Satellite Images, Meteorology, Weather, Space, Planets, Moons, Arctic
GroundMap, a tool to remap satellite images to a desired projection.
Mail the author of the tutorial: fvalk at fvalk dot com (no spaces)
Click on the image above to see an example of results that can be obtained using the GroundMap program.
Geostationary weather satellites situated at an altitude of approximately 36,000 km produce imagery as seen from a fixed viewpoint in space, covering a large portion of a hemisphere on earth. Although such satellites provide a view of larger areas and continents, their vantage point implies that images suffer significant loss of resolution and perspective skew towards the Polar Regions.
Polar orbiting satellites on the other hand move through space in a so called sun-synchronous 800+ km high orbits, looking vertically down to earth and scanning narrow strips of the globe as it revolves beneath them. The advantage is that true geographic information is provided directly underneath the satellite and that the deviations that occur to the left and right sides of the scan line follow a straightforward mathematical model of a sphere (disregarding the effects of the Earth not being perfectly spherical). Deteriorating resolution and perspective skew towards the ends of the scan line are thus limited and fairly easily corrected.
The disadvantage of images derived from Polar orbiting satellites however is that during any one pass only a limited amount of data is provided corresponding to the width of the scan and the length of time the data from the satellite can be received by the ground station. Furthermore, no two single passes will have the same vantage point towards an area observed. Therefore two consecutive passes can never be joined without serious errors over larger areas unless significant effort is invested and consequently, larger areas exceeding the scan width cannot easily be shown in one single image.
David Taylor has developed a tool that addresses this issue in a clever but straightforward way: GroundMap.
This program has seen several phases of development and enhancements are likely to follow. The current tutorial describes the major new release GroundMap Version 2, available from January 2006. The program can be found at David Taylor's site - see point b) below.
As I have been working extensively with this program and have defined a major portion of the functionality it seems opportune to provide some guidance for its use.
What does GroundMap basically do?
It accepts three types of image files as input and then maps the relevant image pixels of that file towards a pre-established output projection. The input files that are accepted are:
A. Image files associated with a location file <filename>.lcn. Such location files are associated with images derived from polar orbiting satellites, both APT and HRPT.
B. HRIT, LRIT or FSD image files that are distributed via EUMETCast. These image products come from geostationary satellites.
C. Images in Plate Carree projection format. These images can come from any source, but most attractive at the time of writing are those made by NASA’s Earth Observatory and made available for free to the public.
The following section will describe using the GroundMap program in general and of working with any of the three input modalities.
What is needed to get started?
a) Availability of a modern good spec PC running Windows XP or 2000.
b) The latest released version of the program, which is available for download at: http://www.satsignal.net/ . Choose "Satellite Tools" followed by "GroundMap". Unzip and install in any directory you like, but a dedicated GroundMap seems practical.
c) Make sure that the required libraries are in place for the program to run without problems. These can be obtained as a full bundle from http://www.satsignal.net following Software => Runtime Libraries => Library Bundle => Download the runtime library bundle, unzip and install.
d) Source or input images as defined above as well as an adequate image processing package such as PhotoShop or PaintShop Pro.
A - Remapping images that come with an associated location file.
A location file contains the actual coordinates of a subset of pixels of a particular APT or HRPT image. Polar orbiting satellites view a certain area during two or more passes at a different time and at a different angle. Having the coordinates of the pixels of a pass stored in the location file then allows one to map pixels of different passes and thus viewpoints to a unified projection, effectively merging image coverage of different passes onto a single canvas.
To accomplish this, follow these steps:
1) Generate either an APT or HRPT image with the associated *.lcn file. In the case of APT, the produced location file will usually reside either in the same folder as the image data (when saved from SatSignal) or in the WXtrack Results folder if generated by WXtrack. It is essential that both the image file (bmp or jpg) and the location data file have an identical name and that they are located in the same folder. For practical purposes and to eliminate confusion, copy or move the files to a dedicated GroundMap working directory. In the case of HRPT the location file may be generated automatically in the target location specified in the program, next to the image file, or you may need to select the Save Location Data Menu item. It is ESSENTIAL that no geo-correction is applied in the process of generating the image as that would result in mapping relocated pixels to the wrong target place. The location file is hence only produced when no geo-correction has been applied. It should be noted that the filename format must follow the convention as normalized for SatSignal, WXtrack and ReadHRPT (in case of APT and HRPT respectively) so that the later creation of a proper Map List is assured.
2) For convenience sake we assume that the created files have been named <N18-test.bmp> and <N18-test.lcn> and that they have been placed in folder C:/GroundMapWorks/
3) Open the program and click “Options” and “Output size”. In the Standard version sizes can be selected in the range from 600 x 480 to 3600 x 2700 pixels. Initially it is wise to select ‘small’ (800 x 600 pixels) in order to get a quicker preview of the results of different settings options. Larger sizes require longer processing times and are therefore better used when making final image output.
