Gappy POD-based reconstruction of the temperature field in Tibet

Meteorological observations in Tibet are poor in quality with a severe amount of missing data; this is mostly caused by extreme climatological conditions and higher maintenance costs. This paper focuses on the imputation of missing data and the reconstruction of the regional temperature field. Due to insufficient observation stations and complicated topography, we employ the weather research and forecasting (WRF) model to produce the proper orthogonal decomposition (POD) basis for the study region. We then develop the gappy POD method for the imputation of missing data. Both methods are compared and tested for various missing data cases, and the results show that the gappy POD method dramatically outperforms the regularized EM algorithm when the amount of missing spatial data is not severe. Furthermore, between the two methods, only the gappy POD method is capable of reconstructing the temperature field at locations where the data are absent. The gappy POD method can also be generalized for data assimilation with the assumption that the data across all model grids have missing values.