Progress 04/01/02 to 03/31/08
Outputs
OUTPUTS: At the completion of the project, the following items have been achieved: 1) the completion of the Arc Macro Language (AML) program that relates forest inventory data to other environmental factors in geospatial context, 2) the development of a remote sensing model to identify forest landscape changes over time, 3) the completion of the AML program for thinning practice at individual tree level, 4) the integration of multispectral satellite imagery and lidar derived data for land cover/land use classification, 5) the creation of a GIS database model for forest inventory and 6) the comparison on timber volume between field measurement and lidar derived estimation. This project demonstrated the use of geospatial technology in forestry including GIS, GPS, and remote sensing. With the increasing analytical efficiencies, forest inventory and assessment can be achieved with higher accuracy, but lower cost. Forest management can be planned with higher precision and flexibility since the information can be attained more rapidly. PARTICIPANTS: Jason Raines, Graduate Assistant of College of Forestry and Agriculture at Stephen F. Austin State University. Jeff Williams, GIS System Administrator of College of Forestry and Agriculture at Stephen F. Austin State University. Jason Grogan, Geospatial Data Acquisition Specialist II of Forest Resources Institute at Stephen F. Austin State University. Ron Thill, Team Leader of Wildlife Habitat & Silviculture Lab, USFS Southern Research Station. Jim Stevens, Manager of East Texas Plant Materials Center, USDA Natural Resources Conservation Service. TARGET AUDIENCES: The results of this project are intended to bring attention to the forestry research community, forestry industry, as well as forest landowners. It demonstrated alternative approaches to assess forest resources at different scales, both spatially and temporally. With a forestry database updated constantly, it also supports decision makers with quality information so that the entire community can benefit. PROJECT MODIFICATIONS: Not relevant to this project.
Impacts
This project has demonstrated a cost and time effective approach to estimate forest resources. Through multispectral and multitemporal remote sensing, forest inventory and assessment can be achieved in a more efficient, frequent, and affordable way. Facilitated with lidar remote sensing and GPS technology, precision forest management can be expected in the near future.
Publications
- Hung, I., Williams, J.M., Kroll, J.C., and Unger, D.R. 2004. Forest landscape changes in east Texas from 1974 to 2002. In: Proceedings of the 4th Southern Forestry and Natural Resources GIS Conference, Athens, Georgia.
- Siska, P.P. and Hung, I. 2004. Advanced digital terrain analysis using roughness-dissectivity parameters in GIS. In: Proceedings of the 24th Annual ESRI International User Conference, San Diego, California.
- Hung, I., McNally, B.C., Farrish, K.W., and Oswald, B.P. 2005. Using GIS for selecting tress for thinning. In: Proceedings of the 25th Annual ESRI International User Conference, San Diego, California.
- Siska, P.P., Goovaerts P., Hung I. and V.M. Bryant. 2005. Predicting ordinary kriging errors caused by surface roughness and dissectivity. Earth Surface Processes and Landforms, 30(5):601-612.
