Browsing by Author "Coulston, John W."
Now showing 1 - 20 of 24
Results Per Page
Sort Options
- Approximating Prediction Uncertainty for Random Forest Regression ModelsCoulston, John W.; Blinn, Christine E.; Thomas, Valerie A.; Wynne, Randolph H. (2016-03)Machine learning approaches such as random forest have increased for the spatial modeling and mapping of continuous variables. Random forest is a non-parametric ensemble approach, and unlike traditional regression approaches there is no direct quantification of prediction error. Understanding prediction uncertainty is important when using model-based continuous maps as inputs to other modeling applications such as fire modeling. Here we use a Monte Carlo approach to quantify prediction uncertainty for random forest regression models. We test the approach by simulating maps of dependent and independent variables with known characteristics and comparing actual errors with prediction errors. Our approach produced conservative prediction intervals across most of the range of predicted values. However, because the Monte Carlo approach was data driven, prediction intervals were either too wide or too narrow in sparse parts of the prediction distribution. Overall, our approach provides reasonable estimates of prediction uncertainty for random forest regression models.
- Auxiliary information resolution effects on small area estimation in plantation forest inventoryGreen, P. Corey; Burkhart, Harold E.; Coulston, John W.; Radtke, Philip J.; Thomas, Valerie A. (2020-10)In forest inventory, traditional ground-based resource assessments are often expensive and time-consuming forcing managers to reduce sample sizes to meet budgetary and logistical constraints. Small area estimation (SAE) is a class of statistical estimators that uses a combination of traditional survey data and linearly related auxiliary information to improve estimate precision. These techniques have been shown to improve the precision of stand-level inventory estimates in loblolly pine plantations using lidar height percentiles and thinning status as covariates. In this study, the effects of reduced lidar point-cloud densities and lower digital elevation model (DEM) spatial resolutions were investigated for total planted volume estimates using area-level SAE models. In the managed Piedmont pine plantation conditions evaluated, lower lidar point-cloud densities and DEM spatial resolutions were found to have minimal effects on estimates and precision. The results of this study are promising to those interested in incorporating SAE methods into forest inventory programs.
- Biomass Estimation Using the Component Ratio Method for White OakDeYoung, Clara (Virginia Tech, 2014-08-26)With higher demands on biomass, the ability to accurately estimate the amount in a stand is more important now than ever before. Existing models currently in use by the Forest Inventory and Analysis (FIA) program of the United States Department of Agriculture (USDA) Forest Service include the Component Ratio Method (CRM). However, testing of the CRM models is needed to validate and calibrate them. The objective of this research was to test and develop a system of equations capable of producing consistent volume and biomass estimates for standing trees of commercially important hardwood species in the southeastern United States. Testing and comparing was done through use of new and legacy data to establish component ratios of trees and contrast these results to those from existing models. Specifically, analyses were completed for models of merchantable and whole stem volume, wood densities models and averages, and the component ratios for wood, bark, branches, and foliage. The existing models were then calibrated and adjusted. Results on accuracy and fitted results of updated models are reported, along with testing the effects of applying updated models over the state of Virginia.
- Creating Landscape-Scale Site Index Maps for the Southeastern US Is Possible with Airborne LiDAR and Landsat ImageryGopalakrishnan, Ranjith; Kauffman, Jobriath S.; Fagan, Matthew E.; Coulston, John W.; Thomas, Valerie A.; Wynne, Randolph H.; Fox, Thomas R.; Quirino, Valquiria F. (MDPI, 2019-03-06)Sustainable forest management is hugely dependent on high-quality estimates of forest site productivity, but it is challenging to generate productivity maps over large areas. We present a method for generating site index (a measure of such forest productivity) maps for plantation loblolly pine (Pinus taeda L.) forests over large areas in the southeastern United States by combining airborne laser scanning (ALS) data from disparate acquisitions and Landsat-based estimates of forest age. For predicting canopy heights, a linear regression model was developed using ALS data and field measurements from the Forest Inventory and Analysis (FIA) program of the US Forest Service (n = 211 plots). The model was strong (R2 = 0.84, RMSE = 1.85 m), and applicable over a large area (~208,000 sq. km). To estimate the site index, we combined the ALS estimated heights with Landsat-derived maps of stand age and planted pine area. The estimated bias was low (−0.28 m) and the RMSE (3.8 m, relative RMSE: 19.7%, base age 25 years) was consistent with other similar approaches. Due to Landsat-related constraints, our methodology is valid only for relatively young pine plantations established after 1984. We generated 30 m resolution site index maps over a large area (~832 sq. km). The site index distribution had a median value of 19.4 m, the 5th percentile value of 13.0 m and the 95th percentile value of 23.3 m. Further, using a watershed level analysis, we ranked these regions by their estimated productivity. These results demonstrate the potential and value of remote sensing based large-area site index maps.
