Progress 10/01/08 to 09/30/14
Outputs
Target Audience: Target audiences of this project include those who are interested in natural resource management. During this period, they ranged from students and faculty of forestry and nature resources schools, forest professionals, forest landowners, and government agencies. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided? This project has provided tremendous opportunities for both our undergraduate and graduate students. All of the tasks completed involved students either collecting data in the field or analyzing data in the lab. We have two graduate students who completed their research that is related to this project. With this project, we were able to present our works in professional meetings and shared experiences and knowledge with others. That also enhanced our teaching where are able to keep up to date with the technology. How have the results been disseminated to communities of interest? Findings from this project have been presented as paper or poster in conferences. Some have been published as journal papers. Each of them can be found in the Products section with detailed information. For the use of Pictometry, we also hosted a two-day workshop at our GIS Lab of Stephen F. Austin State University, December 11-12. What do you plan to do during the next reporting period to accomplish the goals? This is the final report for our project. In order to continue to stay up to date in technology for better managing forests and natural resources of east Texas, we have submitted a new project titled “Geospatial Technologies for Forest Resources Management in East Texas” for consideration.
Impacts
What was accomplished under these goals?
In order to meet the overall goal of assessing and monitoring forest resources in east Texas using GIS/GPS and remote sensing, we completed the following tasks. We derived forest measurement from LiDAR (Light Detection and Ranging) data for the Stephen F. Austin Experimental Forest and assessed its accuracy through field sampling. The assessment was conducted at the FIA (USDA Forest Inventory and Analysis) plot and sub-plot levels. The comparison included the geographic location of individual trees, tree height, crown diameter, the finally estimated stand volume. We applied sub-pixel classification algorithm to classify Landsat satellite imagery to a forest cover type map with the classes of pine, hardwood, mixed forest, and non-forest in the Stephen F. Austin Experimental Forest. In contrast to traditional per-pixel classification, the sub-pixel classification allowed us to derive the proportion of each forest cover type component within each pixel area. We introduced a new technique for measuring height through the Pictometry web-based interface. Compared with DBH (diameter at breast height) tree height measurement is hard to accomplish in the field and the accuracy is often problematic. With the Pictometry web-based interface, height measurement can be done with relative ease and it accuracy was found even higher than traditional height measurement using clinometer or laser range finder. All of the data collected have been built into a geodatabase that can be accessed by multiple users. This dataset will be integrated into a much larger database of our East Texas Forest Inventory project. We are in the process of developing a web map with the goal of allowing general public to browse the map and estimate for forest cover type composition and timber volume of their own area of interest.
Publications
- Type:
Journal Articles
Status:
Accepted
Year Published:
2014
Citation:
Unger, D. R., D. L. Kulhavy, J. Williams, D. Creech, and I. Hung. 2014. Urban Tree Height Assessment Using Pictometry Hyperspatial 4-inch Multispectral Imagery. Journal of Forestry, 112.
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Unger, D. R., I. Hung, and D. L. Kulhavy. 2014. Comparing Remotely Sensed Pictometry Web-based Height Estimates with in situ Clinometer and Laser Range Finder Height Estimates. Journal of Applied Remote Sensing, Vol. 8, No. 1, P. 083590.
- Type:
Journal Articles
Status:
Published
Year Published:
2014
Citation:
Unger, D. R., D. L. Kulhavy, I. Hung and Y. Zhang. 2014. Quantifying Natural Resources Using Field-Based Instruction and Hands-on Applications. Journal of Studies in Education, Vol. 4, No. 2, P. 1-14.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2013
Citation:
Unger, D.R., M.A. Wade, D.L. Kulhavy, J. Williams, D. Creech, and I-K. Hung. 2013. Evaluating Tree Height Using Pictometry Hyperspatial Imagery versus Traditional Measurements. Paper presentation at Society of American Foresters 2013 National Convention, October 23-27, North Charleston, South Carolina.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2013
Citation:
Unger, D.R., M.A. Wade, D.L. Kulhavy, J. Williams, D. Creech, I-K. Hung, and Y. Zhang. 2013. Measuring Tree Height Using Pictometry Hyperspatial 4-inch Multispectral Imagery. Paper presentation at 9th Southern Forestry and Natural Resource Management GIS Conference, December 8-10, 2013, Athens, Georgia.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2012
Citation:
Westbrook, J., I. Hung, D. Unger, and Y. Zhang. 2012. Sub-Pixel Classification of Forest Cover Types in East Texas. Poster presentation at ESRI International User Conference, July 23-27, 2012, San Diego, California.