- Unger, D., Kroll J., Hung, I., Williams, J., Coble, D., and Grogan, J. 2008. A standardized, cost-effective, and repeatable remote sensing methodology to quantify forested resources in Texas. Southern Journal of Applied Forestry 32(1):12-20
- Raines, J.C., Grogan, J., Hung, I., and Kroll, J. 2008. Multitemporal analysis using Landsat Thematic Mapper (TM) bands for forest cover classification in east Texas. Southern Journal of Applied Forestry 32(1):21-27
|
Progress 01/01/07 to 12/31/07
Outputs
OUTPUTS: The main focus of this period is to estimate forest measurement through remote sensing. Both multispectral imagery and lidar (light detection and ranging) data were used, separately and combined. The intent is to achieve forest inventory with higher accuracy, but lower cost. Multispectral satellite imagery has been used for land cover type classification for decades. By using appropriate classification algorithm, a forest type map can be attained with acceptable accuracy. The map can be updated whenever a new scene of imagery is available. It provides the tool for monitoring forest status. With the advancement of sensor capability, imagery of higher spectral, spatial, and temporal resolution has become more available and affordable. In the meantime, more classification techniques have been introduced such as object-oriented vs. traditional pixel-based classification and sub-pixel vs. traditional pixel-level classification. All of these give foresters more alternatives in
measuring forest remotely. However, no matter what imagery and algorithm used, the accuracy of forest cover map remains unknown until ground measurement is carried out. Thus, a well planned ground truthing for accuracy assessment is required. A classification map derived from multispectral image provides the information of spatial distribution and acreage of forest by different cover types. It does not tell the timber volume on the ground, which is what foresters are most interested. Lidar remote sensing, a technique analyzing objects in 3D space, sheds some lights in this aspect. An airborne lidar system flown upon forest landscape collects surface information based upon laser reflectance from ground objects. A lidar dataset, usually referred to as "lidar point cloud" is composed of numerous points with the x-, y- and z-coordinate of each point recorded. Data derived from lidar points can be used to measure forest attributes at individual tree level (total height and crown diameter,
etc.) or stand level (volume, basal area and biomass, etc.) Through the fusion of multispectral imagery and lidar derived data, answers are expected to questions of not only where the forests are but also how much they are. In this period, an area of 2,590 acre of the Stephen F. Austin Experimental Forest in east Texas was flown upon for lidar and digital aerial photo data acquisition. Two scenes of satellite image, IKONOS and QuickBird, for the same area were also acquired. Deriving forest measurement from lidar data was tested using both commercially available software packages (TIFFS and Lidar Analyst) and self developed program written in Python language. Forty eight sampling plots for ground truthing were setup throughout the study area. The development of a program for segmenting lidar point cloud data was initiated. The intent is to convert lidar points into the volumetric pixel structure of raster for the fusion with multispectral imagery. In turn, the composite image will be
used for better quantifying forest resources.
PARTICIPANTS: James Kroll, Director of Columbia Regional Geospatial Service Center, Stephen F. Austin State University. Minho Park, Assistant Professor, Computer Science Department, Stephen F. Austin State University, Nacogdoches, Texas. Jeff Williams, System Administrator, College of Forestry and Agriculture, Stephen F. Austin State University. Jason Raines, Graduate Assistant, 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.
Impacts
This project combines traditional forest cover type delineation through image classification and timber measurement through more recent lidar remote sensing techniques. In the same spatial context, both questions: "Where are they?" and "How much are there?" are expected to be answered. It should provide foresters more reliable and updated information for decision making.
Publications
- Bowes, C., Unger, D., Farrish, K., and Hung, I. 2007. Using remotely sensed data to quantify contaminated brine sites in southwest Texas. In Proceedings of 21st Biennial Workshop on Aerial Photography, Videography, and High Resolution Digital Imagery for Resource Assessment, May 15-17, 2007, Terre Haute, Indiana.
|
Progress 01/01/06 to 12/31/06
Outputs
Two major aspects of the project were accomplished during this period. They are: a) the creation of a GIS database model for forest inventory, and b) the continuation of integrating digital imagery and lidar derived data for forest type classification. Since the turn of the century, GIS database has been moving from georelational data model to object-based data model. Incorporating relational database management system in GIS, data can be stored in geodatabases instead of file system that allows for multiple-user editing, versioning and history checking. A GIS database model designed for forest inventory consolidates information into one centralized location, where spatial information such as individual tree location, plot boundary, stand boundary etc. and the associated attributes such as species, DBH, tree height, timber volume etc. are integrated. The forest inventory geodatabase is the permanent storage for the information of a dynamic forest. Data can be checked
out for update in the field using mobile GIS facilitated with high accuracy GPS. In this period, the geodatabase model was created in a desktop GIS. It was checked out and tested in the field. A manual for this operation was also prepared. With the forest inventory geodatabase being up-to-date, monitoring and predicting timber growth becomes more reliable. The other aspect of this period is the continuation of forest type classification using remote sensing. The focus of this period was to use different combinations of bands from multispectral images at different times for the same study area. Preliminary results showed that more bands used for classification does not necessarily increase accuracy. Once determined, the best scenario of classification will be integrated with lidar derived data to test for even higher accuracy. As soon as forest type classification is achieved with confidence, other environmental parameters such as soils, hydrology, topography etc. will be included in
the geodatabase model mentioned above. With all of the factors integrated in a GIS, better forest inventory and monitoring can be expected.
