Geology topics

Chengquan Huang

Publications and source records attributed to Chengquan Huang.

26 records · Page 2Linked to original sources

USGS2003

Predictive modeling of forest cover type and tree canopy height in the central Rocky Mountains of Utah

Maps of forest cover type and canopy height are needed for LANDFIRE, a multi-scale fire risk assessment project designed to generate intermediate-resolution data of vegetation and fire fuel characteristics for the U.S. Here we describe an evaluation study in the central Rockies of Utah, comparing tree-based methods, multivariate adaptive regression splines (MARS), and a hybrid method for mapping forest cover and canopy height on the basis of more than 2,000 forest inventory ground plots in the seven million ha mapping zone. The two forest attributes were modeled as functions of a variety of predictor variables, including: Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images acquired at three different seasons; Tasseled-cap brightness, greenness, and wetness; a forest type group map; and topographic variables derived from Digital Elevation Models (DEMs); and other ancillary variables. The hybrid modeling approach showed a marked increase in overall and within forest cover type accuracies, outperforming the tree-based and MARS approaches. Little difference was seen in global performance measures of forest canopy height models, but patterns in residual plots resulting from different modeling approaches raise questions about utility of height predictions in different applications.

Idaho, Utah, Wyoming↗

USGS2003

Deriving rangeland structural attributes using Landsat ETM+, ERS-1/ERS-2

The purpose of this study is to determine if Synthetic Aperture Radar (SAR) can be used independently, or in conjunction with Landsat Enhanced Thematic Mapper Plus (ETM+) to improve the classification accuracy of structural attributes of rangeland vegetation, particularly percent shrub cover and top shrub canopy height. Such information, if mapped accurately, can be used in models to better characterize fuel conditions and fire regimes, as well as to evaluate fire hazard status, called for by the U.S. National Fire Plan. The input datasets utilized in this investigation included eighteen bands of Landsat ETM+ path 38 / row 32 (three image dates, six bands each), backscattering and interferometic data derived from tandem ERS-1/2 SAR image pairs (C-band), and extensive field point data. The results showed the use of SAR data provided no significant improvement over the ETM+ data for estimating percent cover or shrub canopy height. The lack of improvement in classification accuracy is possibly due to the influence of topography on the radar backscattering signal. Additional results demonstrated improved model accuracies when a 3x3-averaging filter was applied to the eighteen bands of ETM+ imagery.

Conference Paper↗

USGS2003

Exploration of satellite-measured vegetation seasonality for Landfire land cover

The purpose of this study is to explore the use of satellite data and other sources of spatial data for large area classification in the western United States to support research on potential fire hazards. Extensive field information was made available to this project from two sources: Forest Inventory and Assessment (FIA) and Utah State University. Seasonal spectral patterns of reflectance generated for select vegetation communities indicated that substantial spectral changes occurred through the growing season for most land cover types. In many cases, pronounced spectral differences characterized different types of vegetation, indicating a high probability that classification will accurately separate these particular types of land cover. However, spectral similarities between other types of land cover, such as Douglas fir and white fir, indicate potential classification challenges. Results from this study also show that decision tree analysis is highly effective for assessing quality of input field data and for generating large area land cover classification data sets. It was found that a 5-7% improvement in classification results could be achieved simply by not using those field plots that appeared to be sub-optimal for classification purposes based on image interpretation.

Conference Paper↗

USGS2003

A strategy for mapping mid-scale existing vegetation in support of national fire fuel assessment

Geospatial distribution of natural vegetation is among the very important environmental parameters required for applications ranging from global climate change to monitoring of natural hazards, monitoring of ecosystem vitality, and fire management practices. Increasingly sophisticated applications require vegetation datasets to cover large areas at a suitable scale and provide sufficiently detailed information. In this paper, we describe a research effort to develop a remote sensing methodology capable of producing 30-meter resolution, wall-to-wall coverage of existing vegetation types and structure variables in support of a multi-agency fire fuels and fire risks assessment project. Success of this remote sensing research effort is dependent on improved sensor and data qualities, a thorough understanding of regional and local vegetation ecology, successful integration of remote sensing with a large amount of field plot data, and flexible mapping algorithms. Preliminary results produced in the Wasatch Range and Uinta Mountains of central Utah include 28 vegetation types with an overall accuracy of 60% (average by life forms), percent canopy density (sub-pixel density) of forest, shrub, and herbaceous cover (correlation coefficient of 89, 60, and 55% respectively), and average top canopy height of forest, shrub, and herbaceous cover (correlation coefficient of 73, 50, 20% respectively). Techniques to improve the first-round results are discussed, including refinements of mapping models and use of relevant environmental gradients and potential vegetation classification associated with actual vegetation types.

