Spatial Accuracy 2016, Montpellier, France, 5 - 08 July 2016, vol.1, no.1, pp.1-6
Predictions of the variables at points with no measurements are obtained by interpolation techniques. Space-time interpolation techniques that consider variation both in space and time provided a new research area. Temperature is an important climatic parameter varying both in space and time. Like other meteorological, hydrologic and environmental variables, temperature is measured at specific locations. In order to obtain predictions for all grid locations, kriging methods have been applied for a long time. In here, space-time Ordinary kriging (ST-OK) and space-time Universal kriging (ST-UK) have been used in annual temperature estimation over Lake Van Basin using at 13 meteorological station observations for 2002-2012. Elevation, land cover, distance to Van Lake are used as secondary information in ST-UK. Elevation at 500 m resolution are obtained by Nearest Neighbour resampling of 3 arc second Shuttle Radar Topography Mission (SRTM). MOD12Q1 land cover data set has been downloaded from USGS organization. This dataset has 500 m spatial resolution and 17 subclasses. Distance to nearest coast variable is obtained by calculating the Euclidean distances of each SRTM pixel to the nearest boundary of the Van Lake coast vector. Annual temperature values are analysed and predicted at 500m*500 m resolution for 11 year-period. One-fold cross-validation is used to assess accuracy performance of both methods. R-square and Root Mean Square Error (RMSE) are calculated and evaluated for each technique. Comparison of kriging methods and inclusion of secondary information is assessed.