Reconstruction of mesoscale precipitation fields from sparse observations in complex terrain

The feasibility of a statistical reconstruction of mesoscale precipitation fields over complex topography from a sparse rain gauge network is examined . Reconstructions of gridded monthly precipitation for the European Alps (r esolution 25 km, 1202 grid points) are derived from rain gauge samples (70- 200-km interstation distance, 25-150 stations). The statistical model is ca librated over a 15-yr period, and the reconstructed fields are evaluated fo r the remaining 5 yr of the period 1971-90. The experiments are used to def ine the statistical setup, to assess the data requirements, and to describe the error statistics of a centennial reconstruction to be used in a forthc oming study. Reduced-space optimal interpolation is employed as the reconst ruction method, involving data reduction by empirical orthogonal functions (EOFs) and least squares optimal estimation of EOF coefficients. Also, a pr ocedure to define covariance-guided station samples with a "representative' ' spatial distribution for the reconstruction is proposed. Using a covariance-guided reference sample of 53 stations, the reconstructi on accounts for 77% of the total variance. For individual grid points the r elative reconstruction error (error variance divided by data variance) vari es between 10% and 40%; this value drops to 2%-10% when considering subdoma in means of 100 x 100 km(2). The mesoscale patterns of the fields and multi year precipitation anomalies are accurately reproduced. The EOF truncation is identified as the major limitation of the reconstruction skill but is ne cessary to avoid overfitting. Reconstructions from covariance-guided repres entative samples exhibit superior skill in comparison with those from rando mly distributed stations. The skill of the reconstruction was found to depe nd marginally on the choice of the calibration period within the 20 yr, eve n when months with exclusively positive or negative values of the North Atl antic oscillation index were selected for calibration. This result indicate s that the reconstruction model provides appreciable temporal stationarity.