- Decision Support for Operational Plantation Forest Inventories through Auxiliary Information and SimulationGreen, Patrick Corey (Virginia Tech, 2019-10-25)Informed forest management requires accurate, up-to-date information. Ground-based forest inventory is commonly conducted to generate estimates of forest characteristics with a predetermined level of statistical confidence. As the importance of monitoring forest resources has increased, budgetary and logistical constraints often limit the resources needed for precise estimates. In this research, the incorporation of ancillary information in planted loblolly pine (Pinus taeda L.) forest inventory was investigated. Additionally, a simulation study using synthetic populations provided the basis for investigating the effects of plot and stand-level inventory aggregations on predictions and projections of future forest conditions. Forest regeneration surveys are important for assessing conditions immediately after plantation establishment. An unmanned aircraft system was evaluated for its ability to capture imagery that could be used to automate seedling counting using two computer vision approaches. The imagery was found to be unreliable for consistent detection in the conditions evaluated. Following establishment, conditions are assessed throughout the lifespan of forest plantations. Using small area estimation (SAE) methods, the incorporation of light detection and ranging (lidar) and thinning status improved the precision of inventory estimates compared with ground data alone. Further investigation found that reduced density lidar point clouds and lower resolution elevation models could be used to generate estimates with similar increases in precision. Individual tree detection estimates of stand density were found to provide minimal improvements in estimation precision when incorporated into the SAE models. Plot and stand level inventory aggregations were found to provide similar estimates of future conditions in simulated stands without high levels of spatial heterogeneity. Significant differences were noted when spatial heterogeneity was high. Model form was found to have a more significant effect on the observed differences than plot size or thinning status. The results of this research are of interest to forest managers who regularly conduct forest inventories and generate estimates of future stand conditions. The incorporation of auxiliary data in mid-rotation stands using SAE techniques improved estimate precision in most cases. Further, guidance on strategies for using this information for predicting future conditions is provided.
- Enhancing the precision of broad-scale forestland removals estimates with small area estimation techniquesCoulston, John W.; Green, P. Corey; Radtke, Philip J.; Prisley, Stephen P.; Brooks, Evan B.; Thomas, Valerie A.; Wynne, Randolph H.; Burkhart, Harold E. (2021-07)Naional Forest Inventories (NFI) are designed to produce unbiased estimates of forest parameters at a variety of scales. These parameters include means and totals of current forest area and volume, as well as components of change such as means and totals of growth and harvest removals. Over the last several decades, there has been a steadily increasing demand for estimates for smaller geographic areas and/or for finer temporal resolutions. However, the current sampling intensities of many NFI and the reliance on design-based estimators often leads to inadequate precision of estimates at these scales. This research focuses on improving the precision of forest removal estimates both in terms of spatial and temporal resolution through the use of small area estimation techniques (SAE). In this application, a Landsat-derived tree cover loss product and the information from mill surveys were used as auxiliary data for area-level SAE. Results from the southeastern US suggest improvements in precision can be realized when using NFI data to make estimates at relatively fine spatial and temporal scales. Specifically, the estimated precision of removal volume estimates by species group and size class was improved when SAE methods were employed over post-stratified, design-based estimates alone. The findings of this research have broad implications for NFI analysts or users interested in providing estimates with increased precision at finer scales than those generally supported by post-stratified estimators.
