Potential of High-resolution Satellite Imagery

for National Mapping Products[1]

Dr. Rongxing (Ron) Li, Associate Professor

Department of Civil and Environmental Engineering and Geodetic Science

The Ohio State University

2070 Neil Avenue, Columbus, OH 43210

Tel. (614) 292-6946, Fax (614) 292-2957, Email: li.282@osu.edu

AbstracT

This paper discusses the potential of the upcoming high-resolution (1m ground resolution) satellite imagery for national mapping products. An analysis of the capabilities of these high-resolution imaging systems and existing satellite imaging systems for the representation and extraction of elevation information, such as terrain relief displacement and parallaxes, is given. In-track and cross-track stereo mapping techniques using satellite pushbroom CCD linear arrays are described. A photogrammetric processing model considering such geometry is introduced. Based on an error estimation and analysis, it is concluded that if the strict photogrammetric processing model and ground control points are employed, high-resolution satellite imagery can be used for the generation and update of national mapping products (7.5 minute quadrants or at a map scale of 1/24,000), including Digital Elevation Models (DEM), Digital Orthophoto Quadrants (DOQ), Digital Line Graph (DLG) databases, and Digital Shoreline (DSL) databases.

Introduction

The launch of the new generation of high-resolution (upto 0.82m or 1m) commercial earth imaging satellites in late 1997 and after will mark the start of a new era of space imaging for earth observations (Fritz 1996). Among several commercial high-resolution imaging satellites, Early Bird (3m resolution) of EarthWatch, Incorporated, IKONOS (1m resolution) of Space Imaging, Inc, and OrbView-1 (1, 2 and 4m resolution) of The Orbit Sciences Corporation are planned to be launched in late 1997. Quick Bird (1m resolution) of EarthWatch, Incorporated will be launched in 1998. The imagery will maintain dominant spectral advantages demonstrated by lower resolution satellite imaging systems such as Landsat TM and SPOT. More important is that the new generation of high-resolution satellite imagery will provide strong geometric capabilities that have not been available from existing satellite imaging systems. Specific geometric aspects of the imagery that are interesting to the mapping community include, for example, high-resolution, photogrammetric stereo capability, and revisit rate. With one-meter ground resolution, objects appearing in most digital national mapping products, such as Digital Elevation Models (DEM), Digital Orthophoto Quadrants (DOQ), Digital Line Graphs (DLG), and Digital Shorelines (DSL), can be represented in the imagery (USGS 1997, Ellis 1987, Lockwood 1997). Although linear CCD (Charge-Coupled Device) arrays are used in most one-meter resolution imaging systems and require more complicated photogrammetric models, they will for the first time provide horizontal and vertical positions of measured objects in the ground coordinate system at accuracies of several meters. The revisit rate of 1-4 days, depending on satellites and latitude, makes it possible to map an area frequently without special flight planning and scheduling as required in aerial photogrammetric data acquisition. Stereo models thus formed are valuable for mapping product updating and accurate change detection.

Accurate elevation information is crucial to national mapping products if satellite data are used. The SPOT imaging system has an enhanced resolution of 10m of the panchromatic band with a cross-track stereo capability only. Accuracies of the elevation derived from the data have achieved around 10m with Ground Control Points (Al-Rousan et al. 1997) and are not sufficient for many mapping products (USGS 1997). Airborne mapping is currently still the primary technique employed for national mapping because of its advantages such as high accuracy, flexible scheduling, and easy-to-change configurations. Among the new commercial high-resolution satellite imaging systems, The EarlyBird system of EarthWatch, Incorporated has a ground resolution of 3m in the panchromatic band and captures images frame by frame. The pushbroom imaging technique with two or more linear CCD arrays is adopted by systems of 1m resolution. Such a configuration provides a so-called in-track stereo mode where stereo pairs necessary for deriving the horizontal and elevation information of objects can be acquired in real-time; the cross-track stereo requires additional time to allow the satellite to point to the same ground area from a neighboring track. Stereo mapping capabilities of similar airborne imagers and systems mounted on board space shuttles were demonstrated (Heipke et al. 1996, Fraser et al. 1997). There are four advantages of the high-resolution satellites: a) the highest resolution ever available to the civilian mapping community, b) extremely long camera focal length, for example ten meters, for capturing terrain relief information from the satellite orbit, c) fore-, nadir-, and aft- looking linear CCD arrays supplying in-track stereo strips and “pointing” capabilities generating cross-track stereo strips, and d) a base-height (sensor baseline vs. orbit height) ratio of 0.6 and higher that is similar to aerial photographs. Table 1 gives some selected technical specifications of the high-resolution commercial satellites, which are important for three-dimensional mapping applications (Fritz 1996). If not specified, parameters used in equations and estimations in the rest of the paper are based on the IKONOS system.

