An inexpensive video data capture system

Digital mapping techniques are used increasingly in hydrology. One of the constraints on the application of those techniques is the digitizing process, which can be costly and time consuming. This paper describes the background and development of a low cost, easy-to-use system for digitizing maps, based on the use of a video camera as a scanner, and goes on to discuss the application of this system to the task of digitizing the hydrological soil maps of Europe. Une système peu coûteux d'enregistrement des données hydrologiques par caméra vidéo Résumé Les techniques de digitilisation de cartes sont de plus en plus utilisées en hydrologie. L'une des contraintes de l'application des ces techniques est la digitilisation, qui peut être longue et coûteuse. Cet article décrit le principe et la mise au point d'un système de digitilisation de cartes, facile et peu coûteux, basé sur l'utilisation d'une caméra vidéo comme scanner. L'application de ce système à la digitilisation des cartes de sols hydrologiques d'Europe est ensuite analysée. INTRODUCTION The association of river flow behaviour with the physical and climatic characteristics of catchments is fundamental to estimating floods and low flows for ungauged rivers. The application of this technique to the analysis of river flows across Europe has been an important aspect of the FREND (Flow Regimes from Experimental and Network Data) project. This international study is a major contribution to the third phase of the UNESCO International Hydrological Programme. The objective of the FREND project is to relate flow statistics to catchment characteristics, such as soil type, making use of the large amount of river flow data available from organisations within the study area, in order to estimate flows for ungauged rivers and changes in the flow regime likely to result from human influence on the catchment or from climatic change. An extensive data set is being built up, consisting of flow data for over 2000 gauged European rivers. In addition, data on the various catchment characteristics are also required (Beran, 1982; Institute of Hydrology, 1984), including rainfall, morphology, land use and geology, as well as soil type, for Open for discussion until 1 October 1989 157 Roswitha Gross 158 which the source data are generally available in map form. Soil type is one of the physical catchment characteristics expected to influence river flow behaviour significantly, and a series of maps has been derived from the EC Soil Map of Europe in conjunction with national hydrological and agricultural services within the FREND project study area (Gustard, 1983), according to the WRAP (Winter Rainfall Acceptance Potential) classification scheme (Farquharson et al, 1978). These maps display the spatial variability of the characteristics of the soils of Europe at a scale of 1:1 000 000 in terms of five hydrological soil classes, Class 1 referring to soil with a very high ability to accept rainfall, Class 5 indicating very low acceptance potential. Part of the map covering Brittany, France, is shown in Fig. 1. Fig. 1 Hydrological soil map: western Brittany. The catchment values of various characteristics, such as the fraction of area covered by a specific soil type, are required for the entire study area. The large number of catchments to be analyzed in the project made the use of automation a necessity. The thematic data available on maps must 159 An inexpensive video data capture system for hydrological maps therefore be converted to computer-compatible form. This step of data capture can be time-consuming and expensive, and remains a constraint on the use of Geographical Information Systems (GIS) and other analytical software. This paper outlines the methods of map digitization currently used and the demand for a low cost, easy-to-use method. The development of the video data capture system, and its application to the task of digitizing the hydrological soil maps of Europe, is then described, followed by a discussion of the advantages and limitations of the system and of future work. DIGITAL CARTOGRAPHY Catchment characteristics have traditionally been derived manually from maps, by overlaying the catchment boundary on a map of the variable of interest, then using area measurement (e.g. square-counting or planimetry) to calculate the catchment average value of that variable. Such manual methods are time-consuming, prone to error and not easily repeatable and therefore difficult to check. Digital techniques allow the derivation of catchment characteristics to be rapid, objective and reproducible. A catchment boundary, digitized and held in vector form (i.e. as a series of x, y coordinate pairs) can be converted to a raster or grid by considering those squares that lie inside the boundary to have a value of 1, others of 0 (Fig. 2a). If thematic data, for example rainfall, are also available in gridded form on a suitable scale (Fig. 2b), the two grids can be combined by matrix multiplication and the catchment average value of that variable calculated automatically (Fig. 2c) (Wiltshire, 1983). As well as making feasible the analysis of large numbers of catchments, digital maps can be automatically overlaid and compared to investigate the relationships between two or more mapped variables. A prerequisite of these automated techniques is the availability on a computer grid of the thematic data. Digitizing techniques available vary according to the accuracy and resolution required, and according to the data source and type. Two types of map data can be distinguished. The first, continuously varying data such as altitude or annual rainfall, are usually depicted by isopleths (lines of equal magnitude) which are conveniently digitized as vectors with the aid of a digitizing table and cursor. Subsequent rastering of the vector data can be quite easily achieved using an interpolation routine. In contrast, 'plateau' type variables have sharp discontinuities at the junction of areas of constant value. Examples are soil type, geology or forest, and these are most commonly obtained directly as a raster or grid of values. For example, a digital forest data set has been derived by overlaying manually an appropriate grid on a map and counting those squares more than 50% filled by forest, a relatively subjective and time-consuming task. However, it is possible also to obtain a computerized version of such a map using conventional table and cursor means. In a recent exercise, the soil boundaries were digitized as vectors, each feature coded according to the soil types on either side, with special care taken to ensure that each limb closed precisely