Showing papers in "Photogrammetric Engineering and Remote Sensing in 1997"

Fusion of satellite images of different spatial resolutions: Assessing the quality of resulting images

Abstract: Methods have been proposed to produce multispectral images with enhanced spatial resolution using one or more images of the same scene of better spatial resolution. Assuming that the main concern of the user is the quality of the transformation of the multispectral content when increasing the spatial resolution, this paper defines the properties of such enhanced multispectral images. It then proposes both a formal approach and some criteria to provide a quantitative assessment of the spectral quality of these products. Five sets of criteria are defined. They measure the pe$ormance of a method to synthesize the radiometry in a single spectral band as well as the multispectral information when increasing the spatial resolution. The influence of the type of landscape present in the scene upon the assessment of the quality is underlined, as well as its dependence with scale. The whole approach is illustrated by the case of a SPOT image and three different standard methods to enhance the spatial resolution.

Mapping City Lights With Nighttime Data from the DMSP Operational Linescan System

Abstract: The Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) has a unique capability to detect low levels of visible and near-infrared (VNIR) radiance at night. With the OLS "VIS" band data, it is possible to detect clouds illuminated by moonlight, plus lights from cities, towns, industrial sites, gas pares, and ephemeral events such as fires and lightning illuminated clouds. This paper presents methods which have been developed for detecting and geolocating VNIR emission sources with nighttime DMSP-OLS data and the analysis of image time series to identify spatially stable emissions from cities, towns, and industrial sites. Results are presented for the United States.

A Comparison of Nighttime Satellite Imagery and Population Density for the Continental United States

Abstract: The striking apparent correlation between nighttime satellite imagery and human population density was explored for the continental United States. The nighttime stable-lights imagery was derived from the visible near-ZR band of 231 orbits of the Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS). The population density data were generated from a gridded vector dataset of the 1992 United States census block group polygons. Both datasets are at a one-square-kilometre resolution. The two images were co-registered and correlation between them was measured at a range of spatial scales, including aggregation to state and county levels. DMSP imagery showed strong correlations at aggregate scales, and analysis of the saturated areas of the images showed strong correlations between the areas of saturated clusters and the populations those areas cover. The non-zero pixels of the DMSP imagery correspond to only 10 percent of the land cover yet account for over 80 percent of the continental United States population. Spatial analysis of the clusters of the saturated pixels predicts population with an R2 of 0.63. Consequently, the DMSP imagery may prove to be useful to inform a "smart interpolation" program to improve maps and datasets of human population distributions in areas of the world where good census data may not be available or do not exist.

Using Thematic Mapper Data to Identify Contrasting Soil Plains and Tillage Practices

Abstract: Landsat-5 Thematic Mapper (TM) data from 11 May 1990 for Seneca County, Ohio were used to develop TM-based probability models for classifying agricultural management practices and soil properties. Logistic regression techniques were used to relate field data for 27 fields to TM bands and indices. Field data, including tillage and drainage practices, soil plain, and soil texture, were collected from 1988 to 1990. Both soil plain and tillage logistic regression models classified 89 percent of the fields correctly. Simple ratio and normalized differences of i"M bands 5 and 7 proved most useful for classifying tillage practices. TM bands 1, 2, 3, and 4 were found useful for identifying soil plain. Spectral differences were attributed to soil color differences between lake and till plain soils and surface residue differences between conservation and conventional tillage. The developed models were tested with independent data from 15 additional fields and classified 88 percent of the soil plain and 93 percent of the tillage attributes correctly. Using TM data to identify drainage practices, organic matter content, and soil texture was generally inadequate for scientific purposes.