Item	Technical data
Ground resolution	1 - 3 m
Orbit height	460 - 680 km
Focal length	Up to 10 m
Image recording	CCD linear or frame arrays
Stereo mode	In-track and/or cross-track stereo
Base-height ratio	0.6 - 2.0
Ground swath width	6 - 36 km
Revisit rate	1-4 days for most of northern hemisphere

Table 1. Some selected technical specifications of the high-resolution commercial satellite imagery

This paper gives an analytical assessment of the mapping potential of high-resolution commercial satellite imagery. Aspects of photogrammetric processing of the imagery are discussed. An estimation of ground accuracies of measured objects from the imagery is given considering various error sources. Technical requirements of U.S. national mapping products including DEM, DOQ, DLG, and DSL are compared with the estimated measuring accuracies. Conclusions and recommendations based on the analysis are made for applications of the imagery in national mapping production.

three-dimensional mapping from satellite imagery

Presentation of elevation information

Although it may not be appropriate, photogrammetry has often referred to techniques handling aerial or terrestrial images, while remote sensing dealt with satellite imagery. This simple separation between photogrammetry and remote sensing was probably based on the fact that each of them provides some capabilities which cannot be achieved by the other. Among others, the comparative capabilities include ground coverage, repeatability of observations, spectral ranges, and geometry for three-dimensional mapping. Compared to aerial photographs, satellite images cover larger ground areas because of the orbit height. Furthermore, the images are acquired repetitively within the lifetime of the satellites, which is usually a number of years. The continuity of an imaging system is maintained by launching subsequent, often improved, satellites of the same series. The satellite images usually possess a broad range of spectrum and have great radiometric advantages over early black-and-white aerial photographs. However, it is the geometric strength of photogrammetry that enables highly accurate horizontal and vertical positions for national mapping. With the current development and future applications of high-resolution satellite imagery, the gap between photogrammetry and remote sensing may soon be reduced and vanish for some products in the near future.

The photo-scale of a vertical photograph (horizontal image plane) is defined as f/H. For example, aerial photographs for national mapping taken by a camera with a focal length of 152mm at a flying height of 20,000ft give a photo-scale of 1/40,000. The high-resolution satellite imagery of IKONOS has a photo-scale of 1/68,000 (f/H=10m/680km). On the other hand, the photo-scale can also be expressed as a ratio of an image distance over its corresponding ground distance: d_photo/d_ground. Taking the ground pixel size as d_ground=1m and photo-scale as 1/68,000, the pixel size on the CCD chip is computed as 15mm. Therefore, the higher the ground resolution, the smaller the ground pixel size, and furthermore, the smaller the pixel size on the chip. This may impose the ultimate limit on the ground resolution. In order to derive the elevation, the relevant information must be present in the images in some form. One of them is terrain relief displacement (Figure 1). Assume that an object on the ground has a height of Dh and a distance R from the nadir point of a vertical aerial photograph with a flying height of H. The focal length of the aerial camera is f. The relief displacement d of the object caused by the elevation difference Dh in the aerial photograph can be calculated as (Moffitt and Mikhail 1980, Wolf 1983)

d = Dh f/H R/(H- Dh). (1)

Figure 1. Relief displacement in an aerial photograph and a satellite image

It is obvious that the relief displacement is proportional to the elevation difference Dh. Here f/H is the photo-scale. Small scale photographs are thus less capable of representing the relief displacement information. Objects close to the nadir point have small distances R and therefore, will have small relief displacements. The same elevation difference Dh would produce a larger relief displacement as the object is situated farther away from the nadir point. The terms of Equation (1) are examined for both aerial photography and satellite image cases (Figure 1) where the object height Dh remains the same.