- Type:
Theses/Dissertations
Status:
Other
Year Published:
2012
Citation:
Westbrook, J. 2012. Sub-Pixel Classification of Forest Cover Types in East Texas. Thesis for MS in Spatial Science, Stephen F. Austin State University, Nacogdoches, Texas.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2011
Citation:
Westbrook, J., I. Hung, D. Unger, and Y. Zhang. 2011. Sub-pixel Classification of Forest Cover Types. Presentation at the Eighth Southern Forestry and Natural Resource Management GIS Conference, Dec. 11-13, 2011, Athens, Georgia.
- Type:
Journal Articles
Status:
Published
Year Published:
2010
Citation:
Chapman, J., I. Hung, and J. Tippen. 2010. Evaluating Tiffs (Toolbox for LiDAR Data Filtering and Forest Studies) in Deriving Forest Measurements from LiDAR Data. International Journal of Mathematical and Computational Forestry & Natural-Recourses Sciences, Vol. 2, No. 2, P. 145-152
- Type:
Theses/Dissertations
Status:
Other
Year Published:
2010
Citation:
Evaluation of LiDAR Derived Estimates of Forest Measurement. Thesis for MS in Spatial Science, Stephen F. Austin State University, Nacogdoches, Texas.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2009
Citation:
Chapman, J., I. Hung, and J. Tippen. 2009. Evaluation of Lidar Derived Forest Measurement. In proceedings of 7th Southern Forestry and Natural Resources GIS Conference, Dec. 7-9, 2009, Athens, Georgia.
|
Progress 01/01/13 to 09/30/13
Outputs
Target Audience: Target audiences of this project include those who are interested in natural resource management. During this period, they ranged from students and faculty of forestry and nature resources schools, forestry professionals, forest landowners, and government agencies. Changes/Problems: Nothing significant to report during this reporting period. What opportunities for training and professional development has the project provided? Compared with conventional timber cruising, height measurement through Pictometry significantly reduces time and cost needed for timber volume estimation while resulting in highly accurate outcomes. We conducted several accurate assessment projects involving our graduate and undergraduate students that provided them opportunities working both in the lab and in the field. Furthermore, we teamed up withour local county office, organizing a two-day workshop for the use of Pictometry web applications. How have the results been disseminated to communities of interest? As a cooperative effort with the local county office, the use of Pictometry has been considered for management issues such as urban forestry, planning, inventory, and appraisal of the land. We also presented our project findings in professional conferences including theSociety of American Foresters National Convention and theSouthern Forestry and Natural Resources GIS Conference. What do you plan to do during the next reporting period to accomplish the goals? The next report period will be the final stage of the project. We will review all of the results we have achieved in the past reporting periods and add new components for the dissemination of the findings to ensure that our overall goals are met.
Impacts
What was accomplished under these goals?