Semi-Automatic Linear Feature Extraction by Dynamic Programming and LSB-Snakes

Abstract: This paper deals with semi-automatic linear feature extraction from digital images for GIS data capture, where the identification task is pe$ormed manually on a single image, while a special automatic digital module performs the high precision feature tracking in two-dimensional (2-0) image space or even three-dimensional (3-0) object space. A human operator identifies the object from an on-screen display of a digital image, selects the particular class this object belongs to, and provides a very few coarsely distributed seed points. subseq;ently, with th;?sk seed as an approximation of the ~osition and sham the linear feature will be extracted automatically by either a dynamic programming approach or by LSB-S~~~~S [Least-Squares E-spline Snakes). With dynamic programming, the optimization problem is set up as a discrete multistage decision process and is solved by a "timedelayed" algorithm. It ensures global optimality, is numerically stable, and allows for hard constraints to be enforced on the solution. In the least-squares approach, we combine three types of observation equations, one radiometric, formulating the matching of a generic object model with image data, and two that express the internal geometric constraints of a curve and the location of operator-given seed points. The solution is obtained by solving a pair of independent normal equations to estimate the parameters of the spline curve. Both techniques can be used in a monoplotting mode, which combines one image with its underlying DTM. The LSB-S~~~~S approach is also implemented in a multi-image mode, which uses multiple images simultaneously and provides for a robust and mathematically sound full 3D approach. These techniques are not restricted to aerial images. They can be applied to satellite and close-range images as well. The issues related to the mathematical modeling of the proposed methods are discussed and experimental results are shown in this paper too.

Radiometric Calibration of Landsat

Abstract: The radiometric calibration of the sensors on the Landsat se- ries of satellites is a contributing factor to the success of the Landsat data set. The calibration of these sensors has relied on the preflight laboratory work as well as on inflight tech- niques using on-board calibrators and vicarious techniques. Descriptions of these methods and systems are presented. Results of the on-board calibrators and reflectance-based, ground reference calibrations of Landsat 5 Thematic Mapper are presented that indicate the absolute radiometric calibra- tion of bands 1 to 4 should have an uncertainty of less than 5.0 percent. Bands 5 and 7 have slightly higher uncertainties, but should be less than 10 percent. The results also show that the on-board calibrators are of higher precision than the vicarious calibration but that the vicarious calibration results should have higher accuracy. Introduction The Landsat series of satellites provides the longest running continuous data set of high spatial-resolution imagery dating back to the launch of Landsat 1 in 1972. Part of the success of the Landsat program has been the ability to understand the radiometric properties of the sensors. This understanding has been due to the combination of prelaunch and post- launch efforts using laboratory, on-board, and vicarious cali- bration methods. The radiometric calibration of these systems helps characterize the operation of the sensors, but more importantly, the calibration allows the full Landsat data set to be used in a quantitative sense. A brief overview of the Landsat systems is given here, but the reader is directed to Engel and Weinstein (1983), Lansing and Cline (1975), Markham and Barker (1987), and Slater (1980) for detailed descriptions. The Landsat series of satellites can be viewed in two distinct parts. The first in- cludes Landsats 1, 2, and 3 that carried two sensor systems: the return beam vidicon (RBV) and the Multispectral Scanner (MSS) system. The RBv camera systems on Landsats 1 and 2 were multispectral with three cameras, while the system on Landsat 3 used only two cameras in a panchromatic mode. Landsats 1, 2, and 3 operated in a glg-krn, sun-synchronous orbit with an 18-day repeat cycle. The second phase of Land- sat includes Landsats 4 and 5. These platforms omitted the RBV cameras but still carried the Mss. These two platforms also carried the Thematic Mapper (TM), and their orbits were lowered to 705 krn with a 16-day repeat cycle. The MSS is a 6-bit, whiskbroom sensor with six detectors for each of its four bands. These bands are centered roughly at 0.55, 0.65, 0.75, and 0.85 pm (the MSS on Landsat 3 also had a fifth band between 10.4 and 12.6 pm). Bands 1 to 3 use photomultiplier tubes, while band 4 uses photodiodes. The MSS only collects data in one scan direction, and there is no compensation in the scan for the forward motion of the platform. At the end of every other scan, a rotating shutter and mirror assembly allows light from a calibration lamp to reach the detectors. The TM is also a whiskbroom system but it scans in both the forward and backward cross-track directions, and it cor- rects for the forward motion of the platform. In addition, the TM has 8-bit radiometric resolution and seven bands. Bands 1 to 5 and 7 each have 16 detectors with center wavelengths of roughly 0.49, 0.56, 0.66, 0.83, 1.67, and 2.24 Fm. Band 6 has four detectors and is centered around 11.5 pm. Bands 1 to 4 use silicon-based detectors, bands 5 and 7 use indium antimonide detectors, and band 6 uses mercury-cadmium-tel- luride detectors. Bands 5, 6, and 7 are part of the cold-focal plane that is cooled to 85°K through the use of a radiative cooler. The TM has an on-board calibration system composed of a shutter that oscillates rather than rotates and allows cali- bration data to be collected at the end of each scan. A great deal of research was done during the early days