a) The flying height increases greatly from aerial photography to satellite imaging. However, a high-resolution camera, for example, IKONOS has a focal length of 10m that makes the second term f/H (photo-scale) comparable to an aerial photo-scale.

b) In the third term R/(H- Dh), since H is much greater than Dh, the value of the term decreases rapidly from aerial photography to satellite imaging. Note that the satellite image covers a much larger ground area. In its nadir area, the distance from the nadir point to the object, R, is small and therefore, the corresponding terrain relief displacement will be small; in the most imaged area, R is sufficiently large in comparison to H and a ratio of R/(H- Dh) becomes close to that of aerial photography.

Generally, it is expected that sufficient terrain relief displacement information will be contained in the high-resolution satellite imagery. Fine topographic variations should be reflected.

In addition to the relief displacement information in single images, parallax, a positional change in the flying direction of an object point in a vertical stereo image pair caused by the imaging platform motion, can be used to derived the elevation or elevation difference (Wolf 1983):

p = f B / (H- Dh), (2)

where p is the parallax of the top point of the object in Figure 1 and B is the baseline distance between the two exposure centers of the camera. In the high-resolution imaging case, the orbit height H is great and makes parallax p small. However, B is great in both cross-track stereo mode and in-track stereo mode because the pointing angle of 30^o (Quick Bird) to 45^o (IKONOS) yields large baselines by different combinations of fore-, nadir-, and aft-looking strips. The extremely long focal length f makes a further contribution to a greater parallax. The images will have a typical parallax of 85mm to 170mm, or 6 pixels to 12 pixels, which is sufficiently large to be measured and used for deriving elevation information.

Stereo Model Formation

Figure 2 illustrates three CCD linear arrays (fore, nadir, and aft) “looking” at the same ground profile across the track. Note that the three looking angles are, for instance, a^o, 0^o, and -a^o, respectively. Each CCD linear array produces one strip along the track. An object on the ground is usually covered by three image strips. The specific image lines of the strips containing the object are taken at different times. Thus, the in-track stereo mode has three combinations of stereo pairs, namely F-N (fore-nadir), N-A (nadir-aft) and F-A (fore-aft). Baselines are then 271km (Quick Bird) - 680km (IKONOS) for F-N and N-A stereo pairs, and 542km to 1360km for F-A pairs.

Figure 2. Fore-, nadir-, and aft-looking CCD linear arrays and their looking angles

The base-height ratio is a critical factor for accurate 3-D mapping. Aerial photogrammetry for national mapping usually has a base-height ratio of 0.6. In the case of high-resolution satellite imaging, the orbit height is, for instance, 680km. Considering the possible combinations, the base-height ratio of F-N and N-A pairs is 0.6 (Quick Bird) to 1.0 (IKONOS) and an F-A pair 1.2 to 2.0. This means that the satellite images are able to give a similar base-height ratio as aerial photographs. This partly ensures the quality of the 3-D spatial data derived from the imagery.

The pointing capability across the track with an angle of 45^o (IKONOS) makes it possible to flexibly form cross-track stereo strips. The accuracies of objects derived from the cross-track stereo strips should be similar to those from in-track strips, except that the “side-look” geometry may produce some refractive effects positive to ground relief along track.

Photogrammteric Modeling

Interior and Exterior Orientation Parameters

To simplify the explanation without losing generality, Figure 3 illustrates the imaging system with fore- and nadir-looking CCD linear arrays (array I and array II respectively). In order to accurately model the imaging geometry, the following interior orientation parameters are defined for each linear array:

· Focal length f,

· Ratio of pixel size dy and dx in y and x direction, dy/dx,

· Principal point coordinate x_o (only one image coordinate is needed for a linear array), and

· Lens distortion correction coefficients: p₁, p₂, k₁, k₂, and k₃.

