ASPRS

PE&RS August 2003

VOLUME 69, NUMBER 8
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
JOURNAL OF THE AMERICAN SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING

Peer-Reviewed Articles

873 Geometric Processing of Ikonos Stereo Imagery for Coastal Mapping Applications
Kaichang Di, Ruijin Ma, and Rongxing Li

Abstract  Download Full Article
The results of photogrammetric mapping of a Lake Erie coastal area from 1-m-resolution Ikonos Geo stereo images are presented. The nominal accuracy of ground point de-termination with vendor-provided Rational Function (RF) coefficients is evaluated and systematic errors are found when compared with ground control points. A significant improvement in the accuracy is achieved by a refining process that applies a three-dimensional affine transformation to the RF-calculated 3D ground points to correct the systematic errors. A DEM is automatically generated by a chain of processes: area-based image matching, ground point calculation, outlier elimination, TIN construction, and interpolation. Following DEM generation, an orthoimage is produced using the DEM and the refined geometric model. Accuracies of the DEM and the orthoimage as assessed from independent checkpoints (ICPs) are approximately 2 m in planimetry and 3 m in height. Finally, a 3D shoreline is extracted through manual digitization in one image of a stereo pair and then automatic matching in the other image. Issues concerning the production of digital tide-coordinated shorelines using instantaneous observations are also discussed.

881 Fast Approximation of Visibility Dominance Using Topographic Features as Targets and the Associated Uncertainty
Sanjay Rana

Abstract  Download Full Article
An approach to reduce visibility index computation time and measure the associated uncertainty in terrain visibility analyses is presented. It is demonstrated that the visibility index computation time in mountainous terrain can be reduced substantially, without any significant information loss, if the line of sight from each observer on the terrain is drawn only to the fundamental topographic features, i.e., peaks, pits, passes, ridges, and channels. However, the selected sampling of targets results in an underestimation of the visibility index of each observer. Two simple methods based on iterative comparisons between the real visibility indices and the estimated visibility indices have been proposed for a preliminary assessment of this uncertainty. The method has been demonstrated for gridded digital elevation models.

889 Mapping Multiple Variables for Predicting Soil Loss by Geostatistical Methods with TM Images and a Slope Map
Guangxing Wang, George Gertner, Shoufan Fang, and Alan B. Anderson

Abstract  Download Full Article
Soil erosion is widely predicted as a function of six input factors, including rainfall erosivity, soil erodibility, slope length, slope steepness, cover management, and support practice. Because of the multiple factors, their interactions, and their spatial and temporal variability, accurately mapping the factors and further soil loss is very difficult. This paper compares two geostatistical methods and a traditional stratification to map the factors and to estimate soil loss. Soil loss is estimated by integrating a sample ground data set, TM images, and a slope map. The geostatistical methods include collocated cokriging and a joint sequential co-simulation model. With both geostatistical methods, local estimates and variances at any location where the factors and soil loss are unknown can be computed. The results showed that the two geostatistical methods performed significantly better than traditional stratification in terms of overall and spatially explicit estimates. Further more, the cokriging led to higher accuracy of mean estimates than did the co-simulation, while the latter provided decision makers with reliable uncertainties of the local estimates as useful information to assess risk when making decisions based on the prediction maps.

899 AVHRR-Based Spectral Vegetation Index for Quantitative Assessment of Vegetaton State and Productivity: Calibration and Validation
Felix Kogan, Anatoly Gitelson, Edige Zakarin, Lev Spivak, and Lubov Lebed

Abstract  Download Full Article
The goal of the work was to estimate, quantitatively, vegetation state and productivity using AVHRR-based Vegetation Condition Index (VCI). The VCI algorithm includes application of post-launch calibration to visible channels, calculation of NDVI from channels' reflectance, removal of high-frequency noise from NDVI's annual time series, stratification of ecosystem resources, and separation of ecosystem and weather components in the NDVI value. The weather component was calculated by normalizing the NDVI to the difference of the extreme NDVI fluctuations (maximum and minimum), derived from multi-year data for each week and land pixel. The VCI was compared with wheat density measured in Kazakhstan. Six test fields were located in different climatic (annual precipitation 150 to 700 mm) and ecological (semi-desert to steppe-forest) zones with elevations from 200 to 700 m and a wide range of NDVI variation over space and season from 0.05 to 0.47. Plant density (PD) was measured in wheat fields by calculating the number of stems per unit area. PD deviation from year to year (PDD) was expressed as a deviation from median density calculated from multi-year data. The correlation between PDD and VCI for all stations was positive and quite strong (r² > 0.75) with the Standard Errors of Estimates (SEE) of PDD less than 16 percent; for individual stations, the SEE was less than 11 percent. The results indicate that VCI is an appropriate index for monitoring weather impact on vegetation and for assessment of pasture and crop productivity in Kazakhstan. Because satellite observations provide better spatial and temporal coverage, the VCI-based system will provide efficient tools for management of water resources and the improvement of agricultural planning. This system will serve as a prototype in the other parts of the world where ground observations are limited or not available.

907 An Explicit Index for Assessing the Accuracy of Cover-Class Areas
Guofan Shao, Wenchun We, Gang Wu, Xinhua Zhou, and Jianguo Wu

Abstract  Download Full Article
We present a new index, called Relative Errors of Area (REA), for assessing the accuracy of cover-class areal percentage (%LAND) that is extracted from thematic maps after classifying remotely sensed data. We demonstrate how to derive REA from an error matrix and its relationship with user's and producer's accuracy. We compare the REA index with other accuracy indices in a hypothetical and two real case studies. The accuracy of cover-class areal estimates is highly correlated with the REA index, but not with other classification accuracy indices such as the overall classification accuracy. In general, users should beware of using thematic maps with low REA values. Moreover, the estimates of cover-class area can be revised by using REA if cell values of the major diagonal in an error matrix are available.

915 Potential of Digital Color Imagery for Censusing Haleakala Silverswords in Hawaii
Rick E. Landenberger, James B. McGraw, Timothy A. Warner, and Tomas Brandtberg

Abstract  Download Full Article
Spatially explicit, high spatial resolution remotely sensed imagery offers a largely untapped potential for censusing and monitoring rare plant populations that exist in remote, exposed environments. Using digital color imagery acquired over the Haleakala Crater on Maui, Hawai'i, we evaluated the accuracy of photointerpretation and automated censuses by imaging nine silversword census plots characterized by individuals of known size, life cycle status, and location. Due to spatial resolution limitations, both methods tended to omit small individuals, but omissions varied by size class and type of omission. Omission rates were low for demographically important medium and large plants; however, the automated method often failed to segment and census tightly clustered plants. The photointerpreter commission error rate was lower than that of the automated method, and both methods tended to overestimate mean silversword size. These data outline the issues and challenges that will likely emerge as spatially explicit, high spatial resolution aerial censuses become more common.
Top Home