Monitoring South Florida Wetlands Using ERS-1 SAR Imagery

Abstract: This study evaluates the utility of synthetic aperture radar (SAR) imagery collected by the ERS-1 satellite for monitoring wetland vegetation communities in southwestern Florida. Two images were analyzed, one collected at the end of the dry season in April 1994 and one collected at the end of the wet season in October 1994. The range of image intensity values from the different test sites varied by a factor of 6.2 (7.9 dB) on the dry season ERS-I SAR image and by a factor of 2.6 (4.1 dB) for the wet season ERS-1 SAR image. The results from the radar observations were found to be consistent with theoretical micro wave scattering models that predict variations in backscatter as a function of vegetation structure, soil moisture, surface roughness, and the presence or absence of standing water. Both the radar data and models show that, in wetlands dominated by herbaceous vegetation, the presence of standing water results in a decrease in backscatter. Conversely, in wetlands with woody plants (trees and shrubs), the radar data and models show that the presence of water results in an increase in backscatter. The results of this study illustrate that radar imagery is uniquely suited to detect and monitor changes in soil moisture, flooding, and aboveground biomass in these wetland ecosystems.

Performance of a neural network: mapping forests using GIS and remotely sensed data

Abstract: Neural networks have been proposed to classify remotely sensed and ancillary CIS data. In this paper, the backpropagation algorithm is critically evaluated, using as an example, the mapping of a eucalypt forest on the far south coast of New South Wales, Australia. A GIS database was combined with Landsat thematic mapper data, and 190 plots were field sampled in order to train the neural network model and to evaluate the resulting classifications. The results show that the neural network did not accurately classify GIS and remotely sensed data at the forest type level (Anderson Level III), though conventional classifiers also perjGorm poorly with this type of problem. Previous studies using neural networks have classified more general (e.g., Anderson Level I, II) landcover types at a higher accuracy than those obtained here, but mapped land cover into more general themes. Given the poor classification results and the difficulties associated with the setting up of suitable parameters for the neural-network (backpropagation) algorithm, it is concluded that the neuralnetwork approach does not offer significant advantages over conventional classification schemes for mapping eucalypt forests from Landsat TM and ancillary GIs data at the Anderson Level 111 forest type level.

Sensitivity of selected landscape pattern metrics to land-cover misclassification and differences in land-cover composition

Abstract: investigation of the sensitivity of these measurements to clasCalcuiation of landscape metrics from land-cover data is be- sification error (Hess, 1994). he objectives of this paper are coming increasingly common. Some studies have shown that two-fold: (1) to determine the sensitivity of landscape these measurements are sensitive to differences in land-cover metrics to land-cover misclassification, and (2) to determine composition, but none are known to have tested also their the sensitivity of landscape metrics to differences in landto land-cover misc~ass~ficat~o n~ error sirnula- scape condition. These objectives are necessarily connected. tion model was written to test the sensitivity of selected land- Ideally, landscape pattern metrics would be insensitive to scape pattern metrics to misclassification, and regression misclassification but sensitive to differences in land cover. analysis was used to determine if these metrics were signifi- Land-cover data, mapped from Landsat TM for the Chescantly related to differences in land-cover composition. Corn- apeake Bay Regional Watershed, were used for this study. parison of sensitivity and regression results suggests that The data were divided into 57 eight-digit U.S. Geological differences in land-cover composition need to be about 5 Survey (us~s) hydrologic units or watersheds. Sensitivity to percent greater than the misclassification rate to be confident misclassification is tested using a simulation model based on that differences in landscape metrjcs are not due to mjsclas- a published land-cover error matrix (Green et a]., 1993). Sensification. sitivity to differences in landscape condition is tested by comparison to the amount of human land use in the watershed. Landscape condition is measured as the ratio of an

Choosing an appropriate spatial resolution for remote sensing investigations

Abstract: Choosing rationally the spatial resolution for remote sensing requires a formal relation between the size of support and some measure of the information content. The local variance in the image has been used to help choose an appropriate spatial resolution. Here we choose spatial resolutions to map continuous variation in properties, such as biomass, using the variogram. The experimental variogram can be separated into components of underlying spatially dependent variation and measurement error. The spatially dependent component can be deregularized to a punctual support, and then regularized to any spatial resolution. The regularized variogram summarizes the information attainable by imaging at that spatial resolution because information exists in the relations between observations only. The investigator can use it to select a combination of spatial resolution and method of analysis for a given investigation. Two examples demonstrate the method.

