GIS Ostrava 2012 - Surface models for geosciences January 23. – 25., 2012, Ostrava OPTIMAL INTERPOLATION OF AIRBORNE LASER SCANNING DATA FOR FINE-SCALE DEM VALIDATION PURPOSES Michal GALLAY 1 , Chris LLOYD 2 , Jennifer MCKINLEY 2 , 1 Ústav geografie, Prírodovedecká fakulta, Univerzita Pavla Jozefa Ģafárika, Jesenná 5, 040 01 Koģice, michal.gallay@upjs.sk 2 School of Geography, Archaeology and Palaeoecology, Queen's University Belfast, BT7 1NN Belfast, c.lloyd@qub.ac.uk, j.mckinley@qub.ac.uk Abstract Airborne laser scanning (ALS) is becoming a widely used method for detailed remote sensing across European and other countries. It provides dense sets of highly accurate data points which can be further used for creation of digital elevation models (DEMs). However, the high density also determines higher processing power demands. In case grid-based DEMs are concerned and interpolation is required, it is therefore useful to test the performance of particular interpolation methods using smaller samples in order to find the suitable approach and provide guidelines for other users. In this paper, inverse distance weighting (IDW) and regularized spline with tension and smoothing (RST) were tested to find the optimal DEM based on last return airborne laser scanning (ALS) data. Data from four sites differing in terrain complexity were used. Summary statistics of cross-validation errors were used to select the parameters for generating the optimal ALS DEM. The results show that IDW and RST generate very similar cross-validation statistics. Cross-validation errors increase with increasing short-range variation of elevation for IDW as well as for the RST interpolation method. Rescaled tension between 200 – 400 with smoothing provided acceptable cross- validation statistics for RST. Supplemental validation of the interpolated ALS DEMs against ground surveyed measurements provided means for absolute accuracy assessment. It suggests that both IDW and RST are equivalent for the outlined purpose. For IDW being a simpler method demanding less processing time, this was chosen for deriving the optimal ALS based DEM. The choice of interpolation method is less influential when a surface is interpolated to coarser or similar resolutions than the resolution of the input data which supports finding by other authors. Keywords: inverse distance weighting, spline, laser scanning, lidar, interpolation INTRODUCTION Airborne laser scanning (ALS) can be considered as the most accurate method for mapping land surface by remote sensing. It provides rapid and dense collection of data points with submeter to subdecimeter measurement precision. Such data properties are difficult to achieve in an efficient way by other remote sensing methods such as photogrammetry, synthetic aperture radar (SAR). ALS is capable of collecting altitude of several surface levels depending on the penetration of laser beam down through the ground. The height data are further used to generate digital elevation models (DEM) and the recorded intensity of the backscattered laser beam can be used for classification of surface objects. Generally in vegetated areas, the first returns correspond to the upper landscape canopy level (e.g. vegetation tops) and the last returns to the ground (terrain surface). While the first returns are used to generate digital surface models (DSM) the last returns are used for generation of digital terrain models (DTM). In cases, where impenetrable objects such as buildings are present or the ground is exposed the first returns refer to the last returns. Detailed background on ALS can be found in Baltsavias (1999) or Pfeifer and Briese (2007). Nowadays, ALS as a technology is becoming more accessible for a wide range of users who deal with the problem of effective processing of millions of data points. This has also stimulated research on suitable approaches for generation of digital elevation model (DEM) from the ALS data (Rees 2000, Lloyd and Atkinson 2002, Mitášová et al. 2005). The studies compared performance of several interpolation methods used to derive gridded DEMs for a certain region. Comparison of various kinds of interpolation methods for DEM generation for various terrain types is presented in Carrara et al. (1997) or Svobodová and Tuček (2009) who, however,