In this period, we focused on tree height measurements using remote sensing. Conventionally, timber volume measurement relies on timber cruising in the field. Although tree diameters can be measured accurately in the field, tree heights are hard to measure precisely when the observer isstanding on the forest floor and looking up to the canopy. Therefore, tree height measurements are usually sampled and interpolated with some degree of uncertainty. With the Pictometry data, a multi-aspect digital aerial photography system combined with a digital elevation model, we are able to measure tree heights on thecomputer screen with ahigh level of accuracy.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2013
Citation:
Unger, D.R., M.A. Wade, D.L. Kulhavy, J. Williams, D. Creech, and I-K. Hung. 2013. Evaluating Tree Height Using Pictometry Hyperspatial Imagery versus Traditional Measurements. Paper presentation at Society of American Foresters 2013 National Convention, October 23-27, North Charleston, South Carolina.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2013
Citation:
Unger, D.R., M.A. Wade, D.L. Kulhavy, J. Williams, D. Creech, I-K. Hung, and Y. Zhang. 2013. Measuring Tree Height Using Pictometry Hyperspatial 4-inch Multispectral Imagery. Paper presentation at 9th Southern Forestry and Natural Resource Management GIS Conference, December 8-10, 2013, Athens, Georgia.
|
Progress 01/01/12 to 12/31/12
Outputs
OUTPUTS: Remote sensing has been used to assess the biomass and monitor the status of forestlands of large areas. In this period, we completed the sub-pixel classification to classify satellite imagery to forest cover type maps with the classes of pine, hardwood, mixed forest, and non-forest in Stephen F. Austin Experimental Forest, east Texas. Compared with traditional per-pixel multispectral classifiers which assumes materials within a pixel are homogenous, the sub-pixel classification allows for the un-mixing of pixels to show the proportion of each material of interest. In our study, percentage cover type maps, each for a class including pine, hardwood, mixed forest, and non-forest, were generated. The accuracy for each classified map was also assessed. PARTICIPANTS: Joey Westbrook, Graduate Assistant, College of Forestry and Agriculture, Stephen F. Austin State University. Yanli Zhang, Assistant Professor, College of Forestry and Agriculture, Stephen F. Austin State University. Dan Unger, Associate Professor, College of Forestry and Agriculture, Stephen F. Austin State University. TARGET AUDIENCES: Target audiences of this project include those who are interested in natural resource management. They range from students and faculty of forestry and nature resources schools, forest landowners, consulting firms, government agencies, to non-profit organizations. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
The sub-pixel classification was done through different combinations of using high resolution aerial photography or field measured data as training and/or reference data. A fuzzy error matrix was constructed for accuracy assessment. Results showed that using the aerial photo for both training and reference data achieved the highest overall accuracy (65%). While the traditional pixel-based classification requires 85% accuracy as acceptable, the acceptance level for sub-pixel classifications is not yet clearly defined. However, sub-pixel classification does provide an alternative that breaks a pixel area from a pure cover type into different components and calculates their proportions. When combined with field sample data, or other data sources such as LiDAR, the sub-pixel forest cover type classification will enable precise timber estimation based on a cover type map.
Publications
- Westbrook, J., I. Hung, D. Unger, and Y. Zhang. 2012. Sub-Pixel Classification of Forest Cover Types in East Texas. Poster presentation at ESRI International User Conference, July 23-27, 2012, San Diego, California.
|
Progress 01/01/11 to 12/31/11
Outputs
OUTPUTS: Remote sensing is an efficient way to assess the biomass and monitor the status of forestlands especially on large areas. In this period, we applied sub-pixel classification algorithm to classify satellite imagery to a forest cover type map with the classes of pine, hardwood, mixed forest, and non-forest in Stephen F. Austin Experimental Forest, east Texas. Instead of classifying each pixel of the image as one pure forest type, which is the outcome of traditional per-pixel multispectral classifiers, sub-pixel classification allows for the un-mixing of pixels to show the proportion of each material of interest. In our case, we were able to derive the percentage of each forest type component, including pine, hardwood, mixed forest, and non-forest, within each pixel area. The following accuracy assessment will compare the results of sub-pixel classification to those of traditional per-pixel classification, and to field measured data in order to evaluate its efficacy. PARTICIPANTS: Joey Westbrook, Graduate Assistant, College of Forestry and Agriculture, Stephen F. Austin State University. Yanli Zhang, Assistant Professor, College of Forestry and Agriculture, Stephen F. Austin State University. Dan Unger, Associate Professor, College of Forestry and Agriculture, Stephen F. Austin State University. TARGET AUDIENCES: Target audiences of this project include those who are interested in natural resource management. They range from students and faculty of forestry and nature resources schools, forest landowners, consulting firms, government agencies, to non-profit organizations. PROJECT MODIFICATIONS: Not relevant to this project.
