Private Bag 11008, Palmerston North, New Zealand
Telephone +64 6 351 8064; Fax +64 6 351 8032
Topographic map information with 20m contours exists at the 1:25000 scale for the whole of New Zealand. Is this data adequate for use in farm management models or is more detailed contour information required? In order to validate such map data how much ground-truthing is required?
In an earlier study, slope analysis had shown that digital terrain models (DTM’s) based on contour height data overestimated areas for flat land (zero slope), particularly in the regions of ridge tops and gully bottoms, would RTK point heights remedy this?
Maps from three sources of topographic information for the AgResearch Ballantrae research hill farm, including the national 1:25000 survey, were compared with measurements made with an real time kinematic differential geographic positioning system (RTK) unit. In particular, the locations of specific topographic units and fence-lines were compared.
Results showed significant misalignment of the earlier topographic maps compared with the RTK recordings. Map alignment problems prevented the addition of RTK point heights to existing DTM’s for slope analysis. Combining map information from existing sources with RTK survey data highlighted original map projection errors. It was also evident that the 1:25000 survey maps were insufficiently accurate to align with major topographic features.
Commercial hill farmers, buying a 20m map, are therefore faced with the need for considerable ground-truthing GPS measurements in order to produce a GIS map accurate enough to be useful for management purposes. It is recommended that a better alternative is to derive new maps from aerial photographs in conjunction with a series of RTK GPS ground-truthing measurements.
GIS, RTK, contour, hill farms
Topographic map information with 20m contours exists at the 1:25000 scale for the whole of New Zealand. Is this data adequate for use in farm management decisions or is more detailed contour information required. An earlier study (Costall et al., 2001) showed that there were some discrepancies between commercially available 20m contour maps and more detailed mapping sources available for two AgResearch hill farms. Also in the earlier study slope analysis based on triangulated irregular network (TIN) models using “ARCINFO” software showed that contour based TIN’s tended to over estimate areas of flat land. The question is how accurate are the various sources of map information, and in order to validate purchased map data how much ground-truthing is required? Also could the addition of RTK derived point heights to TIN’s improve the slope analysis result from ARCINFO?
Results from a study conducted at the AgResearch Ballantrae research hill farm are presented here
Map information for the Ballantrae research farm was compiled from three sources and stored in ARCINFO databases. TIN models were used for the extraction of slope and aspect statistics. These map sources were: -
- 5-metre contour map produced in 1974 by the Photogrammetric Branch of the Department of Lands and Survey. Mapped in North Island National units (yards).
- 20-metre contours, extracted from the New Zealand 1:25000 Topographic map series
- Property boundaries and internal fence lines originally produced by Dept of Lands and Survey and supplied in digital format by Survey Services Hawkes Bay Ltd. Mapped with Hawkes Bay Circuit (HAWK) coordinates.
- Point height data collected with a ‘Trimble” real time kinematic differential geographic positioning system (RTK)
The 5-metre contour lines were digitised from the original contour map using ARCINFO. Any errors were removed. Subsequently 5, 10 & 20 metre interval contour maps were created.
Map information was also created for property boundaries, fence lines, streams, point heights, roads and tracks from map information extracted from both the original 5-metre contour map and the maps provided by Survey Services Hawkes Bay Ltd, all maps were converted to HAWK coordinates.
The point height data collected with the RTK system was from specific features that had been identified earlier as showing misalignment when the above data sources were compared. Point heights were recorded around the perimeter of several paddocks and at points within that identified gross changes in the landscape. A small area was also mapped with an RTK rover with point heights being recorded every two metres. Two Trig stations in the area were identified from the NZ Geodetic database; and measured to calibrate the RTK readings. The trig stations used were rated as 2nd and 3rd order stations, in terms of co-ordinate accuracy within the NZ Geodetic database. All data was adjusted to match the co-ordinates for the two trig stations measured, and converted to HAWK co-ordinates.
The RTK data was converted into ARCINFO coverage information using the same co-ordinate system as the existing map information to compare with data from the other map sources.