4) Under “Options” choose “Location”. Here you are offered the choice of area to which youwould like to remap. The program allows you to select from a range of preset views such as UK, Europe, USA, Australasia and the Poles as well as a user defined view. Basically choosing any of the presets (assuming that the input data corresponds to the region requested) provides you with a shortcut to remap to a region with the advantage that results can be consistency reproduced over time. Setting a location implies that the same canvas is always used, whether one adds to an existing image during the same session or maybe weeks later. In most cases, once a preset is available for the region that can be received from the home location, there will be little need to change that setting anymore. The User setting option is available to accommodate the non-preset locations as well or to be able to deviate in the size of the standard coverage area. This also comes in handy for those consecutive passes that do not cover your local region but have been downloaded, for example, from CLASS or obtained via EUMETCast. More on User settings follows in the description of “Settings” in point 6a below.
5) Under “Options” select “Map projection” and choose any of the indicated clickable projections, which are self evident in their description. Refer to the section Projections below for further clarification on the different types of projections.
6) The last field that can be selected under “Options” is “Setup”. Three tabs are available: General, Overlay and Mapping.
a. General: In case “User” had been selected in point 4) the desired map settings can be defined here. In order to do so, activate “Use my map settings” first. When determining the field of view you want to remap to, you must take into account that Meridian and Parallel relate to the centre point of the output field. The notation of the numerical values of longitude to be entered is allowed to be -180 to 180 degrees or 0 to 360 degrees. Latitude values are allowed in the range of 90 via 0 to -90 degrees. Depending on the type of projection selected earlier, the fields “lon span” and “lat span” are available or greyed out. The span determines how many degrees the field of view will be. For fixed ratio projections the latitude span is used to define the view, for non fixed ratio projections such as Cylindrical Equidistant zoomed and Mercator zoomed, the height and width of view are selectable at will. The values entered for Span correspond to the total output view and do not count twice from the centre focus point as defined. Lastly, the Picture Margin option on this tab allows you to set the margin in pixels for the image. The purpose of this is to correct for corners that may go amiss during the mapping process. Initially leave this at 25. The settings on this tab can be saved for future reference to a location of choice by selecting “save user”. Please note that the saved settings refer to this tab only and do not include settings of size and type of projection.
b. Overlay: This rather self explanatory tab provides the choice to select overlay data such as Boundaries and Grid lines to the output image. The line width and the colour are selectable as is the spacing between latitude and longitude grid lines. Boundary data can be based on the countries.dat file or the gshhs_i.b file or similar and the file location can be defined at will. In the start-up process it is advisable to have borders and grids active (white colour, as the background is black) to navigate better duringthe process of image generation.
c. Mapping: this tab offers a quick mode of mapping (“use fast mapping”) which is the right choice for high quality data. If selected, the coordinates of fewer samples will be used in the remapping process than is the case when this option remains unselected. In case noisy input data is used the slow mode of operation, which reads many samples more, provides much more accurate output images. For larger images the processing time however increases significantly. The Quadrant Mapping option on this tab belongs to the Pro version of the program and is explained later on in the tutorial.
7) Having made all the initial settings, we can now move to “Input mapping” on the main screen and select “from .lcn file”
8) Choose “File” on the main screen and select one out of two possibilities:
a. In case no file map_list.mpl is available in the directory C:/GroundMapWorks/ (this is our example working area) then choose “open location data” and select the example N18-test.lcn file. A Map List file will automatically be created. On the “Source picture” tab the original image to which the location data matches will be displayed and on the “map image” tab a remapped image will appear.
b. In case the map_list.mpl file is available in the directory C:/GroundMapWorks/ then choose “Open Mapping List”, followed by file =>Open location data. Also in this case the source image is available on the “Source picture” tab and the remapped image on the “mapped image” tab. The map list stores the image file names for which a location file is available and these names can be seen (and selected) on the main program page, Picture tab.
9) You will now have an image on the “Mapped Image tab” showing the original image re-projected to the projection as specified earlier via the Options tab. The grid lines and borders show for the whole specified preset region while the image covers the canvas fully or partially (provided that the image corresponded to the region specified). If you now look with Explorer in the GroundMap working directory, a new file will have been created there, called remapped.bmp. This is an automatic save of the remapped image created in the previous point. The name of the save will always be the same, hence any other remapping instruction will overwrite the file preventing clogging the directory without need. If the image is OK and should be kept, the action ’file > save image’ provides the option to save to a specific location with a custom name for further processing. Note: Only one single image can be remapped at a time. Combining images belongs to the realm of image processing programs.