Impacts
Unless it is a pure stand through intensive management, a forestland often comprises of different tree species while some of them are dominant. When estimating timber volume through remote sensing, traditional per-pixel classification classifies a pixel as one forest type but not others, even though in reality it might be a mix of different types within the pixel area. Sub-pixel classification provides an alternative that breaks a pixel area of pure cover type into different components and calculates their proportions. When combined with field sample data, other data sources such as LiDAR, the sub-pixel forest cover type classification will enable the more precise timber estimation based on a cover type map.
Publications
- Westbrook, J., I. Hung, D. Unger, and Y. Zhang. 2011. Sub-pixel classification of forest cover types. Presentation at the Eighth Southern Forestry and Natural Resource Management GIS Conference, Dec. 11-13, 2011, Athens, Georgia.
|
Progress 01/01/10 to 12/31/10
Outputs
OUTPUTS: In this period, we continued on the accuracy assessment of LiDAR (Light Detection and Ranging) derived forest measurements, which were compared to traditional field measurements within the Stephen F. Austin Experimental Forest in east Texas. We evaluated three off-the-shelf LiDAR data processing programs in terms of their capability in estimating forest measurements including number of trees, tree diameter at breast height (DBH), tree height, crown radius, and timber volume. The three software programs examined were TiFFS (Toolbox for LiDAR Data Filtering and Forest Studies), TreeVaW (Tree Variable Window), and LiDAR Analyst 4.2. The reference data used for accuracy assessment were collected in the previous year by following the sampling scheme outlined by the USDA Forest Inventory and Analysis (FIA) program. The comparison between LiDAR derived and field measured data was conducted at both plot and sub-plot level. Statistics calculated between the two included root mean square error (RMSE), percent error, and correlation coefficient for each of the forest measurements. An analysis of variance (ANOVA) on the absolute errors of each software program was also conducted to determine if there is any statistically significant difference among the programs tested. PARTICIPANTS: John Chapman, Graduate Assistant, College of Forestry and Agriculture, Stephen F. Austin State University. Jeff Williams, System Administrator, College of Forestry and Agriculture, Stephen F. Austin State University. Jeff Tippen, Senior Project Engineer, SURDEX Corp. Houston Office. TARGET AUDIENCES: Target audiences of this project include those who are interested in natural resource management. They range from students and faculty of forestry and nature resources schools, forest landowners, consulting firms, government agencies, to non-profit organizations. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
TreeVaW performed well compared to the other software programs in regards to absolute error as the measure for accuracy. However, its linear correlation between the LiDAR-derived and field-measures was generally low, indicating that the software did not follow the field-measured changes between plots. TreeVaW is not designed for commercial use and takes a preprocessed canopy height model as the only input instead of the raw LiDAR point data. TiFFS was able to estimate tree height with fair accuracy and had high correlation values, but drastically overestimated tree count and stand volume while underestimating tree crown radius. Its main strength is its generation of surface models reflecting elevation and canopy height. Unfortunately, TiFFS does not allow the user to adjust parameters in order to overcome the over-segmentation of the canopy height model. LiDAR Analyst was found to be difficult to use and faulty at identifying trees, making stand volume almost impossible to predict. However, for those trees it did detect, the algorithm for delineating its features seems to work well, except for DBH. All these programs generally tend to perform better in pine forests, due to the uniformity of crowns. Overall, these LiDAR processing software programs need to be better at identifying individual trees if they are to be used to estimate stand volume. All of the programs tested in this study are currently available to foresters as options for estimating forest measurements. Cost, accuracy, ease of use, and management goals should be considered when assessing which of these programs to use. With new algorithms being implemented and studies being put into practice for evaluating their performance, keeping up to date with the latest software development is important. Even though studies have shown promising results that LiDAR can be used for delineating individual trees and estimating forest properties with satisfactory accuracy, the performance of an algorithm might vary from one forest type to another, or from one region to another. As commercially available LiDAR data processing programs, it should be made clear what types of landscapes and forest types will work best with the software. In the meantime, allowing for the calibration of estimates in conjunction with field-measured training data would increase the accuracy. If a forest manager is to choose a LiDAR data processing software program for operational purposes, the ability to fine-tune the results is a criterion that should be considered in addition to the cost of the software.