- Estimating tree canopy cover using harmonic regression coefficients derived from multitemporal Landsat dataDerwin, Jill M.; Thomas, Valerie A.; Wynne, Randolph H.; Coulston, John W.; Liknes, Greg C.; Bender, Stacie; Blinn, Christine E.; Brooks, Evan B.; Ruefenacht, Bonnie; Benton, Robert; Finco, Mark V.; Megown, Kevin (2020-04)The goal of this study was to evaluate whether harmonic regression coefficients derived using all available cloud free observations in a given Landsat pixel for a three-year period can be used to estimate tree canopy cover (TCC), and whether models developed using harmonic regression coefficients as predictor variables are better than models developed using median composite predictor variables, the previous operational standard for the National Land Cover Database (NLCD). The two study areas in the conterminous USA were as follows: West (Oregon), bounded by Landsat Worldwide Reference System 2 (WRS-2) paths/rows 43/30, 44/30, and 45/30; and South (Georgia/South Carolina), bounded by WRS-2 paths/rows 16/37, 17/37, and 18/37. Plot-specific tree canopy cover (the response variable) was collected by experienced interpreters using a dot grid overlaid on 1 m spatial resolution National Agricultural Imagery Program (NAIP) images at two different times per region, circa 2010 and circa 2014. Random forest model comparisons (using 500 independent model runs for each comparison) revealed the following (1) harmonic regression coefficients (one harmonic) are better predictors for every time/region of TCC than median composite focal means and standard deviations (across times/regions, mean increase in pseudo R-2 of 6.7% and mean decrease in RMSE of 1.7% TCC) and (2) harmonic regression coefficients (one harmonic, from NDVI, SWIR1, and SWIR2), when added to the full suite of median composite and terrain variables used for the NLCD 2011 product, improve the quality of TCC models for every time/region (mean increase in pseudo R-2 of 3.6% and mean decrease in RMSE of 1.0% TCC). The harmonic regression NDVI constant was always one of the top four most important predictors across times/regions, and is more correlated with TCC than the NDVI median composite focal mean. Eigen analysis revealed that there is little to no additional information in the full suite of predictor variables (47 bands) when compared to the harmonic regression coefficients alone (using NDVI, SWIR1, and SWIR2; 9 bands), a finding echoed by both model fit statistics and the resulting maps. We conclude that harmonic regression coefficients derived from Landsat (or, by extension, other comparable earth resource satellite data) can be used to map TCC, either alone or in combination with other TCC-related variables.
- Evaluating and improvement of tree stump volume prediction models in the eastern United StatesBarker, Ethan Jefferson (Virginia Tech, 2017-06-06)Forests are considered among the best carbon stocks on the planet. After forest harvest, the residual tree stumps persist on the site for years after harvest continuing to store carbon. A bigger concern is that the component ratio method requires a way to get stump volume to obtain total tree aboveground biomass. Therefore, the stump volumes contribute to the National Carbon Inventory. Agencies and organizations that are concerned with carbon accounting would benefit from an improved method for predicting tree stump volume. In this work, many model forms are evaluated for their accuracy in predicting stump volume. Stump profile and stump volume predictions were among the types of estimates done here for both outside and inside bark measurements. Fitting previously used models to a larger data set allows for improved regression coefficients and potentially more flexible and accurate models. The data set was compiled from a large selection of legacy data as well as some newly collected field measurements. Analysis was conducted for thirty of the most numerous tree species in the eastern United States as well as provide an improved method for inside and outside bark stump volume estimation.
