Representativeness impacts on accuracy and precision of climate spatial interpolation in data-scarce regions

Data scarcity is a major scientific challenge for accuracy and precision of the spatial interpolation of climatic fields, especially in climate-stressed developing countries. Methodologies have been suggested for coping with data scarcity but data have rarely been checked for their representativeness of corresponding climatic fields. This study proved that satisfactory accuracy and precision can be ensured in spatial interpolation if data are satisfactorily representative of corresponding climatic fields despite their scarcity. The influence of number and representativeness of climate data on accuracy and precision of their spatial interpolation has been investigated and compared. Two precipitation and temperature indices were computed for a long time series in Bangladesh, which is a data-scarce region. The representativeness was quantified by dispersion in the data and the accuracy and precision of spatial interpolation were computed by four commonly used error statistics derived through cross-validation. The precipitation data showed very little and sometimes null representativeness whereas the temperature data showed very high representativeness of the corresponding fields. Consequently, precipitation data denoted scarcity but the temperature data denoted sufficiency regarding the required number of data for ensuring satisfactory accuracy and precision for spatial interpolation. It was also found that with the available data, accurate and precise precipitation surfaces can be produced only for representative synoptic spatial scales whereas such temperature surfaces can be generated for the regional scale of Bangladesh. It is highly recommended that the rain-gauge network of Bangladesh be increased or redistributed for computing representative regional precipitation surfaces.