Evaluation of interpolation techniques for the creation of gridded daily precipitation (1 × 1 km²); Cyprus, 1980–2010

Abstract

High-resolution gridded daily data sets are essential for natural resource management and the analyses of climate changes and their effects. This study aims to evaluate the performance of 15 simple or complex interpolation techniques in reproducing daily precipitation at a resolution of 1 km 2 over topographically complex areas. Methods are tested considering two different sets of observation densities and different rainfall amounts. We used rainfall data that were recorded at 74 and 145 observational stations, respectively, spread over the 5760 km 2 of the Republic of Cyprus, in the Eastern Mediterranean. Regression analyses utilizing geographical copredictors and neighboring interpolation techniques were evaluated both in isolation and combined. Linear multiple regression (LMR) and geographically weighted regression methods (GWR) were tested. These included a step-wise selection of covariables, as well as inverse distance weighting (IDW), kriging, and 3D-thin plate splines (TPS). The relative rank of the different techniques changes with different station density and rainfall amounts. Our results indicate that TPS performs well for low station density and large-scale events and also when coupled with regression models. It performs poorly for high station density. The opposite is observed when using IDW. Simple IDW performs best for local events, while a combination of step-wise GWR and IDW proves to be the best method for large-scale events and high station density. This study indicates that the use of step-wise regression with a variable set of geographic parameters can improve the interpolation of large-scale events because it facilitates the representation of local climate dynamics.

Evaluation of a spatial rainfall generator for generating high resolution precipitation projections over orographically complex terrain

“…The method developed and tested by Camera et al (2014) to derive 1 9 1 km 2 daily gridded data sets of precipitation for Cyprus (1980Cyprus ( -2010 was applied. The method involves inverse distance weighting (IDW) to interpolate local events, while a combination of step-wise geographically weighted regression and IDW is used for large scale events.…”

“…Hence, a simplified gridding scheme is developed. Thiessen polygons are established around each observational station and raster cells, from the 1 9 1 km 2 mask map of the gridded data sets (Camera et al 2014), are assigned to the different polygons. A scaling coefficient is calculated for each cell as the ratio between the mean annual rainfall (from observations 1980-2010) of the cell itself and the same value of the reference station, for the Thiessen polygon in which they fall.…”

A conceptually superior variant of Shepard's method with modified neighbourhood selection for precipitation interpolation

“…Several spatial interpolation techniques such as inverse distance weighting (IDW) method, thin plate splines (TPS), trend surface analysis (including linear and polynomial regression), kriging and parameter elevation regression on independent slopes model (PRISM; Hutchinson, ; American Society of Civil Engineers (ASCE), ; Goovaerts, ; Hay et al , ; Daly et al , ) were used either independently or in a combination (Lin and Chen, ; Perry and Hollis, ; Camera et al , ) to generate gridded precipitation data in the last two decades. Among those, the deterministic IDW and the kriging methods with their respective variants are widely used both for the estimation of missing rainfall records and interpolation (Teegavarapu, ; Li and Heap, ; Ly et al , ; Yanto et al , ).…”

Impact of Rain Gauge Quality Control and Interpolation on Streamflow Simulation: An Application to the Warwick Catchment, Australia

“…Therefore, there has been little consensus on the relative superiority of these methods. For instance, Camera et al (2014) showed that different interpolation methods work better for local and large scale events (frontal or convective, etc. ).…”