- Fourier Series Applications in Multitemporal Remote Sensing Analysis using Landsat DataBrooks, Evan B. (Virginia Tech, 2013-06-27)Researchers now have unprecedented access to free Landsat data, enabling detailed monitoring of the Earth's land surface and vegetation. There are gaps in the data, due in part to cloud cover. The gaps are aperiodic and localized, forcing any detailed multitemporal analysis based on Landsat data to compensate. Harmonic regression approximates Landsat data for any point in time with minimal training images and reduced storage requirements. In two study areas in North Carolina, USA, harmonic regression approaches were least as good at simulating missing data as STAR-FM for images from 2001. Harmonic regression had an R^2"0.9 over three quarters of all pixels. It gave the highest R_Predicted^2 values on two thirds of the pixels. Applying harmonic regression with the same number of harmonics to consecutive years yielded an improved fit, R^2"0.99 for most pixels. We next demonstrate a change detection method based on exponentially weighted moving average (EWMA) charts of harmonic residuals. In the process, a data-driven cloud filter is created, enabling use of partially clouded data. The approach is shown capable of detecting thins and subtle forest degradations in Alabama, USA, considerably finer than the Landsat spatial resolution in an on-the-fly fashion, with new images easily incorporated into the algorithm. EWMA detection accurately showed the location, timing, and magnitude of 85% of known harvests in the study area, verified by aerial imagery. We use harmonic regression to improve the precision of dynamic forest parameter estimates, generating a robust time series of vegetation index values. These values are classified into strata maps in Alabama, USA, depicting regions of similar growth potential. These maps are applied to Forest Service Forest Inventory and Analysis (FIA) plots, generating post-stratified estimates of static and dynamic forest parameters. Improvements to efficiency for all parameters were such that a comparable random sample would require at least 20% more sampling units, with the improvement for the growth parameter requiring a 50% increase. These applications demonstrate the utility of harmonic regression for Landsat data. They suggest further applications in environmental monitoring and improved estimation of landscape parameters, critical to improving large-scale models of ecosystems and climate effects.
- Growth, Removals, and Management IntensityWynne, Randolph H.; Thomas, Valerie A.; Bender, Stacie; Brooks, Evan B.; Coulston, John W.; Derwin, Jill M.; Gopalakrishnan, Ranjith; Green, Patrick; Harding, David; Sumnall, Matthew; Joshi, Pratik; Ranson, Jon; Schleeweis, Karen; Thomas, R. Quinn; Yang, Zhiqiang (2019-05-01)
- Improved accuracy of aboveground biomass and carbon estimates for live trees in forests of the eastern United StatesRadtke, Philip J.; Walker, David; Frank, Jereme; Weiskittel, Aaron R.; DeYoung, Clara; MacFarlane, David W.; Domke, Grant M.; Woodall, Christopher W.; Coulston, John W.; Westfall, James A. (2017-01)Accurate estimation of forest biomass and carbon stocks at regional to national scales is a key requirement in determining terrestrial carbon sources and sinks on United States (US) forest lands. To that end, comprehensive assessment and testing of alternative volume and biomass models were conducted for individual tree models employed in the component ratio method (CRM) currently used in the US' National Greenhouse Gas Inventory. The CRM applies species-specific stem volume equations along with specific gravity conversions and component expansion factors to ensure consistency between predicted stem volumes and weights, and additivity of predicted live tree component weights to match aboveground biomass (AGB). Data from over 76 600 stem volumes and 6600 AGB observations were compiled from individual studies conducted in the past 115 years - what we refer to as legacy data - to perform the assessment. Scenarios formulated to incrementally replace constituent equations in the CRM with models fitted to legacy data were tested using cross-validation methods, and estimates of AGB were scaled using forest inventory data to compare across 33 states in the eastern US. Modifications all indicated that the CRM in its present formulation underestimates AGB in eastern forests, with the range of underestimation ranging from 6.2 to 17 per cent. Cross-validation results indicated the greatest reductions in estimation bias and root-mean squared error could be achieved by scenarios that replaced stem volume, sapling AGB, and component ratio equations in the CRM. A change in the definitions used in apportioning biomass to aboveground components was also shown to increase prediction accuracy. Adopting modifications tested here would increase AGB estimates for the eastern US by 15 per cent, accounting for 1.5 Pg of C currently unaccounted for in live tree aboveground forest C stock assessments. Expansion of the legacy data set currently underway should be useful for further testing, such as whether similar gains in accuracy can be achieved in estimates of regional or national-scale C sequestration rates.