Landform characterization with geographic information systems

Abstract: The ability to analyze and quantify morphology of the surface of the Earth in terms of landform characteristics is essential for understanding of the physical, chemical, and biological processes that occur within the landscape. However, because of the complexity of taxonomic schema for landforms which include their provenance, composition, and function, these features are difficult to map and quantify using automated methods. The author suggests geographic information systems (GIS) based methods for mapping and classification of the landscape surface into what can be understood as fourth-order-of-relief features and include convex areas and their crests, concave areas and their troughs, open concavities and enclosed basins, and horizontal and sloping flats. The features can then be analyzed statistically, aggregated into higher-order-of-relief forms, and correlated with other aspects of the environment to aid fuller classification of landforms.

Landscape Cover-Type Modeling Using a Multi-Scene Thematic Mapper Mosaic

Abstract: Landscape ecological applications of remotely sensed data are needed over increasingly larger areas and at finer spatial scales. Within the framework of the National Biological Service Gap Analysis program, 36 Utah cover types were modeled from a state-wide Landsat TM mosaic created from 24 scenes at 30-metre resolution (219,883 sq km). The state was subset into three ecoregions for classification, with cover-type association to spectral classes defined using a two-step modeling approach. Steps included post-classification correlation of 1,758 state-wide field training sites to spectral classes, and post-classification ancillary GIS modeling using ecological parameters of elevation, slope, aspect, and location to further refine spectral classes representing multiple cover types. Thirty-four of 36 cover-type classes were totally or partially identified using digital modeling, with five of 36 classes requiring both digital and analog methods. This methodology provides a framework to optimize landscape remote sensing cover-type modeling using a multiple scene mosaic.

Spectral reflectance with varying suspended sediment concentrations in clear and algae-laden waters.

Abstract: The objective of the study was to characterize and compare the relationship between suspended sediment concentration (SSC) and reflectance in clear and algae-laden waters. A controlled experiment was conducted outdoors in a 7510-litre water tank using natural sunlight. A red loam soil was added and suspended in the tank filled with clear and algae-laden waters, respectively. A total of 20 levels of SSC (from 25 to 500 mg I-*) were created for each type of treatment. Reflectance was recorded using an ASD spectroradiometer, and the bi-directional reflectance factor was computed and analyzed. The same amount of suspended sediment generated higher reflectance between 400 and 700 nm in clear water than in algae-laden water due to the blue and red absorption of chlorophyll. The effect of chlorophyll on the SSC-reflectance relationship was minimum at wavelengths between 700 and 900 nm. For both clear and algae-laden waters, the linearity in the SSC-reflectance relationship increased with wavelength between 400 and 900 nm. A near-linear relationship between ssc and reflectance was found between 720 and 900 nm.

A Machine-Learning Approach to Automated Knowledge-Base Building for Remote Sensing Image Analysis with GIs Data

Abstract: A machine learning approach to automated building of knowledge bases for image analysis expert systems incorporating GIS data is presented. The method uses an inductive learning algorithm to generate production rules from training data. With this method, building a knowledge base for a rule-based expert system is easier than using the conventional knowledge acquisition approach. The knowledge base built by this method was used by an expert system to pe$orm a wetland classification of Par Pond on the Savannah River Site, South Carolina using SPOT multispectral imagery and GIs data. To evaluate the peqformance of the resultant knowledge base, the classification result was compared to classifications with two conventional methods. The accuracy assessment and the analysis of the resultant production rules suggest that the knowledge base built by the machine learning method was of good quality for image analysis with GIS data.