Impacts
This project consolidates forest inventory information into relational database management system with spatial component in a GIS. It helps for better monitoring forest status over time. With timely updates and the capability of history tracking, better management decision can be made.
Publications
- Siska, P.P., I. Hung and V. M. Bryant. 2006. Correlation between Pollen Dispersion and Forest Spatial Distribution Patterns in the Southeastern United States. Paper in the Proceedings of the 5th Southern Forestry and Natural Resources GIS Conference, Asheville, North Carolina.
|
Progress 01/01/05 to 12/31/05
Outputs
Two major aspects of the project were accomplished during this period. They are: a) the completion of the Arc Macro Language (AML) program for thinning practice at individual tree level, and b) the integration of multispectral satellite imagery and lidar derived data for land cover/land use classification. Thinning is the silvicultural practice of adjusting the number and arrangements of trees in a stand to promote an increase in growth and productivity. It removes trees within a stand to regulate the level of site occupancy and subsequent stand development. In order to determine the outcome of a thinning practice before trees are cut, a GIS program was developed to simulate the post-thinning trees spacing based upon parameters assigned by the planner. The program is written in AML and applies a moving circular quadrat system from one tree location to another. In GIS, with the location of each individual tree and its associated attributes including diameter and height
available, the program is able to identify those trees to remove so that after thinning an evenly distributed basal area can be achieved resulting the stand stock desired by the user. This accomplishment was presented as a paper at the 25th Annual ESRI International User Conference. Since GIS applications are moving from georelational data model to object-based data model, the next step is to migrate this program from AML to scripting language environment. That will benefit not only performance enhancement but also data interoperability. The other aspect of this period is a series of land cover/land use classification using remote sensing data. Landsat multispectral satellite imagery has been widely used for forest type classification and its accuracy can be assessed. With the ever increasing availability of high spatial resolution imagery, its appropriateness in forestry should be tested. In addition, the ever evolving lidar technology makes it possible to depict the 3-demesional
structure of forest stand. During this period, different combinations of data sources mentioned above were used for image classification by employing supervised classification methodology including maximum likelihood and artificial neural network. Accuracy assessment will be done by comparing each classification map to relevant aerial photography on randomly selected locations. The optimum combinations of data sources and classification methods will be determined through kappa statistics and z-test.
Impacts
This project applies GIS and remote sensing for forest management at different scales. Combining field measurement in forest inventory and periodically acquired remote sensing data, it allows natural resource managers to monitor the environment dynamically. With programs designed for specific forest practices, timber yield and harvest planning can be implemented within the GIS model.
Publications
- I-Kuai Hung, Benjamin C. McNally, Kenneth W. Farrish, and Brian P. Oswald. 2005. Using GIS for Selecting Tress for Thinning. In: Proceedings of the 25th Annual ESRI International User Conference, San Diego, California.
- Siska, P.P., Goovaerts P., Hung I. and V. M. Bryant. 2005. Predicting Ordinary Kriging Errors Caused by Surface Roughness and Dissectivity. Earth Surface Processes and Landforms, 30(5):601-612.
