Original Articles

Using the Nearest Neighbour method to substitute missing daily solar radiation data


Abstract

Ground level solar radiation reflects the amount of energy that reaches the earth's surface and is utilised by people, animals and plants. Biological models often require such radiation records for long periods of time, however, a lack of radiation data is common to many countries. Consequently various methods were developed to estimate daily radiation from other meteorological measurements. These methods normally need site specific calibration, require a fixed amount of input variables and do not include uncertainties introduced by global climate change. In this paper an attempt was made to develop a station independent substitution method without sacrificing accuracy. Meteorological data from different climatic regions in South Africa were used to assess the suitability of the Nearest Neighbour (NN) method. This method is based on the re-occurrence of events similar to those in the past. Different statistical approaches were assessed to calibrate the distance equation. An attempt was made to find suitable weights, calibrated universally, for the distance equation that would still produce good estimates for individual stations. The universally calibrated NN method outperformed previously developed equations (locally calibrated) by as much as 20% and overcomes various shortcomings identified in these equations. More detailed analyses also confirmed that the NN approach generates more representative statistical distributions and estimates extreme instances of solar radiation more accurately.

Get new issue alerts for South African Journal of Plant and Soil