Precipitation reconstruction for the Czech Lands, AD 1501–2010

Abstract

ABSTRACTThis contribution employs documentary‐based precipitation indices and long homogenized series of precipitation totals for quantitative reconstruction of seasonal and annual precipitation in the Czech Lands (now the Czech Republic) from AD 1501. Final calibration is based on linear regression using fully independent indices and data measured during the 1804–1854 overlap period, with subsequent variance scaling. Correlation analysis demonstrates that Czech documentary indices explain a significant amount of precipitation variability in all months and seasons of the entire overlapping period. Reconstruction results are best for annual values and for autumn (SON), for which proxy and target data share 36% of common variability. The coefficient of determination for summer (June‐July‐August – JJA) is 35%, for spring (March‐April‐May – MAM) 33%, while for winter (December‐January‐February – DJF) it is only 26%. Verification statistics [reduction of error (RE), coefficient of efficiency (CE)] computed for early (1804–1829) and late (1830–1854) overlapping periods indicate acceptable reconstruction skill for precipitation indices in JJA and annual values. However, for the other seasons they failed in the early or late calibration period, indicating possible chronological instability of reconstruction results in MAM, SON (September‐October‐November), and DJF seasons. The final reconstructions are complemented with uncertainty estimates. Reconstructed Czech precipitation series do not indicate long‐term trends but reveal quite high inter‐annual and inter‐decadal variability. Smoothed reconstructed DJF and JJA precipitation totals show the highest values in the second part of the 16th century, while the driest 30‐year period occurred during the 18th century in DJF, MAM, JJA, and in annual series. Direct comparisons with two other reconstructions (tree‐ring‐based for southern Moravia and gridded multi‐proxy for Central Europe) not only show significant correlations for a substantial part of the common period, but also disclose several periods with loss of coherence. Finally, uncertainties in reconstructions are discussed.