10) It may be that the overlay created on the output image is misaligned with the image itself. This may occur when during the creation of the APT or HRPT input images Kepler files were used of a different date than of the actual image recording. There are nudging arrows at the bottom of the main screen to adjust the position of the boundary lines on the image with a precision of seconds. In case large discrepancies (minutes) need to be dealt with a course adjustment (double arrow) is available. The whole overlay can be switched on and off without recreating the image using the radio button right of the double arrows. Hovering over the text left of the arrows provides information on the files currently in use and the zoom function allows one to see the whole output canvas on the screen or the image in true size.
11) The other drop-down menus are: Picture, Edit, View and Help. As mentioned in 8b the ‘Picture’ tab will list the images that have been processed and are within the directory where the .mpl file resides. Once several images are in that list, clicking the desired one immediately initiates the generation of a re-map of that image without the need to call for a location file first. The ‘Edit’ tab allows you to flip wrongly oriented images to the desired position without going through the full remapping calculation process. ‘View’ comes in very handy to see relevant image information such as size, projection and coverage. The Help function offers Registration and program version info. Real sources of Help will be found in the tutorial and the GroundMap.txt file in the Program directory.
As a general note: it is advisable to do all the experimenting at a small output size until satisfied, only then to proceed to select the final larger size. Correcting the overlay to match the image contours perfectly is best done at the larger magnification because differences may not be apparent at small scale but highly disturbing at large scale.
Once two (or more) passes have been re-mapped onto the same canvas settings and the resulting files have been saved, it is only a matter of being handy with your favourite image editing program to join images together, merge , use masks for fluent transitions and any other manipulative action. That is an art in itself and I will not go into further detail here.
Up to this point the working of the GroundMap program has been described in the Location data mode only (point A in the introduction). We will now proceed with the EUMETCast derived images from point B, albeit without all the generalities of the program outlined above.
B - Remapping HRIT/LRIT and FSD images as distributed via
EUMETSAT makes image files of various geostationary satellites covering the whole world available to license holders via the EUMETCast distribution system. Images of each of the distributed satellites (Meteosat 5, 7, 8, Goes 10, 12, MTSAT-1R) come in a fixed format, which facilitates easy incorporation in the GroundMap environment. It is justified to ask the question why opening geostationary image files in GroundMap and not to use the ample functionalities GeoSatSignal offers. The reason is that GroundMap offers a number of extra projections that can be used for very specific applications or demands (see section Projections below) which would unnecessarily complicate the GeoSatSignal program when included there. The access to all of the different output sizes available in GroundMap and access to the Pro functions are other reasons.
In order to access the EUMETCast option go to the “input mapping” tab on the main screen and select any of the lower portion options (GOES-E to MTSAT-1R). The program will then ignore looking for location files or files with a specific name format and instead offers the choice to specify file name and location of the image to be input by responding: <You can now use the: File, Open mapped image… menu to load an image>. Press OK and navigate to the desired file to be opened as instructed. The file used for input will now be remapped to the output canvas specified in accordance with all settings specified for the occasion. Care is to be taken that the file type specified (for example Meteosat 7) coincides with the file actually used for input, as GroundMap does the remapping based on the file type it has been told. Using in the example a Meteosat-8 image will provide a remapped image but this will be useless do to the difference in input image size.
From this point on, one can use all GroundMap functions on the geostationary derived input image. The flip image function should of course be ignored and the Picture tab will not be available.
C - Remapping images that are in Plate Carree projection format
The world has been pleasantly surprised when the Earth Observatory group of NASA made highly detailed imagery of the whole world available from which all cloud cover has been removed to a very high degree. The imagery, dubbed Blue Marble, can be had in different resolutions, ranging from 8 km per pixel to a stunning 500 m per pixel. What is more, in order to show seasonal effects such as extent of ice cover and vegetation, all resolutions have been created for each month of the year. One also has the option to have topographic data included and/or bathymetric data. The projection in which it comes is standard Plate Carree and thus highly accessible. GroundMap Version 2 accepts this projection as input and offers its wide range of remapping options for the creation of maps of all sorts for a multitude of applications.
It should be noted that images covering the whole world in high detail will be extraordinarily large. People will therefore be restricted to handling images not exceeding the limits set by the current 32 bits PC systems. In practice the maximum manageable image at present is 10k x 10k pixels RGB, but this makes a strangling demand on memory resources and processing capacity.
In order to cope with system and program limitations, the program has been adapted to allow the input of part of the total world view at a higher resolution without compromising the restrictions. The following provides a reference of maximum permissible sizes, but be aware that for agility of work it is better to stay well below these maxima.
The East and West hemisphere need to be created once by the user, as is the case for the four quadrants A, B, C and D. The maximum resolution to be obtained with the hemisphere and quadrant presets is therefore 2 km/pixel.
