Porcentaje de polvo de bambú

A check of the accuracy of the DTM was carried out using the flood levels surveyed. The process to obtain these flood levels is discussed in Chapter 5 of this dissertation. These levels were able to be compared with the DTM as they were either ground levels or using the depth of flooding at the point to obtain the ground level.

3.3.1 Field Survey Details

The field survey measured levels to +/-0.02 m. This survey also did a check on the difference between the digital terrain model survey datum and the datum that the global positioning system uses to put the levels into the digital terrain model datum. This was done by checking the levels of the photograph control mark points in the flooded area (see Figure 3-5). This check used 10 of these points and 3 Lands and Survey control bench

marks in this area. After the adjustment of the two datums the root mean errors of these points from the adjusted datum was 0.008 m with a maximum of 0.017 m. The difference was expressed as a constant and a tilt in both the north and east directions.

The survey was carried out by the surveyor standing at the flood level position with the antenna still on his back. This caused another source of error in the survey as the surveyor was not standing at exactly the right level for the survey point compared to placing the antenna on a pole. This error was estimated to be about +/-0.02 m. The survey was done in this manner as it saved time and also the error was well below the other errors on the project. Adding all these errors together (using the square root of the sum of squares of their errors ) gives an error of +/-0.029 m for the field survey data.

These levels were either at ground level or, if they were not, their height above ground was measured to give the ground level at that point. The levels used were assessed to ensure that the ground level could be established. The point of the flood survey was primarily to obtain flood levels and secondarily to check the Digital Terrain Model. In some cases the depth was not measured. The depth was found in these cases either by:

a) using the difference between the normal height of the GPS antenna of 1.79 m (which was used for levels on the ground) and the lesser height of a level where it was above the ground. b) measuring the depth in the field.

There were 8 depths inside buildings that were not used as one could not measure the ground surface under the buildings. Also a further 9 points that had two levels at the same point were not used for the comparison.

3.3.2 Comparison of Surveyed Ground Levels with Digital Terrain Model

3.3.2.1 Statistical Analysis of the Data

These levels were compared with the ‘TIN” surface level that the DTM gives at that point. The full comparison is given in Appendix 3. This lists the points, the depths and the digital terrain model level for each point used in the comparison.

The results of this show that the mean level of the DTM is 0.044 m lower than the average level of the ground for the points surveyed. The standard error of the differences about this mean was 0.264 m and the standard error of the mean 0.015 m.

Figure 3-4 is a plot of the cumulative distribution curve for these data. Comparing this with the specification requirement that 68% of the values (one standard error) for any position on the DTM had to be within +/- 0.3 m showed that 220 or 77% of the values were within this range. This also shows that there were 25 (8%) points with DTM levels over 0.3 m higher than the ground level and 44 (15%) DTM levels below 0.3 m below the ground level. This statistic with the standard error under +/-0.3 m reflects the requirement that the standard error for the points had to be within +/- 0.2 m and +/-0.3 m for any point on the DTM. The source of largest error was for a point under a tree near a terrace where the horizontal position can make a large difference in vertical level. The latter means that it would be difficult to accurately define the position of the terrace at this point.

All the differences over 0.5 m were investigated and this showed that these were mostly due to errors reading the values or in making the adjustments using the photogrammetry equipment. In several cases a low terrace or feature was missed.

Another possible source of the differences was the levels on the fence lines. The ground at the fence lines is sometimes higher than the ground in the surrounding paddock. The reason for this was that the paddock is ploughed and this causes a small terrace at the fence line. The photogrammetrist did not take levels at fence lines for this reason. This was not

the level ground and not on any windrow at a fence line. This was generally the case as at one point two levels were surveyed for this very reason. A field check of the survey positions that I could remember shows that only one case could be found where the level of the ground was above the adjacent ground. This point was 16 1994 and a subsequent adjustment was made. This changed the difference between the point and the DTM surface from 0.105 m to -.245 m. This latter however matched the differences with the other points in this area far better (see

Figure 3-4).

However before another field survey was seriously considered another analysis was done without the fence data. This gave a mean difference of 0.05 m and a standard deviation of 0.269 m. This result showed that the levels not on the fence lines are actually slightly higher relative to the DTM than the levels on fence lines. Therefore no further work was necessary.

Figure 3-4. Cumulative Distribution Plot of the Differences between the Surveyed Levels and the Digital Terrain Model Levels (before adjustment)

-1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.10 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.91 0 36 72 108 144 180 216 252 288

data points in ascending order of difference (302 values) differences in level (m)

True ground level minus the DTM level

Mean = 0.044 Standard Deviation about mean = 0.264

3.3.2.2 Area Analysis of the Data

Arcinfo provided another check of the errors as the program could do a ‘TIN” surface of the differences in the DTM ‘TIN’ levels and surveyed levels. Figure 3-5 (see attached folder of plans) shows the contours of these differences.

This shows high and low areas and only a few random differences. Areas over +/- 0.3m in difference have been highlighted and there are few cases of random differences over this amount. The factor of particular significance was that the photograph control points were well out in a few cases. The worst difference after adjustment was 0.414 m at photograph control point number 8. This was in the largest area (area 1 of Figure 3-5) of differences over 0.3 m. It would therefore seem that these worst areas could be improved if the model was calibrated using these field data.

The actual difference surface could not be used as it contains irregularities as it used the differences from every point. Any adjustment surface needed to smooth these differences with the average difference for a point in high and low area to give a gradually changing surface.

The photogrammetry personnel were approached with this information to see if it was possible to improve the accuracy of the overall levels with this information.

They examined these differences and recommended that the levels be adjusted by 0.044 m as this was the only significant difference. The standard error of the mean was 0.015 m hence the 0.044 m was significant at 1 %.

However they considered that the high and low areas of the DTM were not significant enough to change and as the DTM was within the specified error did no more work on it. After receiving this information, the worst high areas were investigated (Areas 1, 2 and 3 of Figure 3-5). They consisted of 21 points, 17 points and 14 points (see Appendix 10). The analysis of the points showed that their mean was significant at 0.1 %. This meant

points and area 6 of 4 points that are significant at 1 %, where similar adjustments could be done. All the other areas consist of only 1 or 2 points. This was too small a number to base an adjustment upon.

These adjustments were not done as the surveyor had more experience in this field. It was decided to use the results of the analyses to investigate whether allowing for the differences in the DTM and the flood levels would improve the standard error of the results. This would be done in a similar manner to the analysis of the differences of the ground level using both the statistical analysis of the data adjusted for the difference in the DTM and surveyed levels and calculating a ‘TIN’ surface of the differences. It was expected that these would confirm the value of adjusting only part of the DTM and any further decision could be made at that stage.

Figure 3-5. A ‘TIN” surface of the differences between the field survey of the ground levels and the DTM ‘TIN’ surface (after the 0.044 m adjustment) (see drawings at the end of the Volume).

4. GATHERING OF OBSERVED FLOOD PLAIN FLOOD LEVEL