Figure 3. Imaging system with fore- and nadir-looking CCD linear arrays

in the ground coordinate system

Three lenses can be used, one for each CCD linear array. Or one lens can be used for all three CCD linear arrays. At this time, this information is not available from the imaging companies. The following discussion is based on a three lens configuration. The above parameters are then for all three CCD linear arrays (fore, nadir, and aft). Each array has its own 2-D image coordinate system x_i-O_i-z_i. The linear arrays are subsequently defined in the platform coordinate system X_pl-Y_pl-Z_pl. All coordinate systems are referenced to the ground coordinate system. The two linear arrays have a fixed geometric relationship, namely, a translation vector O_I-O_II or Dr=(DX, DY, DZ)_pl and a rotation matrix M_Dpl = M_D(Dw, Df, Dk)_pl from x_I-O_I-z_I to x_II-O_II-z_II defined in the platform coordinate system. For a system with three linear arrays, there are 36 interior orientation parameters, including eight intra-array parameters for each array and 12 inter-array parameters (two sets of translation and rotation parameters among the three arrays). Here it is assumed that the image coordinate system of array I is used as the platform coordinate system in order to reduce the parameters involved. Since a laboratory calibration is usually impossible during the lifetime of the satellite, it is recommended that frequent evaluations of the interior orientation parameters be carried out using calibration ranges.

Exterior orientation parameters are used to relate the platform to the ground coordinate system. At any time, the exterior orientation parameters of the platform are coordinates of the origin of the platform coordinate system (X_o,Y_o,Z_o) and rotational angles (w, f, k) with respect to the ground coordinate system. The rotation is described by a 3-D rotation matrix M_pl which is a function of the rotation angles (Moffitt and Mikhail 1980). For simplicity mentioned above, the exterior orientation parameters of array I are identical to the platform: (X_o,Y_o,Z_o) and M_pl. Those of array II are calculated by applying the inter-array parameters (translation and rotation) between array I and array II to the exterior orientation parameters of array I: (X_o,Y_o,Z_o)+ M_pl M_Dpl Dr and M_pl M_Dpl. The exterior orientation parameters of array III can be computed in the same way.

Photogrammetric Intersection

When interior and exterior orientation parameters of all lines in the image strips are known, image coordinates of an object in one or more stereo pairs of strips can be measured and used to calculate the ground coordinates of the object. This computational procedure is called photogrammetric intersection. Suppose that a stereo line pair consists of an image line of array I (fore-looking) at time t and an image line of array II (nadir looking) at time t+Dt (Figure 4). Image coordinates, x_I and x_II, of a ground point P in both image lines are measured. The ground coordinates of P, or vector r, are to be computed

Figure 4. Photogrammetric intersection in in-track stereo mode

With the known interior and exterior orientation parameters of Dr, M_Dpl, (X_o,Y_o,Z_o)_t, (X_o,Y_o,Z_o)_t+Dt, M_pl,t, and M_pl,t+Dt, the rotation matrices of array I and array II are:

M_I,t = M_pl,t

M_II,t+Dt = M_pl,t+Dt M_Dpl . (3)

The corresponding exposure centers are:

r_Io = (X_o,Y_o,Z_o)_t

r_IIo = (X_o,Y_o,Z_o)_t+Dt + M_II,t+Dt Dr . (4)

The baseline is the difference between r_IIo and r_Io:

B= r_IIo- r_Io . (5)

A coplanarity equation can be established by requiring that the three vectors, namely, B, r_I, and r_II, be on the same plane:

B ^. (r_I r_II) = 0 . (6)

r_II and r_I are expressed as

r_I = l_I M_I,t (x_I,0,f_I)^T ,

r_II = l_II M_II,t+Dt (x_II,0,f_II)^T . (7)

Equation (7) is then inserted into Equation (6). With known baseline B, known interior and exterior orientation parameters (f_I, f_II, M_I,t, M_II,t+Dt), and measured image coordinates (x_I and x_II) of the object point P, the two scaling factors l_I and l_II are solved by a least squares adjustment of the equations resulting from Equation (6). Using the calculated l_I and l_II in Equation (7), r_I and r_II become known. Finally, the position of point P or r = (X,Y,Z) can be computed by either of the following equations:

r = r_Io + r_I or

r = r_IIo + r_II . (8)