Publications
- Chapman, J., I. Hung, and J. Tippen. 2010. Evaluating Tiffs (Toolbox for LiDAR Data Filtering and Forest Studies) in deriving forest measurements from LiDAR data. International Journal of Mathematical and Computational Forestry & Natural-Recourses Sciences, Vol. 2, No. 2, P. 145-152
|
Progress 01/01/09 to 12/31/09
Outputs
OUTPUTS: In this period, we performed the derivation of forest measurements from LiDAR (Light Detection and Ranging) data and compared it to the filed measured data of the same area. We completed the field measurement of the sampling plots located within the Stephen F. Austin Experimental Forest, part of the Angelina Nation Forest in Texas. The sampling scheme followed the method outlined by the USDA Forest Inventory and Analysis (FIA) so that the data collected are more comparable with other studies. Within the plots, individual tree diameter at breast height (DBH), height, and crown width at four cardinal directions were measured. Each tree's geographic coordinate was attained with a sub-meter accuracy GPS unit. This field measurement dataset is to serve as the reference for accuracy assessment on those measurements derived from LiDAR remote sensing. Forest characterization of the study area was also derived through LiDAR data processing using a software application, TiFFS (Toolbox for LiDAR Data Filtering and Forest Studies). The derived measurements include individual tree coordinate, height, and crown width. They were compared to the field measured data for accuracy. The comparison was conducted at both plot and sub-plot level. PARTICIPANTS: John Chapman, Graduate Assistant, College of Forestry and Agriculture, Stephen F. Austin State University. Jeff Williams, System Administrator, College of Forestry and Agriculture, Stephen F. Austin State University. Jeff Tippen, Senior Project Engineer, SURDEX Corp. Houston Office. TARGET AUDIENCES: Target audiences of this project include those who are interested in natural resource management. They range from students and faculty of forestry and nature resources schools, forest landowners, consulting firms, government agencies, to non-profit organizations. PROJECT MODIFICATIONS: Not relevant to this project.
Impacts
The preliminary results showed that LiDAR was able to detect individual tree height with satisfactory accuracy. Even though LiDAR tends to overestimate tree height when compared with field-measured data, the relationship stays consistent with a correlation coefficient of 0.8223. With some calibration based on test data measured in the field, LiDAR is a reliable tool in detecting individual tree height. LiDAR also tents to overestimate the number of trees and underestimate the crown width. It was caused by the fact that the LiDAR application assumes every single peak in the canopy surface a tree top instead of a branch of a tree. This will result in detecting more trees, each with smaller diameter. If stand volume is to calculate and commercial value to estimate, there is still way to go before LiDAR is widely available in real-world operation. The 3-dimensional forest structure varies from place to place with different species composition. Fusing LiDAR data with multispectral imagery could aid in delineating different forest structures and attain higher accuracy. Ideally, a commercially available LiDAR data processing application for forestry should allow for the calibration with field-measured test data and other parameters that are local to the study area. The outcomes of this period give forestry professionals an insight in terms of where LiDAR technology is standing for real-world forestry applications. As more LiDAR and other remote sensing applications become available, we will continue on researching its accuracy and efficacy.
Publications
- Chapman, J., I. Hung, and J. Tippen. 2009. Evaluation of lidar derived forest measurement. In proceedings of 7th Southern Forestry and Natural Resources GIS Conference, Dec. 7-9, 2009, Athens, Georgia.
|
|