will be to evaluate the cost and feasibility of using each of the hydrologic measurement scenarios across larger landscapes in the Sierra Nevada. We will expand the remote sensing water balance analysis (2003-2016) to the Sierra Nevada section of the American River Basin, which flows into Folsom Reservoir near Sacramento. The 1773 mi2watershed includes 16 hydropower facilities operated by multiple organizations including the Sacramento Municipal Utility District, Placer County Water Agency, and PG&E. Measurement cost and feasibility will be critical considerations for large scale restoration projects and need to be well understood. Accordingly, we will work to assess these by conducting interviews and reviewing records from SNRI projects, reviewing US Forest Service Schedule of Proposed Actions (SOPA) for potential projects that may have meaningful landscape vegetation changes. Feasibility criteria for each scenario will include technical complexity, availability of baseline remote sensing data, accuracy of remote sensing approaches as compared to the more instrumented approaches, stakeholder utility acceptance, and scale-up potential.