- Increased Precision in County-Level Volume Estimates in the United States National Forest Inventory With Area-Level Small Area EstimationCao, Qianqian; Dettman, Garret, T.; Radtke, Philip J.; Coulston, John W.; Derwin, Jill; Thomas, Valerie A.; Burkhart, Harold E.; Wynne, Randolph H. (Frontiers Media, 2022-04-26)Many National Forest Inventory (NFI) stakeholders would benefit from accurate estimates at finer geographic scales than most currently implemented in operational estimates using NFI sample data. In the past decade small area estimation techniques have been shown to increase precision in forest inventory estimates by combining field observations and remote-sensing.We sought to demonstrate the potential for improving the precision of forest inventory growing stock volume estimates for counties in United States of North Carolina, Tennessee, and Virginia, by pairing canopy height models from digital aerial photogrammetry (DAP) and field plot data from the United States NFI. Area-level Fay-Herriot estimators were used to avoid the need for precise (GPS) coordinates of field plots. Reductions in standard errors averaging 30% for North Carolina county estimates were observed, with 19% average reductions in standard errors in both Tennessee and Virginia. Accounting for spatial autocorrelation among adjacent counties provided further gains in precision when the three states were treated as a single forest land population; however, analyses conducted one state at a time showed that good results could be achieved without accounting for spatial autocorrelation. Apparent gains in sample sizes ranged from about 65% in Virginia to 128% in North Carolina, compared to the current number of inventory plots. Results should allow for determining whether acquisition of statewide DAP would be costeffective as a means for increasing the accuracy of county-level forest volume estimates in the United States NFI.
- Large-area forest assessment and monitoring using disparate lidar datasetsGopalakrishnan, Ranjith (Virginia Tech, 2017-02-24)In the past 15 years, a large amount of public-domain lidar data has been collected over the Southeastern United States. Most of these acquisitions were undertaken by government agencies, primarily for non-forestry purposes. That is, they were collected mostly to aid in the creation of digital terrain models and to support hydrological and engineering assessments. Such data is not ideal for forestry purposes mainly due to the low pulse density per square meter, the high scan angles and low swath overlaps associated with these acquisitions. Nevertheless, the large area of coverage involved motivated this work. In this dissertation, I first look at how such lidar data (from non-forestry acquisitions) can be combined with National Forest Inventory tree height data to generate a large-area canopy height model. A simple linear regression model was developed using two lidar-based metrics as predictors: the 85th percentile of heights of canopy first returns and the coefficient of variation of the heights of canopy first returns. This model had good predictive ability over 76 disparate lidar projects, covering an area of approximately 297,000 square kilometers between them. Factors leading to the residual lack-of-fit of the model were also analyzed and quantified. For example, predictive ability was found to be better for softwood forests, forests with more homogeneous vegetation structure and for terrains with gentler slopes. Given that as much as 30% of the US is covered by public domain non-forestry lidar acquisitions, this is a first step for constructing a national wall-to-wall vertical vegetation structure map, which can then be used to ask important questions regarding forest inventories, carbon sequestration, wildlife habitat suitability and fire risk mitigation. Then, I examined whether such lidar data could be further used to predict understory shrub presence over disparate forest types. The predictability of classification model was low (accuracy = 62%, kappa = 0.23). Canopy occlusion factors and the heterogeneity of the understory layer were implicated as the main reasons for this poor performance. An analysis of the metrics chosen by the modeling framework highlighted the importance of non-understory metrics (metrics related to canopy openness and topographic aspect) in influencing shrub presence. As the proposed set of metrics were developed over a wide range of temperate forest types and topographic conditions of Southeastern US, it is expected that it will be useful for more localized future studies. Lastly, I explored the possibility of combining lidar-derived canopy height maps with Landsat-derived stand-age maps to predict plantation pine site index over large areas (site index is a measure of forest productivity). The model performance was assessed using a Monte Carlo technique (RMSE = 3.8 meters, relative RMSE = 19%). A sample site index map for large areas of Virginia and South Carolina was generated (map coverage area: 832 sq. km) and implications were discussed. Analysis of the resulting map revealed the following: (1) there is an increase in site index in most areas, compared to the 1970s, and (2) approximately 83% of the area surveyed had low levels of productivity (defined as site index < 22.0 meters for base age of 25 years). This work highlights the efficacy of combining lidar-based canopy height maps with other similar remote sensing based datasets to understand aspects of forest productivity over large areas, and to help make policy-relevant recommendations.