Commercial implications of topographic terrain mapping using scanning airborne laser radar

Abstract: The airborne remote sensing industry is characterized by strict deadlines, tight budget constraints and an unrelenting demand for higher data densities with increased accuracy at lower cost. End-users, including everyone from urban GIs administrators to engineering firms doing traditional route planning, are demanding faster turnaround times and more accurate data. Therefore, aerial survey operators must look to more advanced technologies to reduce data processing times and field work expenses. Not limited by the environmental conditions restricting aerial photography, airborne laser scanning technology is emerging as an attractive alternative to the traditional techniques for large-scale geospatial data capture. The recent development of commercial laserbased topographic terrain mapping systems is driven by the availability of compact ruggedized solid state lasers, high precision airborne inertial navigation systems, and rugged precise highspeed scanners which combine to make high accuracy airborne scanning laser rangefinders practical. With the completion of the GPS and the availability of low-cost multichannel GPS receivers, these technologies enable the creation of affordable laser-based topographic terrain mapping systems. The purchase price of these commercial instruments has been reduced so that cost is no longer a barrier to companies capable of investing in standard aerial photogrammetry equipment. Also, the operational costs of these systems are generally comparable to or better than existing remote sensing technologies. Laser-based svstems offer distinct advantages over existing survey instruments i n areas such as forestry surveys or coastal zone monitoring, while offering complimentary data collection in other areas such as airborne spectrometry. A unique advantage of these instruments is that they are capable of penetrating vegetation allowing the ground beneath a tree canopy to be mapped directly from the air. Another advantage of laser-based data capture is that during post-processing it is possible to classify each data point as ground, vegetation, a building or other object of interest such as a power line. Once classified, removing the overlying layers is simple and allows the straight forward generation of, for example, a digital terrain model (DTM) of the ground beneath the tree canopy. Because it is an active illumination sensor, a laser system can collect data at night and can be operated in weather and at low sun angles that prohibit aerial photography. While prototypes for these instruments have been operating for several years, only recently have commercial systems been available. Table 1 compares the features of several commercial laserbased terrain mapping systems with information compiled from the literature.'

Landsat and Earth Systems Science : Development of terrestrial monitoring

Abstract: One of the major catalysts leading to the development of the global-scale Earth Systems Science concept, the international Geosphere-Biosphere Program, and the U.S. Global Change Research Program were the unique views of Earth provided by Landsat sensors over the past 25 years. This paper addresses Landsat's contributions in the Earth Systems Science arena. Early successes in observing the Earth's cloud patterns from space led to the use of this new spaceborne perspective to observe surface terrestrial features. Deployment of Landsat demonstrated that significant information about the Earth's land areas could be acquired from such an observatory. Numerous studies indicated that assessments of agricultural production, forest resources, human population surveys, and environmental conditions could be derived from Landsat data. Thus, an unanticipated outcome of the Landsat program was the evolution of unique new insights concerning terrestrial biospheric patterns and dynamics. The electronic, high precision spectral radiometry, combined with Landsat's repetitive coverage, revealed that a critical new environmental measurement, the spectral vegetation index, could be acquired with these sensors. These measurements are also of critical importance in understanding the hydrology, land surface climatology, and biodiversity characteristics of the Earth. Recognition of the value of this vegetation index in regional and global-scale studies of the Earth's environment served as a strong stimulus to the development of the Earth Systems Science research agenda, one of the major foci of NASA's Mission to Planet Earth, Earth Observing System. Since the innovation of the Landsat Thematic Mapper instrument in the early 1980s, significant progress has been achieved in assessing human impacts within the Earth systems. Significant further inputs to Earth Systems Science from Landsat are expected when Landsat 7 is launched in 1998. Refinements in radiometric response and calibration, inclusion of a 15-m panchromatic band, improvement of the spatial resolution of the thermal band to 60 m, and an aggressive acquisition strategy will all contribute to Landsat's new role as a major component of NASA's Mission to Planet Earth, Earth Observing System. Development of technologies for more refined, as well as lower cost, sensors and platforms is now underway to continue the Landsat science mission. These technology advances are expected to further enhance the capability to monitor the Earth's land areas.