|
Progress 01/01/04 to 12/31/04
Outputs
There were three major aspects of the project accomplished during this period: a) the completion of the Arc Macro Language (AML) program that relates forest inventory data to other environmental factors in geospatial context, b) the development of a remote sensing model to identify forest landscape changes over time, and c) the initiative of a GIS model at individual tree level for thinning practice. The AML program is capable of analyzing a variety of parameters at the same time. By superimposing a moving window system across the study area, both vector and raster data can be extracted and related to one another. In addition to forest inventory data, tree height, basal area, and stand volume etc., other parameters were integrated. They included the spot elevation for terrain metrics, the National Hydrology Dataset (NHD) for stream network, and the thematic map of forest cover type derived from remote sensing data. The size of moving window was programmed to be
adjustable so that the Modifiable Areal Unit Problem (MAUP) can be taken into account. The ultimate function of this AML program was to investigate the relationship between timber production and its underlying properties of local environment. This accomplishment was presented as a paper at the 24th Annual ESRI International User Conference. The remote sensing model was designed to identify forest landscape changes over time by using historical satellite imagery. Based on the classified forest cover type map, landscape patches were identified and landscape metrics were calculated. They included patch size, aggregation of patches, and patch shape complexity. The results showed an overall decline of total forestland in the 1980s and a recovery in the 1990s in east Texas. The mean patch size of forest showed a trend of increase, whereas that of non-forest was consistently decreasing and the shape became more complex. The replanting efforts in this region has created buffers between land
development such as urban sprawl and ranching. The forest maintained the overall landscape contagion while non-forest land-use became more fragmented. This accomplishment was presented as a paper at the 4th Southern Forestry and Natural Resources GIS Conference. The last aspect of this period was a preliminary design of a decision making system for selecting trees for thinning in a GIS environment. A loblolly pine plantation was surveyed through a complete tree tally, recording the coordinates of each individual tree. The dataset will be processed in GIS by applying a moving quadrant system throughout the study area. In each quadrant, tree attributes including DBH, height, basal area, and density will be populated as determining factors for tree selection with a goal of equal distribution of trees across the stand. With the promising technology of LiDAR, individual tree location and its associated attributes could be recorded remotely. This tree selection system would assist in
decision making before ground crew move into a forest stand to regulate the level of site occupancy and promote subsequent stand development.
Impacts
This project integrated GIS and remote sensing into traditional forest practices. It reviewed the forest landscape changes over time, linked timber measurement to the environment in geospatial context, and developed tools for forest management. As the project progresses, a better understanding of forest landscape dynamics in time and in space could be expected.
Publications
- Hung, I., Williams, J. M., Kroll, J. C. and Unger, D. R. 2004. Forest landscape changes in east Texas from 1974 to 2002. In: Proceedings of the 4th Southern Forestry and Natural Resources GIS Conference, Athens, Georgia.
- Siska, P. P. and Hung, I. 2004. Advanced digital terrain analysis using roughness-dissectivity parameters in GIS. In: Proceedings of the 24th Annual ESRI International User Conference, San Diego, California.
- Siska, P.P., Goovaerts P., Hung I. and M. Bryant. 2004. Predicting the magnitude of kriging interpolation errors using terrain roughness and dissectivity parameters: multiple regression approach. Earth Surface Processes and Landforms. (in press)
|
Progress 01/01/03 to 12/31/03
Outputs
During this period the project focused on two aspects of forest inventory: a) developing a methodology to assess the local spatial variation of forest parameters in association with environmental properties. An aml program (Arc Macro Language) that was designed at the beginning of this project and new addition to this GIS program was developed during this last period. The program originally computed the statistical and newly developed dissectivity parameters in a moving window environment from spatially georeferenced point data sets in the GIS database. The newest addition to this program also supports the line data structure. As a result of this new addition, the program computes the drainage density in each moving window and the new dissectivity parameter based on TIN (triangular Irregular Network) is computed. This parameter represents a ratio between the surface sum of a...and its base. The next step is to enrich the program with three surface curvature
parameters. The capability of this aml program was used in developing regression models for studying the dependency of the kriging errors magnitude on surface dissectivity. The regression models were successfully established and the magnitude of kriging errors was predicted with 95% of the r-square value. The results were summarized in articles that were submitted to scientific journals and will be presented at conferences. The second focus was on developing the best predictive models for merchantable volume in East Texas. The ground measurements and remote sensing data were used to develop a spatial modeling procedure that would predict the timber volume in cubic feet per acre in unsampled locations in the East Texas studied area. The indicator kriging was applied to ground measurements and a probability map was developed based on 700 cubic feet per acre cutoff value. Later, remote sensing data were added to the analysis and thirty equiprobable realizations of volume estimates were
produced using sequential gaussian simulation. The final map of e-type estimates indicated a summary of 30 realizations that predicted the merchantable volume in the East Texas region.