Navigation Data as Exterior Orientation Parameters

In Equations (1)-(8), both interior and exterior orientation parameters are assumed to be known. The imaging platform positions of (X_o,Y_o,Z_o)_t and (X_o,Y_o,Z_o)_t+Dt at time t and t+ Dt are measured by onboard kinematic DGPS (Differential Global Positioning System) at an accuracy of 3m. The corresponding rotation angles of the imaging platform (w, f, k)_pl,t and (w, f, k)_pl,t+Dt (M_I,t and M_I,t+Dt) are determined by star trackers at an accuracy of two arcseconds. In fact, the sampling rates of DGPS and star trackers are usually lower than that of image recording. There are not DGPS and startracker data available for every image line. There will be polynomial interpolations of navigation data between “orientation lines” which are image lines with actual DGPS and startracker navigation data. Such interpolations can also be performed using orbital parameters (Slama et al. 1980). The interpolated navigation data along with the imagery can be distributed to users. On the other hand, the polynomial interpolation parameters can be estimated through a bundle adjustment procedure using Ground Control Points (GCP), as applied in airborne or space shuttle “three-line” photogrammetric mapping (Hofmann 1986, Heipke et al. 1996). The satellite orbit is supposed to be relatively smooth, so that high-order polynomial interpolations are not expected. An appropriate interpolation model in bundle adjustment should be determined by an experiment using the real data. In addition, DGPS and startracker navigation data can be treated as “nonperfect” positional and attitude observations. The least squares adjustment will make corrections to these observations depending on weights (confidences) assigned accordingly.

potential for national mapping products

Error estimation

According to technical specifications of the satellite imaging companies, positional accuracies of objects derived from high-resolution satellite imagery are 12m (horizontal) and 8m (vertical) without GCPs, and 2m (horizontal) and 3m (vertical) with GCPs (Fritz 1996). An independent estimation of the accuracies is given using certain assumptions and error propagations. Errors of exposure centers provided by DGPS are s_xo=s_yo=s_zo=3m. Errors of the platform attitude from startrackers are s_wo=s_fo=s_ko=2 arcseconds. The elevation error of a ground point derived from a vertical stereo photograph pair caused by flying height error s_H, baseline errors s_B, and image parallax measurement error s_p can be estimated as (Wolf 1983):

s_h1²= s_H² + [(H-h)/B]²s_B² + (H-h) ⁴/ [B²f²] ²s_p² . (9)

Considering relationships between the exposure centers, the flying height, and the baseline, s_H and s_B are evaluated as s_zo=3m and 1.414s_xo=4.24m, respectively. An image coordinate measurement error of 0.5 pixels (7.5mm) is assumed. The parallax measurement error s_p is then 10.6mm. In Equation (9) let (H-h)=H=680km, f=10m, and B=680km (F-N or N-A stereo). s_h1 is evaluated as s_h1=[9m² + 18m² + 0.4m²]^1/2=5.2m. Note that s_h1 considers the impact of positional DGPS navigation errors implicitly through s_H and s_B. Ideally the attitude error of the nadir-looking linear array does not contribute to the elevation error of the ground point. The contribution by the fore or aft linear array is

s_h2= H s_w= 6.6 m. (10)

The overall elevation error of the point without GCPs is s_h=(s _h1² + s _h2²) ^1/2=8.4m, which is close to the specified error of 8m by the imaging companies. In same way, the horizontal error (without GCPs) of 13.2m can be derived under the same assumptions, compared to 12m specified by the imaging companies.