There is however another option available in GroundMap, that gives one access to work with the highest resolution data available at present: 500 m per pixel. This is the “Plate Carree arbitrary X-Y coordinates” option. This option allows the user to create any personalized coverage area at best resolution.
Working with remapping Plate Carree based images is quite straightforward once the input files have been generated. First choose ‘input mapping’ on the main screen and then select one out of the available Plate Carree options. You will be requested to select an input file that meets the requirements (restricted size, aspect ratio, full image, hemisphere or quadrant). Once selected, remapping takes place to the canvas specified in ‘Options’.
A last word on the arbitrary X Y mode: the name of the input file to be created by the user must contain the coordinates of the four corners of the input image. Longitude values must be 3 digits plus e (east) or w (west) and latitude values must be two digits plus ‘n’ or ’s’. When selecting the arbitrary function the program thus looks if the file has a name such as 045w80n041e35n.jpg. If so, than the remapping can take place as with all other remap options. It will be clear that the accuracy of input file creation is decisive for correct output results.
GroundMap - Pro Version
The pro version offers a number of extras that a more professionally oriented user requires.
To complete the GroundMap Version 2 Tutorial an overview of available types of projections is attached.
· Cylindrical Equidistant (full): the version used in GroundMap shows the full earth as if it were divided in rectangular uniform boxes of 10x10 degrees with the Greenwich meridian being at the centre. Scale (horizontally and vertically) is true along the central line (the equator). Shape and scale distortions increase near points 90 degrees from the central line. This projection is also known as Plate Carreé projection.
· Cylindrical equidistant (zoomed): shows the same type of projection as above, but limited to the region specified under Span. As there is not such a thing as “correct” projection one should visually choose the best width-height span in relation to the user’s objective. Setting lon span for example to 10 and lat span to 80 will work but the image will not have many practical uses. Proportionality lon : lat = 9 : 5 is most likely the right choice here.
· Azimuthal Equidistant: provides a standard, fixed, coverage area of the selected preset region (appr. 4,000 km North-South), with all points on the globe that are at the same distance from the centre point depicted as a circle around that centre point. Distances measured from the centre are true, whereas distortion of other properties increases away from the centre point. It is in many cases the most natural looking type of projection when the radius is not too large.
· Azimuthal Equidistant (wide): The same as before but with North-South coverage of a bit over 10,000 km. In both Azimuthal equidistant projections the Lon Span setting is disabled as in fixed ratio projections Lat span is used to determine the field of view.
· Mercator full: shows the full earth again as if it were divided in rectangular boxes with the Greenwich meridian being at the centre. The longitudinal separation is a constant linear, whereas the separation between latitude lines increases as a (aTan) function of distance from the equator. Scale is true at the equator or at two standard parallels equidistant from the equator. The projection is specifically applicable for use in marine navigation because all straight lines on the map are lines of constant azimuth, the bearing.
· Mercator (zoomed): shows the same type of projection as above, but limited to the region specified under Span. As there is not such a thing as “correct” projection one should visually choose the best width-height span in relation to the user’s objective. In line with the observation made under Cylindrical Zoomed the proportion between longitude and latitude that is most likely to be used in practice is 9 : 5.
· Orthographic: A perspective azimuthal projection in which the projecting lines, emanating from a point at infinity, are perpendicular to a tangent plane. The definition implies that only one variable can be set in this case: the latitude span. Proportionality is always assured. Orthographic projections are used for perspective views of hemispheres. Area and shape are distorted. Distances are true along the equator and other parallels.
· Linearised orthographic: A linearised variant to the regular Orthographic projection resulting in a somewhat more curved plane. This variant requires significantly more computing time and is therefore slower to produce.
· Polar stereographic: Stereographic projections are used for navigation in Polar Regions. Directions are true from the centre point and scale increases away from the centre point as does distortion in area and shape. As is the case with Orthographic projection only one variable can be set: the latitude span.
· Lambert conformal conic: the meridians are equally spaced straight lines converging at one of the Poles with angles between meridians being less than true angles. The parallel are represented as unequally spaced concentric circular arcs centred on the chosen Pole of convergence. This results in a projection that has true scale along the one or two selected parallels, constant scale along any parallel and the same in any direction. Thus there is no distortion along the standard parallels.
· Gnomonic: points on the surface of a sphere are projected from the viewpoint of sphere’s centre to a plane tangent to a point on the globe (usually the South or North Pole). Only one hemisphere can be mapped this way at a time. Great circles are mapped to straight lines and provide a view as formed by a spherical lens.
· Van der Grinten: the boundary of the whole view is a circle, in which the meridians and parallels are circular arcs with the exception of the zero meridian and the equator, which are straight lines.
· Great circle: this is in fact the limiting case of the azimuthal equidistant projection with a full 360 degree field of vision. The projection is often used for long distance radio communication as bearing and distance are always true with area and shape being progressively distorted over distance.