- Methods for the spatial modeling and evalution of tree canopy coverDatsko, Jill Marie (Virginia Tech, 2022-05-24)Tree canopy cover is an essential measure of forest health and productivity, which is widely studied due to its relevance to many disciplines. For example, declining tree canopy cover can be an indicator of forest health, insect infestation, or disease. This dissertation consists of three studies, focused on the spatial modeling and evaluation of tree canopy cover, drawing on recent developments and best practices in the fields of remote sensing, data collection, and statistical analysis.newlinenewline The first study evaluates how well harmonic regression variables derived at the pixel-level using a time-series of all available Landsat images predict values of tree canopy cover. Harmonic regression works to approximate the reflectance curve of a given band across time. Therefore the coefficients that result from the harmonic regression model estimate relate to the phenology of the area of each pixel. We use a time-series of all available cloud-free observations in each Landsat pixel for NDVI, SWIR1 and SWIR2 bands to obtain harmonic regression coefficients for each variable and then use those coefficients to estimate tree canopy cover at two discrete points in time. This study compares models estimated using these harmonic regression coefficients to those estimated using Landsat median composite imagery, and combined models. We show that (1) harmonic regression coefficients that use a single harmonic coefficient provided the best quality models, (2) harmonic regression coefficients from Landsat-derived NDVI, SWIR1, and SWIR2 bands improve the quality of tree canopy cover models when added to the full suite of median composite variables, (3) the harmonic regression constant for the NDVI time-series is an important variable across models, and (4) there is little to no additional information in the full suite of predictors compared to the harmonic regression coefficients alone based on the information criterion provided by principal components analysis. The second study presented evaluates the use of crowdsourcing with Amazon's Mechanical Turk platform to obtain photointerpretated tree canopy cover data. We collected multiple interpretations at each plot from both crowd and expert interpreters, and sampled these data using a Monte Carlo framework to estimate a classification model predicting the "reliability" of each crowd interpretation using expert interpretations as a benchmark, and identified the most important variables in estimating this reliability. The results show low agreement between crowd and expert groups, as well as between individual experts. We found that variables related to fatigue had the most bearing on the "reliability" of crowd interpretations followed by whether the interpreter used false color or natural color composite imagery during interpretation. Recommendations for further study and future implementations of crowdsourced photointerpretation are also provided. In the final study, we explored sampling methods for the purpose of model validation. We evaluated a method of stratified random sampling with optimal allocation using measures of prediction uncertainty derived from random forest regression models by comparing the accuracy and precision of estimates from samples drawn using this method to estimates from samples drawn using other common sampling protocols using three large, simulated datasets as case studies. We further tested the effect of reduced sample sizes on one of these datasets and demonstrated a method to report the accuracy of continuous models for domains that are either regionally constrained or numerically defined based on other variables or the modeled quantity itself. We show that stratified random sampling with optimal allocation provides the most precise estimates of the mean of the reference Y and the RMSE of the population. We also demonstrate that all sampling methods provide reasonably accurate estimates on average. Additionally we show that, as sample sizes are increased with each sampling method, the precision generally increases, eventually reaching a level of convergence where gains in estimate precision from adding additional samples would be marginal.