Subpixel classification of Bald Cypress and Tupelo Gum trees in thematic mapper imagery

Abstract: A subpixel spectral analytical process was used to classify Bald Cypress and Tupelo Gum wetland in Landsat Thematic Mapper imagery in Georgia and South Carolina. The subpixel process enabled the detection of Cypress and Tupelo trees in mixed pixels. Two-hundred pixels were field verified for each tree species to independently measure errors of omission and commission. The cypress total accuracy was 89 percent and the tupelo total accuracy was 91 percent. Field investigations revealed that both cypress and tupelo trees were successfully classified when they occurred both as pure stands and when mixed with other tree species and water. In a comparison with traditional classification techniques (ISODATA clustering, maximum likelihood, and minimum distance) the subpixel classification of cypress and tupelo yielded improved results. Large areas of wetland where cypress was heavily mixed with other tree species were correctly classified by the subpixel process and not classified by the traditional classifiers.

Evaluating the uncertainty of area estimates derived from fuzzy land-cover classification.

Abstract: The use of remotely sensed data as input into geographical information systems has promoted new interest in issues related to the accuracy of multispectral classification. This paper investigates the impact of classification uncertainty on the estimation of area from satellite derived land-cover data. Applying four variants of the maximum-likelihood classifier, it is shown that the estimated area for different land-cover classes is highly influenced by the methods which are used for classifier training. To evaluate the uncertainty of area estimates, a new error modeling strategy is proposed. Assuming that attribute uncertainty in image classification is field-based rather than pixel-based, the image is segmented in fields according to similarities in the probability vectors of adjacent pixels. In simulating uncertainty, this field structure is explicitly taken into account. Using different strategies for image segmentation, it is shown that the spatial correlation of classification uncertainty has a major impact on the assessment of the uncertainty of area estimates. Centre for Cartography and GIS, Department of Geography, Vrije Universiteit Brussel, Pleinlaan 2, B 1050 Brussel, Belgium.

Multisource Classification of Complex Rural Areas by Statistical and Neural-Network Approaches

Abstract: The automatic generation of land-cover inventories by using remote-sensing data is a very difficult task when complex rural areas are involved. The main difficulties are related to the characterization of such spectrally complex and heterogeneous environments and to the choice of an effective classification approach. In this paper, the usefulness of spectral (Landsat-5 Thematic Mapper images), texture (grey-level cooccurrence matrix statistics), and ancillary (terrain elevation, slope, and aspect) data to characterize two complex rural areas in central Italy is quantitatively demonstrated. A statistical and a neural-network classification approach are applied to such a multisource data set, and their classification performances are assessed and compared. The classification performances of the two approaches are quantitatively evaluated in terms of global and conditional Kappa accuracies. The Zeta statistics is used to evaluate the statistical significance of the different classification accuracies obtained by the two approaches by using multisource data.

Exploring and Evaluating the Consequences of Vector-to-Raster and Raster-to-Vector Conversion

Abstract: Spatial data can be represented in two formats, raster (grid cell) or vector (polygon). It is inevitable that conversion of the data between these two formats be essential to the best use of the data. Most geographic information systems (GIS) now provide software for such a conversion. The objective of this study was to explore and evaluate the consequences of data conversion on the accuracy of the resulting data layer. Simple shapes were chosen to document the results of the raster-to-vector and vector-to-raster conversion processes. These shapes included a square, a triangle (not aligned with the grid), a circle, a hole within the circle, and a non-convex shape. Error matrices were employed to represent the changes in area through the conversion process. A second set of data including a circle, a thin rectangle, and a wide rectangle were used to examine the effect of grid cell size on both presence/absence of a feature as well as to maintain the feature's shape. Finally, recommendations for continuing this work and its application to information derived from remotely sensed data were presented.

Mapping Desert Shrub Rangeland Using Spectral Unmixing and Modeling Spectral Mixtures with TM Data

Abstract: Spectral unmixing experiments were done to explore the applicability of linear unmixing models, especially the basic least-squares method for mapping sparse vegetation in rangeland Some important theoretical and technical issues involved in physical inversion problems were addressed Based on the field reference spectra of image components, a constrained least-squares method was applied to Landsat Thematic Mapper data over an area in Long Valley, Nevada to calculate vegetation abundance in a pixel A method for formulating a well-conditioned spectral mixture by calculating the cosine of the angles between the candidate sugace components was presented This method provides a way to measure the separability of candidate endmembers quantitatively and derive spectral endmem bers objectively The results of this study suggest that the ambiguity or uncertainty in physical inversion problems arises from the inability to provide a complete set of representative reference spectra and to formulate a well-conditioned spectral mixture, not from the least-squares method itself Some of import implications of the study include the following: (1) the unmixing techniques can provide moderate estimates of vegetation fractions in arid rangeland, where vegetation is sparse, with TM data; and (2) the degree of spectral pureness of endmembers should be consistent between endmember spectra that are used for unmixing