It should be noted that the above error estimation does not take the effect of the interpolation of navigation data into account. This effect can only be determined effectively when actual data are available after the successful launch of the satellites. Furthermore, the specified accuracies of ground points with GCPs are also difficult to verify without actual data. Ridley et al. (1997) simulated the 1m resolution data from 0.2m resolution aerial photographs. Using the geometry of frame cameras instead of CCD linear arrays, the ground point error was around 2m (with GCPs). Airborne three-line scanner imagery of 2m resolution with GCPs demonstrated a ground accuracy of 1m (horizontal) and 2m (vertical) (Heipke et al. 1996). It seems that a bundle adjustment combines interpolation parameters of navigation data, image measurements, navigation measurements, and unknown coordinates of the measured ground points in a strict mathematical model and is capable of providing a better ground point accuracy under global optimal conditions Specific issues such as distribution of GCPs, minimum number of GCPs for each national mapping product, and correlations between parameters of the bundle adjustment need to be addressed based on experiments using actual data.

Application Potential for DEM, DOQ and DLG

Although hard opies of 1/24,000 topographic maps have been scanned and/or digitized for producing digital national mapping products of DEMs, DOQs and DLGs, the aerial photographs are the primary data used to derive information necessary for the products. These photographs usually have a base-height ratio of 0.63. This requirement can be met easily by forming stereo pairs using F-N and N-A combinations. Elevation error s_h in DEM is required to be less than 15m. This requirement can be met without GCPs. DOQ products have a resolution of 1m to 2m. Elevation errors of DEM needed to produce DOQ are required to be less than 7m. The image resolution requirement is met. The elevation error of less than 7m can only be satisfied by applying GCPs according to the above analysis. DLG (1/24,000) allows the horizontal error to be smaller than 12m and the vertical error smaller than 15m. The required elevation error of 15m is greater than the estimated elevation error without GCPs. Although the required horizontal error of 12m equals the specified horizontal error (without GCPs), it is recommended that GCPs are also used.

In addition to geometric accuracy, the capability of representation of topographic objects and extraction of the information from the imagery is of great interest. The 1m resolution stereo imagery of the panchromatic band should be able to give sufficient image features for generation of DEM grid points spaced every 30m manually or automatically by digital image matching techniques. Existing DEMs that may be updated using such imagery can be employed to produce black-and-white DOQs of 1m resolution (3.75minute quadrants) and 2m resolution (7.5 minute quadrants). Additional 4m resolution multispectral DOQs may be produced as byproducts using the multispectral bands and the same DEMs. Attributes of 1/24,000 DLG quadrants are derived from maps of the same scale (USGS 1997). Their base categories include: (1) political boundaries, (2) hydrography, (3) Public Land Survey System (PLSS) data, (4) transportation data, (5) other significant manmade structures, (6) hypsography, (7) surface cover (e.g., vegetative surface cover), (8) nonvegetative surface features (e.g. lava and sand), and (9) surface control and marks. Categories (3) and (9) are usually well documented by relevant federal agencies. Their update in the DLG quadrants can be performed either manually or automatically. The information update of the rest of the categories can be, in principle, carried out by using high-resolution satellite imagery. Positive results of attribute information extraction from simulated high-resolution satellite imagery for Ordnance Survey of UK was reported (Ridley 1997). Experiments using actual data should be performed to verify the capability for DLG quadrants.

Application Potential for DSL

Theoretically, an instantaneous shoreline is the intersecting line between a Coastal Terrain Model (CTM), which is a digital surface model of a strip of land along the shoreline with onshore elevation and nearshore bathymetry, and a water surface. It changes as the water surface increases or decreases. Therefore, a tide coordinated shoreline should be defined based on a certain water datum. National Geodetic Survey (NGS) of NOAA (National Oceanic and Atmospheric Administration) produces the nation’s tide coordinated Mean Lower-Low Water (MLLW) and Mean High Water (MHW) shorelines in nautical charts. These datums refer to the synoptic averages over a 19.2 year lunar/solar cycle. NGS derives the tide coordinated shorelines using periodic aerial photographs taken at the time of the desired water level using coordinated quasi real-time hydrographic observations (Slama et al. 1980). Since a satellite has a prescribed orbit, there is no control to image the coast area at the time of the desired water level. The satellite has a revisit rate of 1-4 days. The CTM may change along with time because of erosion, land use, and other natural or human activities in the coastal zone. The instantaneous shorelines thus derived from the high-resolution satellite imagery vary according to both CTM and water level.

