The Netherlands is characterised by mostly low lying, flat, and engineered agricultural lands, which are sensitive to flooding. To protect against floods, a correct network characterisation is of the utmost importance. A Dutch water resources management company, the HDSR, wishes to have a highly automated method to characterise the water course network. In this thesis, I investigate the possibilities of automatically identifying water courses in flat and engineered landscapes, using the raw points of the AHN3 LiDAR dataset. I found that there are many methods described in literature which identify channel-like features, and some which identify water courses in engineered landscapes, but none of these are suitable for this application. Thus, I designed a new methodology which is based on two concepts; (1) concave hulls, and (2) the MAT. The concave hull approach makes use of the presence of water in the water courses, while the MAT uses the concave profiles of the water courses to identify them. A workflow was implemented which uses the raw AHN3 LiDAR point cloud to identify for every water course the polygons of the water surface, and the geographical position of the water surface's centre lines. The implemented prototype was used for four different areas to test its applicability to different environments; a clay, peat, urban, and sand area. The water course characteristics in terms of water surface width and surface concavity, differ between these areas. The resulting datasets were validated to obtain mapping and positional accuracies. The experiments performed in this thesis show the potential of the designed methodology. The concave hull method is very robust to errors in the identification; there are relatively few errors of commission. However, the method does not perform well for high vegetation coverage or low water surface width. It is particularly suited for use in areas where relative water levels are high, water courses are wide, and vegetation coverage is low. The MAT is able to operate well when water levels are low, or even when water courses are dry, and it is relatively insensitive to vegetation coverage. However, it does not perform well for water courses which show little surface curvature, and is prone to errors of commission caused by local non-watercourse convexities and concavities. The combined prototype provides a strong and promising approach for the automatic identification of water courses in flat and engineered landscapes from the raw AHN3 point cloud. When the methods are combined, they manage to identify 98% of all water courses for the clay area, 97% for the peat area, 95% for the urban area, and 76% for the sand area. Clearly, the identification rates profit from the combination of methods. However, the relatively high error of commission of the MAT also radiates into the combined method. The error of commission is then 8% for the clay and peat area, 47% for the urban area, and 17% for the sand area. A number of possible improvements are identified which could elevate the identification rate for the sand area, but specifically lower the presented commission rates. Although the methods currently require a small amount of calibration when applied to new areas, they can in principle be fully automated.