Conference Paper↗

USGS2002

Synergistic use of FIA plot data and Landsat 7 ETM+ images for large area forest mapping

FIA plot data were used to assist in classifying forest land cover from Landsat imagery and relevant ancillary data in two regions of the U.S.: one around the Chesapeake Bay area and the other around Utah. The overall accuracies for the forest/nonforest classification were over 90 percent and about 80 percent, respectively, in the two regions. The accuracies for deciduous/evergreen/mixed and forest type group classifications were around 80 percent and 65 percent, respectively, and were consistent in the two regions. These results suggest that use of FIA plot data together with satellite imagery and relevant ancillary data may substantially improve the efficiency, accuracy, and consistency of large area forest land cover mapping.

Conterminous United States↗

USGS2002

Development of a circa 2000 land cover database for the United States

Multi-Resolution Land Characterization 2000 (MRLC 2000) is a second-generation federal consortium to create an updated pool of nation-wide Landsat 7 imagery, and derive a second-generation National Land Cover Database (NLCD 2000). This multi-layer, multisource database will include a suite of 30-meter resolution data that will serve as standardized ingredients for the production of land cover – both nationally and locally. This database will also provide the framework to allow flexibility in developing and applying suites of independent data layers. These nationally standardized independent data layers or components, will be useful not only within the land-cover classification but as data themes for other applications. This database will consist of the following components: (1) normalized tasseled cap (TC) transformations of Landsat 7 imagery for three time periods per scene (early, peak and late), (2) ancillary data layers, including 30m DEM derivatives of slope, aspect and elevation and three STATSCO soil derivatives, (4) image shape and texture information, (5) image derivatives of percent imperviousness and percent tree canopy per-pixel, (6) classified land-cover data derived from the Tassel Capped imagery, ancillary data and derivatives, (7) classification rules and metadata from the land cover classification, allowing future users the potential to modify rules to derive land cover products tailored to their specific local applications. In a pilot study application of the database concept, two mapping zones (Utah and Virginia) were selected for full generation of the above data components. Three derivative layers including, per-pixel imperviousness, per-pixel canopy and land cover were classified from the database. Cross validation accuracies for land cover ranged from 65-82%, and mean absolute error values of 10-15% were reported for percent tree canopy and imperviousness.

conterminous United States↗

USGS2001

A Landsat 7 scene selection strategy for a national land cover database

A strategy for selecting Landsat 7 ETM+ imagery for development of a new generation national land cover database of the United States has been developed. This strategy is formulated to target Landsat 7 ETM+ scenes based on land cover and land use, vegetation phenology and image quality (cloudiness, haze). Criteria based on phenology and scene quality provide a national baseline for acquiring Landsat 7 data. Optimal time periods for discriminating land cover types were identified for each Landsat 7 path-row footprint and each proposed land cover mapping zone (mosaic of several path-rows based on landscape and ecoregion), from which three Landsat scenes were selected. This database of selected scenes is used to guide Landsat 7 data purchasing. This methodology provides a consistent framework for populating Landsat 7 imagery to be used for a new national land cover characterization initiative.

Conference Paper↗

USGS2001

A strategy for estimating tree canopy density using Landsat 7 ETM+ and high resolution images over large areas

Forest cover is of great interest to a variety of scientific and land management applications, many of which require not only information on forest categories, but also tree canopy density. In previous studies, large area tree canopy density had been estimated at spatial resolutions of 1km or coarser using coarse resolution satellite images. In this study, a strategy is developed for estimating tree canopy density at a spatial resolution of 30 m. This strategy is based on empirical relationships between tree canopy density and Landsat data, established using linear regression and regression tree techniques. One-meter digital orthophoto quadrangles were used to derive reference tree canopy density data needed for calibrating the relationships between canopy density and Landsat spectral data. This strategy was tested over three areas of the United States. In general, models derived using both linear regression and regression tree techniques were statistically significant. The regression tree was found more robust than linear regression, primary due to its capability of approximating complex non-linear relationships using a set of linear equations. This strategy will be recommended for use in developing a nation wide tree canopy density data set at a 30 m resolution as part of the Multi-Resolution Land Characteristics 2000 project.

Conference Paper↗