The stereo satellite image strips can be used to determine the CTM. Onshore land elevations of grid points of the CTM are computed by either image matching or interactive measurements on panchromatic images. Offshore grid points within the shallow area may be determined in the same way from infrared band images (4 meter resolution), which penetrate water better than visible band images (Slama et al 1980, Lillesand and Kiefer 1994). Within a time period, a set of low water satellite images can be accumulated and used to generate a reliable CTM with maximum number of grid points determined from high-resolution panchromatic images. This CTM can be used wherever no instantaneous CTM is necessary. The accuracy of the CTM is expected to be 2m horizontal and 3m vertical with GCPs. Water surface levels can be modeled by a computer modeling system using water gauge observations and other hydrographic data along the shoreline. A water surface model may have a vertical accuracy of several centimeters (Bedford and Schwab 1994). Since historical water level data have been archived, both MLLW and MHW can be determined.

The instantaneous shoreline at the time of imaging is projected in the image strips. This line separates the land and water areas and can be extracted either interactively by on-screen digitizing or automatically by edge enhancement and edge following. The position of the shoreline in the ground coordinate system is triangulated from the extracted lines in stereo strips. The instantaneous shorelines derived from the satellite imagery at different time should be corrected to tide coordinated shorelines. Data used for this correction include the instantaneous shorelines, updated CTM, and the instantaneous water level and desired water datum. Research on the correction algorithms should be conducted to make the comparison of the shorelines and the estimation of a datum coordinated shoreline possible.

The Digital Shoreline (DSL) is calculated from shorelines generated periodically over a long term. Currently, the National Geographic Data Committee is preparing a National Standard for Digital Shoreline Databases (NGDC 1997). It is expected that the accuracy of CTM grid points will be 2-3m. The water level accuracy is to be submeters. Consequently, the DSL generated using the above method should have the potential to produce DSL. An extreme case would be a very flat coastal area where a small vertical error would yield a large horizontal error of the shoreline. Such a difficulty should be overcome by examining both the estimated shoreline and the shorelines in images.

Discussions and Conclusions

The estimation and analysis presented above show a great potential of the upcoming high-resolution satellite imagery for national mapping products. Verification of the potential and assessment of the general mapping capabilities of the imagery are envisioned by The Ohio State University. A high altitude aerial photogrammetric calibration range in Madison County, Ohio will be used for DEM, DOQ and DLG related experiments. The range consists of approximately 23 Ground Control Points (GCPs) located within a rectangular region of 14 miles east to west and 9 miles north to south. The range center is about latitude 39°56'24" N and longitude 83°24'42" W (NAD-83). The GCPs are distributed mostly in an east-west direction. One of the lines (US 40) has GCPs in an east-west extent of 13-15 km, which gives a cross-track control of satellite images. The targets consist of painted circles on asphalt pavement. Each has an one-meter circle, centered on the monument and painted flat white. A three-meter flat black circle is painted as background to enhance the contrast. The range was surveyed by GPS methods and a network adjustment was applied. All targets along US-40 are known in WGS84 based on a 2nd-order station at the Madison County Airport. They are internally consistent in three dimensions to about 2 cm RMSE. The target points are distributed in three zones. Zone I covers mostly rural and agricultural objects, Zone II is a part of the highway, and Zone III covers more manmade objects, such as buildings, roads and parcels. Since the range does not provide significant elevation variation, an additional site will be needed for a DEM experiment. An experiment of DSL is to be carried out for the shoreline segment of Lake Erie. Shorelines derived from the periodic satellite imaging observations will be used for monitoring shoreline changes from erosion. Causes and impact of shoreline erosion can be analyzed by integration of the shoreline changes and other scientific and social economic data in a GIS environment (Li and Cho 1996).

A number of technical aspects will have to be dealt with for mapping purposes. They include, among others,

a) Increased impact of atmospheric refraction and earth curvature,

b) Development/improvement of photogrammetric CCD linear array models, and

c) Overcoming low visibility caused by cloud cover in certain areas.

Aspect c) may be considered along with applications of other satellite data which are less weather dependent, such as RADARSAT, to achieve so-called all weather mapping capability. This is especially important when dealing with event based mapping, for instance, shoreline monitoring and flood mapping.