Measuring Uncertainty in Class Assignment for Natural Resource Maps under Fuzzy Logic

Abstract: There are two kinds of uncertainty associated with assigning a geographic entity to a class in the classification process. The first is related to the fuzzy belonging of the entity to the prescribed set of classes and the second is associated with the deviation of the entity from the prototype of the class to which the entity is assigned. This paper argues that these two kinds of uncertainty can be estimated if a similarity model is employed in spatial data representation. Under this similarity model, the uncertainty of fuzzy belonging can be approximated by an entropy measure of membership distribution or by a measure of membership residual. The uncertainty associated with the deviation from the prototype definitions can be estimated using a membership exaggeration measure. A case study using a soil map shows that high entropy values occur in areas where soils seem to be transitional and that areas which are mis-classified have higher entropy values. The membership exaggeration is high for areas where soil experts have low confidence in identifi-ing soil types and predicting their spatial distribution. These measures helped in identifying that the high elevation areas were mapped with high accuracy and that error reduction efforts are needed in mapping the soil resource in the low elevation areas.

Detection of vegetation changes associated with extensive flooding in a forested ecosystem.

Abstract: Monitoring broad-scale ecological responses to disturbance can be facilitated by automated change-detection approaches using remotely sensed data. This study evaluated the effectiveness of five unsupem'sed change-detection techniques using multispectral, multitemporal SPOT High Resolution Visible (HRV) data for identifying vegetation responses to extensive flooding of a forested ecosystem associated with Tropical Storm Alberto in July 1994. Standard statistical techniques, logistic multiple regression, and a probability vector model were used to quantitatively and visually assess classification accuracy. The change-detection techniques were (I) spectral-temporal change classification, (2) temporal change classification based on the Normalized Difference Vegetation Index (NDVI), (3) principal components analysis (PCA) of spectral data, (4) PCA of NDVI data, and (5) NDVI image differencing. Spectral-temporal change classification was the least effective of the techniques evaluated. Classification accuracy improved when temporal change classification was based on NDVI data. Both PCA methods were more sensitive to flood-affected vegetation than the temporal change classifications based on spectral and NDVI data. Vegetation changes were most accurately identified by image differencing of NDVI data. Logistic multiple regression and a probability vector model were especially useful for relating spectral responses to vegetation changes observed during field surveys and identifying areas of agreement and disagreement among the different classification methods. to utilize satellite data to assess vegetation responses to flooding in a forested ecosystem and to compare analytical approaches for vegetation change detection. Minimal wind and storm surge damage accompanied Alberto as it made landfall on the Florida panhandle near Fort Walton Beach on 3 July 1994 and traveled inland. However, due to weak steer

Aquatic macrophyte modeling using GIS and logistic multiple regression

Abstract: Aquatic macrophytes are non-woody plants, larger than microscopic size, that grow in water. They are an essential component of wetland communities because they provide food and habitat for a variety of wildlife, and they regulate the chemistry of the open water. Unfortunately, they also hinder human activities by clogging reservoirs and affecting recreational activities. Given their impact on environmental processes as well as on human activities, it is important that aquatic macrophytes be monitored and managed wisely. This research focuses on developing a predictive model, based on several biophysical variables, to determine the future distribution of aquatic macrophytes. Par Pond, a cooling reservoir at the Savannah River Site in South Carolina, was selected as the study area. Four biophysical variables, including water depth, percent slope, fetch, and soils, were digitized into a geographic information system (GIS) database. A logistic multiple regression (LMR) model was developed to derive coefficients for each variable. The model was applied to seven water depths ranging from the 181-foot contour to the 200foot contour at Par Pond to determine the probability of aquatic macrophyte occurrence at each water level. Application of the LMR model showed that the total area of wetland would decline by nearly 114 ha between the 200- and 181foot contours. The modeling techniques described here are useful for predicting areas of aquatic macrophyte growth and distribution, and can be used by environmental scientists to develop effective management strategies.

