DISCUSSION
IMAGES OF GRAVITY FIELD OF INDIA AND THEIR SGLIENT FEATURES by Y. Sreedhar Murthy, Jour. Geol. Soc. India, 11.54, pp.221-236, 1999.
(1) Saif ud din, Remote Sensing Applications Centre, Aligarh Muslim University, Aligarh comments:
I congratulate the author f a r taking up this new approach of ATD, but certainly some modifications are to be made to make it more efficient and accurate. A GIs approach would be a better option if the ATD'is used for digitization and the map is stored in aGIS compatible format.
If integrated with the image data, the efficacy of the methodology will certainly improve. The
following observations need to be incorporated to improve the efficacy of methodology in the paper:
1. Auto Trace and Digitize (PLTD) shall consider points on the contour lines but how many?
Does ATD take into consicieration the X and Y coordinates of all the pixels that contour line passes through?
2. If there are closely spaced contours and more than one contour passes through a single pixel, then how that pixel \will be assigned a Z value?
3. The Z value for each pixel has been calculated by the author ("Digitization is pixel wise and hence the inherent accuracy of contour map is preserved"). If the author means that all
the pixels through which the contour passes have been digitized, then probably the digitized line may not be a smooth line, and instead may be a hatched line.
4. It has not been mentioned whether digitization was carried out in the vector or raster mode.
5. How the colour assignment was done? Was Boolean or fuzzy logic adopted, since a major part of the map is to be extrapolated?
6 . How the individual pixel was assigned a Z value? If done by extrapolation of point data obtained from digitization of contour lines, this is not a recent approach, since people have been doing this for years.
7. Probably, using the ATD designed by the author and storing it in a GIs compatible format may be easy to integrate arrd overlay the different types of maps.
8. The two contour lines may be considered to form a close polygon and the gradient within the polygon is to be established by extrapolation. If the assignment is made using fuzzy logic, it should be more accurate.
Y. Sreedhar Murthy, Centre for Exploration Geophysics, Osmania University, Hyderabad -
500 007 replies:
We hope that the clarifications given below to the queries raised by you will clarify your doubts:
1. The number of pixels contained in a contour line depends upon the resolution at which the map had been scanned. If a given map is scanned at a higher resolution, a line will contain more pixels when compared to the one scanned at a lower resolution. For typical paper maps, scanning resolution from 150 DPI to 600 DPI is sufficient and normally any higher scanning resolution is not needed because most of the paper maps are not produced at that high resolution. Usually the lines to be traced must be at least one pixel thick but must not
line. ATDS does not consider all the points on a line but only the XY coordinates of the pixels, which can be considered to be the midpoint of a line at the given point.
2. Theoretically, no two contour lines must cross each other. Contour lines may be drawn as close as possible, but they must never share a common point. If a map has such closely spaced contours resulting in two contours sharing the same pixel, it may have to be rescanned at a higher resolution or resampled to a higher resolution, which will increase the number of pixels representing their adjacency. Usually when you resample an image at 150 dpi to
300 dpi, a line of one pixel thick becomes at least 2 to 3 pixels thick. Hence distinct identification of the two lines will be possible, thus solving the problem of assigning a correct Z value to the pixels concerned.
3.
Z
values are not calculated. The user has to input the value to be assigned to a selected line when prompted by the programme whose value is equal to the contour value. ATDS doesnot consider all the points on a line but only the pixels which can be considered to be the midpoint of a line at the given point. Since only the centre of most pixels is being considered, the traced line resembles the original line very closely. The digitized data consists of points one pixel apart. When you plot this data using a point-data posting programme or when the data is used to create a vector line, the resulting line is also smooth and not hatched.
4. ATDS has a concept of grouping points lying on a line under a common line ID. Every line ID is associated with a third parameter, Z value, which stores the value of the line, The following diagram illustrates the file structure:
Line, ID Zvalue, Zval XI,YI X2, Y2 X3, Y3 ... ... X4, Y4
Line, 1 Zval, 1000 78.15, 20.12 78.14, 20.12 78.13, 20.13 ... ...
78.85, 20.82
Line, 2 Zval, 1 1 00 78.35, 21.22 78.35, 20.23 78.36, 20.24
....
....