Results and discussion
The RTK data was taken as representing the most accurate source of map information for comparison with other data sources. The un-calibrated RTK data collected had an error of approx. 3.5 metres for easting and 1 metre for northing calculations when RTK measured values for sampled trig station easting and northing were compared with those from the NZ Geodetic database. The ellipsoidal distance measured between Trig station “CCVII” and the base station which was located on a pre identified survey mark “Anemometer”, was 2964.287m based on RTK derived coordinates and 2964.298m based on the survey control maps supplied. The resultant variation being only 11 millimetres between the two.
When map information for fence lines sourced from the Department of Lands and Survey and RTK surveyed point heights for the same area were compared, after conversion to the same coordinate system, several errors became apparent (Figure 1). The two sources showed a misalignment error of approx. 26 metres.
The 20-metre contour data sourced from New Zealand 1:25000 topographic map series was significantly out of alignment in terms of recognisable topographic features. This became apparent only when fence lines and contour lines derived from the Ballantrae 5 metre contour map were overlaid. Recognisable features like hill tops were misplaced by approx 50 metres (Figure 2).
Point height data from the original 5m contour map also differed from the heights obtained with the RTK system.
Combining map information from various sources is not an easy task but all the errors can be corrected given correct coordinates for ground control points. This may be more complicated and time consuming than actually generating new maps as a basis for creating digital elevation models.
Slope analyses conducted by (Costall et al., 2001) using ARCINFO software, showed that areas deemed to be flat i.e. with zero degree of slope and with no aspect were over represented, particularly by ridge tops and gully bottoms. Point heights collected with the RTK were expected to help alleviate this problem but matching the coordinates and elevations for the various data sources and identifying which map sources contained the errors was difficult and extremely time consuming.
Slope analysis using RTK point height data only, eliminated this problem with the accuracy dependent on the intensity of point heights input into the digital elevation model (DTM). A comparison of slope analysis for an area within the Ballantrae research farm based on two different contour intervals and RTK point heights is shown in Figure 3. This shows that whereas the earlier analysis resulted in significant areas of flat land, the RTK point height example eliminates this category completely. The RTK based example had no land defined as flat compared to 25% & 17% for the 5 & 20m models respectively. Point height data used in a DTM will only define areas as flat when they fall within three coordinates having the same height.
- The NZ 1:25000 20m contour topographic maps are too imprecise for use in most hill farm management decisions because of displacement and distortion of topographic features.
- Combining map information from various sources with contrasting projections may not be a viable option because of alignment problems.
- Slope analysis derived from contour maps using ARCINFO software will overestimate the proportion of flat land present. Point heights derived with an RTK, in conjunction with contour maps will improve accuracy.
- Commercial hill farmers buying map information suitable for GIS management purposes are advised to invest in new maps for their farms based on aerial photographs and ground control points provided with RTK GPS. Point heights can be recorded with an RTK system from a vehicle on easy terrain. On steeper land point heights from ridge tops and gully bottoms are needed to enhance the accuracy of the DTM based on contours extracted from the aerial photograph. The resulting map would provide the farmer with accurate paddock areas, and with slope and aspect information for management decisions.
- The purchased aerial photograph provides a convenient and visual reference for farmers in discussing management decisions.
- The derived GIS map information could then be incorporated into an on farm package holding information for a range of farm operations including differential fertiliser application by ground machinery or aircraft.
Costall, D.A., D Williams, D.A.,Gillingham,A.G., Betteridge ,K.R. and Lambert , M.G. (2001). Effect of source and scale of mapping as a decision support tool for Precision Agriculture in hill country. In Eds. L.D. Currie and P Logonathan. Occasional Report No 14. Precision Tools for Improving Land Management. Proceedings of the Massey University Fertiliser and Lime Research Centre Workshop,14-15th February. In Press.
Figure 1. Locational difference between existing map information from Department of Lands and Survey compared with RTK survey data collected May 2001.
Figure 2. Contour displacement between the 20m contour map from the NZ 1:25000 map series compared to 20m contour data from the digitised 5m contour map of Ballantrae showed a displacement of approx. 50m.
Figure 3. Comparison of slope analysis based on RTK point heights, and 5 and 20 metre contours from the Ballantrae contour map source.