Impacts
The project developed a new program that enables GIS users to compute statistical, environmental and surface parameters that were not available before. During this project a new methodology for analyzing spatial data was developed that will enhance the research in natural resource management, GIS and forestry.
Publications
- Siska, P. P., and Pierre Goovaerts. (2004) Predicting the Magnitude of Ordinary Kriging Errors Using Surface Dissectivity Parameters and Multiple Regression. Earth Surface Processes and Landforms (accepted, under revision).
- Siska, P.P., Goovaerts, P. and I-Kuai Hung (2004). Assessment of Kriging Errors and Terrain Modeling Using Moving Windows Technique. Computers and Geosciences (currently under review).
- Eriksson, M. and P. P. Siska (2003). Replay to Marcotte's comments. Mathematical Geology Vol 1. No. 1. 130-135.
|
Progress 01/01/02 to 12/31/02
Outputs
The first year of this research focused on two major objectives: a) investigating LIDAR measurements as a potential tool for forest inventory, and b) modeling and prediction of interpolation errors for the purpose of achieving the higher accuracy of final maps. These maps will be used for representing the spatial distribution of biomass, volume, basal area and additional forest parameters from the ground measurements. The investigation based on the laser data indicated that profiling lasers yield estimates of biometric parameters with reasonable r-square values and are appropriate for large forested areas such as the East Texas region. The classical ground measurements are extremely time consuming and costly. On the other hand the implementation of satellite images is often biased by individual approach in interpretation of the remote sensing data. The lasers however, can yield to reliable measurements of forest parameters once the biometric equations are established
between the ground and laser data. The major advantage of the laser measurements is in the subsequent data collection and analysis that do not require new field measurements to estimate biomass and all forest parameters of interest. A program in the visual basic was developed that automatically reads the raw LIDAR data into GIS. This supports rapid generation of spatial layers based on point data. The method of data input to GIS is based on COM (component object model). This current investigation about using LIDAR data as a new possibility for forest measurements lead to another proposal in cooperation with NASA scientists. This new proposal should complete the study objectives (Forest Inventory) using only LIDAR data in East Texas and will be a parallel project. The second phase of this project was dedicated to the newly proposed GIS analysis of point data sets and to evaluating interpolation errors that emanate from developing continuous surfaces in GIS. This method is based on the
moving window strategy that consists of sets of overlapping regular rectangles moving across the entire study area. Inside each moving window statistical parameters are computed including newly developed parameters. These parameters are three dissectivity indices that capture the surface diversity and the gradient of change. The objective of this procedure is to study the relationship between the interpolation errors and the parameters of surface diversity and to develop a linear regression model predicting the magnitude of errors during the interpolation procedure. The practical application of this research is in creating continuous surfaces in GIS. Forest inventory measurements are essentially point data sets. In order to construct a map in GIS from inventory data interpolation procedures must be used and the quality of this surface evaluated.
Impacts
Geographic Information Science (GIS) is a new technology that significantly contributes to natural resources. In this project GIS is applied to forest inventory and analysis to evaluate spatial distribution and structure of forest parameters.
Publications
- Siska, P.P. and Hung, I-Kuai. 2003. (Under review). The influence of surface roughness and dissectivity on the magnitude of kriging errors. International Journal of Geographic Information Science.
- Siska, P.P. and Bhowmick, A. 2002. Laser based measurements for evaluating natural resources. SWAAG Conference, Laredo, TX.
|
|