It is expected that high-resolution satellite imagery can be, to some extent, used to reduce the demand for aerial photographs for middle scale to some large scale (for example 1/24,000) mapping. Strict photogrammetric models will be employed to derive accurate horizontal and vertical position information from the imagery. Additional characteristics of the imagery such as periodic and global coverage, strong spectral capability, and Internet based ordering and distribution systems make the technology even more attractive to the mapping community. Overall, high-resolution satellite imagery will mark a significant advancement of applications of satellite imaging technology in mapping.

REFERENCES

Al-Rousan, N., P. Cheng, G. Petrie, T. Toutin, and M.J.V. Zoej 1997. Automated DEM Extraction and Orthoimage Generation from SPOT Level 1B Imagery. Photogrammetric Engineering and Remote Sensing, Vol.63, No.8, pp.965-974.

Bedford, K. and D. Schwab 1994. The Great Lakes Forecasting System - An Overview. Proceedings of National Conference on Hydraulic Engineering, August 1-5, 1994, Buffalo, NY, pp.197-201.

Ellis, M.Y. 1978. Coastal Mapping Handbook. USGS/NOS Publication.

FGDC 1997. Public Comment on the Proposal to Develop the ``National Shoreline Data Standard'' as a Federal Geographic Data Committee Standard. Federal Register, Vol.62, No.156, August 13, 1997, pp.43342‑43344.

Fraser, C., P. Collier, J. Shao, and D. Fritsch 1997. Ground Point Determination Using MOMS-02 Earth Observation Imagery. Geomatica, Vol.51, No.1, pp.60-67.

Fritz, L.W., 1996. Commercial Earth Observation Satellites, Intl. Archives of Photogrammetry and Remote Sensing, ISPRS Com.IV, pp.273-282.

Heipke, C., W. Kornus and A. Pfannenstein 1996. The Evaluation of MEOSS Airborne Three-Line Scanner Imagery: Processing Chain and Results, Photogrammetric Engineering and Remote Sensing, Vol.62, No.3, pp.293-299.

Hofmann, O. 1986. Dynamische Photogrammetrie. Bildmessung und Luftbildwesen, Vol.54, No.3, pp.105-121.

Li, R. 1997. Mobile Mapping: An Emerging Technology for Spatial Data Acquisition, Photogrammetric Engineering and Remote Sensing, Vol.63, No.9.

Li, R. and W.K. Cho, 1996. Geographical Information Systems for Shoreline Management - A Malaysian Experience, Proceedings of ADB Coastal Erosion Conference, Kuala Lumpur, Malaysia, 9pp.

Li, R., K.P. Schwarz, M.A. Chapman, and M. Gravel, 1994. Integrate GPS, INS, and CCD Cameras for Rapid GIS Data Acquisition, GIS World, Vol.7, No.4, pp.41-43.

Lillesand, T.M. and R.W. Kiefer, 1994, Remote Sensing and Image Interpretation, John Wiley &Sons, Inc.

Lockwood, M. 1997. NSDI Shoreline Briefing to the FGDC Coordination Group. January 7, 1997, NOAA/NOS (unpublished), 29pp.

Moffitt, F.H and E.M. Mikhail 1980. Photogrammetry. Harper&Row Publishers, Inc., New York, NY.

Ridley, H.M., P.M. Arkinson, P. Aplin, J.-P. Muller, and I. Dowman 1997. Evaluating the Potential of the Forthcoming Commercial U.S. High-Resolution Satellite Sensor Imagery at the Ordnance Survey. Photogrammetric Engineering and Remote Sensing, Vol.63, No.8, pp.997-1005.

Slama, C.C., C. Theurer, and S.W. Henriksen, 1980. Manual of Photogrammetry, American Society of Photogrammetry, Va.

USGS 1997. Fact sheets and WWW pages of GeoData: Digital Line Graphs, Digital Elevation Models, and Digital Orthophotos. U.S. Geological Survey, Reston, Virginia.

Wolf, P.R. 1983. Elements of Photogrammetry. McGraw-Hill, Inc.

---

[1] Paper submitted to PE&RS in September 1997.