The Landsat Program: Its Origins, Evolution, and Impacts

Abstract: Landsat 1 began an era of space-based resource data collection that changed the way science, industry, governments, and the general public view the Earth. For the last 25 years, the Landsat program - despite being hampered by institutional problems and budget uncertainties - has successfully provided a continuous supply of synoptic, repetitive, multispectral data of the Earth's land areas. These data have profoundly affected programs for mapping resources, monitoring environmental changes, and assessing global habitability. The societal applications this program generated are so compelling that international systems have proliferated to carry on the tasks initiated with Landsat data.

Empirical relationships between structural and spectral factors of Yellowstone lodgepole pine forests

Abstract: Multiple regression analysis was used to examine the relationships between spectral and biotic factors within the lodgepole pine (Pinus contorta var. latifolia) forests of Yellowstone National Park. Field-sampled data on forest overstory and understory conditions were regressed against Landsat Thematic Mapper (TM) radiance values and transformed TM data for 70 stands. Factors relating to the physical structure of the forest canopy (height, basal area, biomass, and leaf area index (LAI)) are best predicted using a combination of visible and middle-infrared Thematic Mapper bands. Other overstory factors (density, size diversity, mean diameter, and number of overstory species) were not well explained by the TM data or by combinations of TM data with transformed spectral data. Understory factors [number of seedlings; number of understory species; total cover by forbs, grasses, and shrubs; and total living and nonliving cover) were poorly explained by regression models incorporating spectral and transformed spectral data.

The Effect of Neural-Network Structure on a Multispectral Land-Use/Land-Cover Classification

Abstract: While neural networks are now an accepted alternative to statistical multispectral classification techniques for remote sensing image classification, the network approach presents both unique challenges and abilities. The size of the hidden layer must be determined by trial and error, and the random initial weight settings result in different paths for the training procedure, making the network a non-deterministic classifier. For the sample classification presented here, it was found that there was a range of optimal hidden layer sizes below which the accuracy decreased and above which the training time increased. However, it was also found that, for a fairly wide range, the hidden layer size made little difference to the final classification accuracy. Initial weight randomization was as much of a factor as hidden layer size. Using 3 by 3 windows of data in each band was found, despite increased training time per iteration, to achieve similar accuracy with less overall training time, although with less consistency.

GIS-based evaluation of salmon habitat in the Pacific Northwest

Abstract: Categorization of 164,083 kilometres of stream length has provided the first quantitative measure of the extent and location of potential salmon stream habitat throughout western Washington State. Reach slope and forest seral stage provided a coarse indicator of channel condition across the region. Reach-average slopes calculated for individual stream reaches using 30-metre digital elevation model (DEM) data, correctly identified low-gradient (less than 4.0 percent slope) response reaches that typically provide habitat for anadromous salmon with an accuracy of 96 percent (omission and commission error rates of 24.0 and 4.0 percent, respectively). Almost one-quarter (23.2 percent) of all stream length categorized consisted of response reaches, of which only 8.7 percent were associated with late-sera1 and 20.7 percent with mid-sera1 forest stages. Approximately 70 percent of the total stream length potentially providing anadromous salmon habitat is associated with non-forested and early-sera1 stage forests. GIS-based analytical techniques provided a rapid, objective, and cost-effective tool to assist in prioritizing locations of salmon habitat preservation and restoration efforts in the Pacific Northwest.

GIS modeling of elk calving habitat in a prairie environment with statistics

Abstract: In 1981, elk were first introduced to the prairie environment of the Cimarron National Grassland in Kansas. The lack of information regarding critical elk habitat in the prairie and the demand for integrated land use necessitated elk habitat studies in the grassland. A logistic regression model was developed to assess the relationship between observed calving sites and a set of biophysical and anthropogenic habitat variables. A GIs was used to solicit spatial information and implement the logistic model to predict the spatial distribution of calving probabilities in the grassland. Seep pits, the manmade water supply facilities along the river corridor, and cottonwood and salt cedar in the riparian areas were found statistically significant in explaining elk calving habitat; in contrast, highways and improved gravel roads appear to affect calving habitat in a negative fashion. The results also suggested possible adaptation of elk to human disturbance.