- Near-term investments in forest management support long-term carbon sequestration capacity in forests of the United StatesCoulston, John W.; Domke, Grant M.; Walker, David M.; Brooks, Evan B.; O'Dea, Claire B. (Oxford University Press, 2023-11-21)The forest carbon sink of the United States offsets emissions in other sectors. Recently passed US laws include important climate legislation for wildfire reduction, forest restoration, and forest planting. In this study, we examine how wildfire reduction strategies and planting might alter the forest carbon sink. Our results suggest that wildfire reduction strategies reduce carbon sequestration potential in the near term but provide a longer term benefit. Planting initiatives increase carbon sequestration but at levels that do not offset lost sequestration from wildfire reduction strategies. We conclude that recent legislation may increase near-term carbon emissions due to fuel treatments and reduced wildfire frequency and intensity, and expand long-term US carbon sink strength.
- Not All Biomass is Created Equal: An Assessment of Social and Biophysical Factors Constraining Wood Availability in VirginiaBraff, Pamela Hope (Virginia Tech, 2014-05-19)Most estimates of wood supply do not reflect the true availability of wood resources. The availability of wood resources ultimately depends on collective wood harvesting decisions across the landscape. Both social and biophysical constraints impact harvesting decisions and thus the availability of wood resources. While most constraints do not completely inhibit harvesting, they may significantly reduce the probability of harvest. Realistic assessments of woody availability and distribution are needed for effective forest management and planning. This study focuses on predicting the probability of harvest at forested FIA plot locations in Virginia. Classification and regression trees, conditional inferences trees, random forest, balanced random forest, conditional random forest, and logistic regression models were built to predict harvest as a function of social and biophysical availability constraints. All of the models were evaluated and compared to identify important variables constraining harvest, predict future harvests, and estimate the available wood supply. Variables related to population and resource quality seem to be the best predictors of future harvest. The balanced random forest and logistic regressions models are recommended for predicting future harvests. The balanced random forest model is the best predictor, while the logistic regression model can be most easily shared and replicated. Both models were applied to predict harvest at recently measured FIA plots. Based on the probability of harvest, we estimate that between 2012 and 2017, 10 – 21 percent of total wood volume on timberland will be available for harvesting.
- A novel application of small area estimation in loblolly pine forest inventoryGreen, P. Corey; Burkhart, Harold E.; Coulston, John W.; Radtke, Philip J. (2020-04)Loblolly pine (Pinus taeda L.) is one of the most widely planted tree species globally. As the reliability of estimating forest characteristics such as volume, biomass and carbon becomes more important, the necessary resources available for assessment are often insufficient to meet desired confidence levels. Small area estimation (SAE) methods were investigated for their potential to improve the precision of volume estimates in loblolly pine plantations aged 9-43. Area-level SAE models that included lidar height percentiles and stand thinning status as auxiliary information were developed to test whether precision gains could be achieved. Models that utilized both forms of auxiliary data provided larger gains in precision compared to using lidar alone. Unit-level SAE models were found to offer additional gains compared with area-level models in some cases; however, area-level models that incorporated both lidar and thinning status performed nearly as well or better. Despite their potential gains in precision, unit-level models are more difficult to apply in practice due to the need for highly accurate, spatially defined sample units and the inability to incorporate certain area-level covariates. The results of this study are of interest to those looking to reduce the uncertainty of stand parameter estimates. With improved estimate precision, managers, stakeholders and policy makers can have more confidence in resource assessments for informed decisions.