78.45, 20.98
The output obtained by using the Export as CSV command creates an ASCII file i n Comma Separated Variable format, which contains coordinates (X and Y) and third parameter
(z value) information about each digitized pixel. For example,
6 8 4 DISCUSSION
GeoSoft, ILWIS, etc. This kind of output is neither in vector nor in raster format. It may be
called as Point Data and can be used for creatihg a vector format file. ATDS has an option to export in Autocad DXF format also.
5. The digitized data was used as input data for creating a regularly spaced grid. Minimum curvature interpolation method was used for generating the grid file. This is one of the most widely used interpo1,ation methods in the earth sciences. The interpolated surface generated by minimum curvature is analogous to a thin, linearly-elastic plate passing through
each of the data values witl~, a minimum amount of bending. Minimum curvature generates the smoothest possible surface, while attempting to honour original data as closely as possible. The colour images and shaded relief images were prepared using these grid files.
Image maps use different colours to represent the third parameter (which may be topography, gravity values, magnetic values, etc). Colours on image maps are associated with percentage values. The percentage values are in relation to the minimum and maximum Z values from
the grid file. The colour associated with 0 percent corresponds to the minimum Z value from the grid file, and the colour associated with 100 percent corresponds to the maximum Z value in the grid. The digitized data was only interpolated. Extrapolation was not done (not usually done in earth sciences), since we are only considering regions where original contour data is avaiIable. 19s is seen in the all the image regions over Kashmir, parts of Northeast India have not been included in the image, since the original contours were not available for these regions.
6. The digitized data was used as input data for creating a regularly spaced grid. Minimum curvature interpolation method was used for generating the grid file. No claim was made that interpolationlimaging was done for the first time. In fact, due acknowledgement has been given to the various works in this regard. We have to maintain the fact that the gravity series of India has been presented in digitial form for the first time.
7. ATD is a user friendly software for digitizing contour maps quickly. The output from the programme can be used in any programme and on any platform. The output is compatible with some GIs packages 1.ike ILWIS. ATDS can also export in Autocad DXF format which can be imported by ARC Info and other vector packages. We are working on making ATDS output to be completely compatible with GIs formats.
8. The idea of forming closed polygon is usually employed in vectorizing programmes and a digital elevation model is objained by using methods like triangulation. ATD methodology is different from other methods in the sense that it does not do any vectorization. The point data generated can be interpolated by any of the following interpolation methods:
Inverse Distance to a Power
Kriging
Minimum Curvature Nearest Neighbour Polynomial Regression Radial Basis Functions S hepard's Method
Triangulation with Linear Interpolation
Although we have not used fuzzy logic for interpolation in our work, we are willing to experiment with it.
comments:
In the first place the colour images, their usage, and limitations are not new to the field of earth sciences. However, in view of the oversimplification of converting the contour form of Bouguer gravity, free-air and isotatic anomaly maps of India into the coloured images and overstretching of inferences of surface and hidden geological features from the digital images (Sreedhar Murthy,
1999), 1 would like to raise a few important points.
I . It is necessary to understand certain important restrictions in converting gravity field contour maps into colour images: (a) The very collection of gravity data is at discrete spatial coordinate points on the surface of the earth, (b) anomaly contours are inherently imaginary i n nature and hence are approximated data set, (c) again digitizing the anomaly contour maps (a polynomial form) need interpolations and these invariably create undesirable data called "noise", and above all, (d) the gravity anomaIies are essentially a reflection of geological mass variations, which are hidden beneath the surface of the earth. The conversion of digitized gravity field anomaly contours into colour images would differ from the conventional images formed from data like satellite, videography, TV etc.
The entire matter under "the method" is just a repetition of the earlier publication with a
I
misleading title "innovative methodology" (Sreedhar Murthy et al. 1998). It is not explicit what the x, y and z values are. Neither the earlier work (Sreedhar Murthy et al. 1998) nor the present attempt (Sreedhar Murthy, 1999) does say anything about "... edited and cleaned map ....""Digitization is done for every pixel": It needs certain clarity. It may be mentioned that the smallest physical unit which an image is able to record, represent or produce a discrete amount of light, is called a pixel or picture element and the pixel size defines the spatial resolution of the image (Teubu, 1993). The author does not seem to know about how many samples and grey levels are required for a good approximation. The resolution of an image depends strongly on these two parameters (Gonzalez and Woods, 1993).
