REVEALING CELTIC FIELDS FROM LIDAR DATA USING KRIGING BASED FILTERING Astrid Humme 1 , Roderik Lindenbergh 1 and Chris Sueur 2 1: Delft University of Technology, Delft Institute of Earth Observation and Space Systems; A.J.M.Humme@student.TUDelft.NL, R.C.Lindenbergh@lr.tudelft.nl 2: Vestigia B.V.; c.sueur@vestigia.nl KEY WORDS: LIDAR, archaeology, filtering, geostatistics ABSTRACT: A detailed elevation model of the Netherlands has been acquired with the use of Airborne Laser Altimetry and is managed by AHN (Actueel Hoogtebestand Nederland). This elevation model can be used for a large amount of applications. Using a suited illumination technique it turned out to be possible to find and visualize so-called Celtic fields, agricultural fields systems of about 2500 years old in the AHN data. Here we will show that the visualization of Celtic fields out of AHN data can be enhanced by applying a geostatistical filter technique that removes the raw topography from the data set such that only the micro-topography, including the Celtic fields, remains. Visualization of this micro-topography not only enhances the images of the already known Celtic fields but also shows that the spatial extend of the field system is larger than expected. 1. INTRODUCTION Celtic fields are prehistoric agricultural field systems that are rec- ognizable as rectangular patches of land of about 40 × 40 me- ter surrounded by low earth walls, see Fig. 1. The highlight of the use of Celtic fields was between 800 B.C. and 0 B.C. and they can be found in several countries in North-West Europe, (Spek, Groenman-van Waateringe, Kooistra, and Bakker, 2003). In open land Celtic field systems are often still recognizable, even from the ground, although it is easier to distinguish them by color changes in areal photo’s. Unfortunately it is very difficult to find Celtic field systems in forest areas, as small elevation differences or small color changes are completely masked by the forest cover. Figure 1. Impression of a Celtic field system. (Rob Beentjes) It turns out however that, even in forest areas, Celtic field systems can be found by analyzing airborne laser altimetry data, (Zij- verden and Laan, 2005). Several approaches exist to obtain the bare earth surface out of raw airborne laser scanning point clouds, (Sithole and Vosselman, 2004). After removing points that were reflected by trees or shrubs, the remaining points are assumed to represent the ground surface. In this way it is possible to look for archeological features under the trees, (Da Devereux, Amabable, Crow, and Cliff, 2005). In The Netherlands, the Dutch Ministry of Public Works initi- ated the setup of the so-called Actueel Hoogtebestand Nederland (AHN), that can be translated as ‘Up to date height data base of The Netherlands’. This database consists of interpolated airborne laser altimetry data covering the whole of The Netherlands, con- taining at least one point per 4 m 2 outside and one point per 16 m 2 inside forest areas, (Anonymous, 2000). The accuracy of the laser altimetry technique depends strongly on the amount of vegetation and topography in the area. AHN states that for the accuracy of solid topography (such as roads and parking lots) as well as flat or soft topography (such as beaches and grass-fields) a standard deviation of 15 cm maximum applies, with a system- atical error of 5 cm maximum. Figure 2. Celtic fields, inside the circle, revealed by illumination. This elevation model can be used for a large amount of applica- tions, including many archeological, see also (De Boer, Laan, Waldus, and Van Zijverden, ToAp). The third author of this pa- per applied a technique based on illumination to visualize ancient Celtic fields, near Doorwerth, situated in the East of the Nether- lands, out of AHN data. Figure 2 shows the fields as exposed by the third author. By illuminating the ground surface points in a suited way in a software visualization program, some unknown field systems could be traced by the shadows of the micro-relief of the low earth walls. In our contribution we show that this approach can be still im- proved. In the original approach, the micro-relief is visualized with respect to the natural relief of the ground surface as it is rep- resented in the laser data. By filtering the large scale topography the micro-relief becomes more prominent: only small-scale fea- tures like road beds, foot-paths and the earth walls surrounding