- On-the-Fly Massively Multitemporal Change Detection Using Statistical Quality Control Charts and Landsat DataBrooks, Evan B.; Wynne, Randolph H.; Thomas, Valerie A.; Blinn, Christine E.; Coulston, John W. (Institute of Electrical and Electronics Engineers (IEEE), 2014-06)One challenge to implementing spectral change detection algorithms using multitemporal Landsat data is that key dates and periods are often missing from the record due to weather disturbances and lapses in continuous coverage. This paper presents a method that utilizes residuals from harmonic regression over years of Landsat data, in conjunction with statistical quality control charts, to signal subtle disturbances in vegetative cover. These charts are able to detect changes from both deforestation and subtler forest degradation and thinning. First, harmonic regression residuals are computed after fitting models to interannual training data. These residual time series are then subjected to Shewhart X-bar control charts and exponentially weighted moving average charts. The Shewhart X-bar charts are also utilized in the algorithm to generate a data-driven cloud filter, effectively removing clouds and cloud shadows on a location-specific basis. Disturbed pixels are indicated when the charts signal a deviation from data-driven control limits. The methods are applied to a collection of loblolly pine (Pinus taeda) stands in Alabama, USA. The results are compared with stands for which known thinning has occurred at known times. The method yielded an overall accuracy of 85%, with the particular result that it provided afforestation/deforestation maps on a per-image basis, producing new maps with each successive incorporated image. These maps matched very well with observed changes in aerial photography over the test period. Accordingly, the method is highly recommended for on-the-fly change detection, for changes in both land use and land management within a given land use.
- Prediction of Canopy Heights over a Large Region Using Heterogeneous Lidar Datasets: Efficacy and ChallengesGopalakrishnan, Ranjith; Thomas, Valerie A.; Coulston, John W.; Wynne, Randolph H. (MDPI, 2015-08-27)Generating accurate and unbiased wall-to-wall canopy height maps from airborne lidar data for large regions is useful to forest scientists and natural resource managers. However, mapping large areas often involves using lidar data from different projects, with varying acquisition parameters. In this work, we address the important question of whether one can accurately model canopy heights over large areas of the Southeastern US using a very heterogeneous dataset of small-footprint, discrete-return airborne lidar data (with 76 separate lidar projects). A unique aspect of this effort is the use of nationally uniform and extensive field data (~1800 forested plots) from the Forest Inventory and Analysis (FIA) program of the US Forest Service. Preliminary results are quite promising: Over all lidar projects, we observe a good correlation between the 85th percentile of lidar heights and field-measured height (r = 0.85). We construct a linear regression model to predict subplot-level dominant tree heights from distributional lidar metrics (R2 = 0.74, RMSE = 3.0 m, n = 1755). We also identify and quantify the importance of several factors (like heterogeneity of vegetation, point density, the predominance of hardwoods or softwoods, the average height of the forest stand, slope of the plot, and average scan angle of lidar acquisition) that influence the efficacy of predicting canopy heights from lidar data. For example, a subset of plots (coefficient of variation of vegetation heights <0.2) significantly reduces the RMSE of our model from 3.0–2.4 m (~20% reduction). We conclude that when all these elements are factored into consideration, combining data from disparate lidar projects does not preclude robust estimation of canopy heights.
- Producing a Canopy Height Map Over a Large Region Using Heterogeneous LIDAR DatasetsGopalakrishnan, Ranjith; Thomas, Valerie A.; Coulston, John W.; Wynne, Randolph H. (2014)Accurate and unbiased wall-to-wall canopy height maps for large regions are useful to forest scientists and managers for several reasons such as carbon accounting and wildfire fuel-load monitoring. Airborne lidar is establishing itself as the most promising technology for this. However, mapping large areas often involves using lidar data from different projects executed by different agencies, involving varying acquisition dates, sensors, pulse densities, etc. In this work, we address the important question of how accurately one can predict and model canopy heights over large areas of the Southeastern US using a heterogeneous lidar datasets (with more than 90 separate lidar projects). A unique aspect of this effort is the use of extensive and robust field data from the Forest Inventory and Analysis (FIA) program of the US Forest Service. We construct a simple linear model to predict canopy height at plots from distributional lidar metrics. Preliminary results are quite promising: over all lidar projects, we observe a correlation of 81.8% between the 95th percentile of lidar heights and field-measured height, with an RMSE of 3.66 meters (n=3078). We further estimated that ~1.21 m (33%) of this RMSE could be attributed to co-registration inaccuracies. The RMSE of 3.66 m compares quite well to previous efforts that used spaceborne lidar sensors to estimate canopy heights over large regions. We also identify and quantify the importance of several factors (like point density, the predominance of hardwoods or softwood) that also influence the efficacy of our prediction model.