2. It may be noted that although colour image processing is to some extent based on individual colour perception, as far as the feature detection is concerned, in both qualitative and quantitative sense, it does not vary from person to person, as the images are subjected to certain rigorous processing rules or techniques.
The author has correlated the known structures and tectonic features with the coloured images. These known structures were earlier revealed by laborious exercises of many researchers and correlated and confirmed by different geophysical data sets, particularIy seismic. Similar is the case with identification of Singbhum craton. The Bouguer gravity anomaly shaded images of India, presented by Ram Babu (Fig. 1,1999) and Sreedhar Murthy (Fig. I c , 1999), differ in many respects like the image of southern part of India, especially the Deccan Trap and Cuddapah basin.
In correlating the free-air gravity image (Fig.2b and c) with the geological features the author's attempt is extremely feeble, by saying
"...
resolves the broad features into distinguishable smaller features." In fact, detaiIed satellite imageries would reveal more explicitly the trends of surface topographic features.The author has also not spelt out the criteria for identifying geological fold belts using any of the constructed gravity images.
3. The very purpose of the entire exercise has been virtually nullified by making the contradictory statement "However, caution needs to be exercised in drawing conclusions about the features ,
6 8 6 DISCUSSION
based only on shaded relief images as these are illumination dependent.", since the correlation attempted here is based on shaded relief images.
Y.
Sreedhar Murthy, Centre for Exploration Geophysics, Osrnania University, Hyderabad -500 007 replies:
1. (a) It may be pointed out that collection of almost all geophysical data is at discrete (ordered or random) points. Data and digital images used in Digital Image Processing/Remote Sensing are also discrete. The data iippears continuous if a proper scale to resolution aspects used. When we are discussing a rrlap in the scale of 1 :5,000,000, it is not only unnecessary, but also redundant to ask for gravity observations as a continuous signal for the entire length and breadth of the country.
(b) As mentioned in the paper, the bulk of the geophysical data in the country is still in analogue mode and therefore needs to be digitized in the first instance before imaging is possible. It has been accepted internationally that digitization of available contour maps is a good alternative of obtaining digital data-sets in the absence of original digital data. It is definitely much cheaper and less time consuming than collecting the data for the entire country once again in digital mode. Gravity and magnetic anomaly images of the entire continent have been prepared from data digitized from contour maps. Contouring does involve some amount of imagination, but i F the original data set is as good as the ones used in the compilation of the Gravity Series of India (NGRI), the degree of imagination is considerably reduced.
( c ) Interpolation is used frequently in all imaging applications. Interpolation methods are acceptable scientifically and statistically, as long as the quantity of original data points exceed the interpolated points.
(d) We can consider the gravity anomalies as signals generated by the subsurface features. Images are one of the marry methods of displaying these signals. Imaging provides more insight into the nature and characteristics of the signals recorded. The paper demonstrates that the features seen on the images correlate well with the known geological and tectonic features. Direct interpretation has not been made, but certain striking features brought out by the images call for a closer look.
A brief description of the tracing methodology has been given in this paper. The details of the methodology have been given in the two publications given in the references. More technical details have.been provided in the reply to comments made by Mr. Saif ud din.
2. I would like to stress that the present images correlate well with the known geological structures and features over Indian land mass. They have been dearly described in detail in the article. I would like to mention a vt:ry apt saying to clarify the point further. "The eye sees what the
mind identifies
".
To recognize the features on the images, one needs to have a knowledge of the Indian tectonics, geology etc.3. I do not see anything wrong in exercising caution while making geophysical interpretations and correlating with known geology.
References
GONZALEZ, R.C. and WOODS, R.E. (1993'). Digital Image Processing. Addison Wesley Publ. Co., New York.
RAM BABU, H.V. (1999). Gravity Image. Cum. Sci., v.76, no. 12, pp. 1533-1535.
SREEDHAR MURTHY, Y., GOVINDARAJAN, J. and BABU RAO, V. (1998). Contours to images -Part 1: An innovative methodology. Jour. Geophysics, v.19, 110.3, pp. 14 1-148,
SREEDHAR MURTHY, Y. (1999). Images of the gravity field of India and their salient features. Jour. Geol. Soc. India, v.54, no.3,
pp.221-235.