Abstract:
Rainfall is considered a highly valuable climatologic resource, particularly in arid regions. As one of the primary inputs that drive watershed dynamics, rainfall has been shown to be crucial for accurate distributed hydrologic modeling. Precipitation is known only at certain locations; interpolation procedures are needed to predict this variable in other regions. In this study, the ordinary cokriging (OCK) and collocated cokriging (CCK) methods of interpolation were applied for rainfall depths as the primary variate associated with elevation and surface elevation values as the secondary variate. The different techniques were applied to monthly and annual precipitation data
measured at 37 meteorological stations in the Central Kavir basin. These sequential steps were repeated for the mean monthly rainfall of all twelve months and annual data to generate rainfall prediction maps over the study region. After carrying out cross-validation, the smallest prediction errors were obtained for the two multivariate geostatistical algorithms. The cross-validation error statistics of OCK and CCK presented in terms of root mean square error (RMSE) and average error (AE) were within the acceptable limits for most months. Then the two approaches were compared to select of the most accurate method (AE close to zero and RMSE from 0.53 to 1.46 for 37 rain gauge
locations for all months). The exploratory data analysis, variogram model fitting, and generation precipitation prediction map were accomplished through use of ArcGIS software.
Machine summary:
(1999) adopted kriging with external drift using both satellite data and ground rain gauges to improve the estimation of decadal rainfall and their spatial distribution while Goovaerts (2000) adopted a digital elevation model for monthly and annual rainfall totals in south Portugal.
PardoIgu´zquiza (1998) compared the areal average climatological rainfall mean estimated by the classical Thiessen method, ordinary kriging, cokriging, and kri ging with an external drift (the first two methods used only rainfall information, while the latter two used both precipitation data and orographic information) in the G uadalhorce river basin in so uthern Spain, and concluded that kriging with an external drift seemed to give the most coherent results in accordance with cross-validation statistics and had the advantage of requiring a less demanding variogram analysis than cokriging.
Zhang and Srinivasan (2009) developed nearest-neighbor (NN), inverse distance weighted (IDW), simple kriging (SK), ordinary kriging (OK), simple kriging with local means (SKlm), and k riging with external drift (KED) to facilitate the estimation of automatic spatial precipitation while incorporating the geographic information system program in the Luohe watershed, located downstream of the Yellow River basin.
After calculating OCK and CCK, the cross- validation error statistics were compared to select the best method to perform the CK, Then, the best CK algorithm (either OCK or CCK) was used to predict the primary variate values for more locations within the study region.
The predicted mean monthly rainfall values from the CK algorithm and the observed data of rain gauge locations for all twelve months were subjected to ordinary kriging analysis.