Estimation of missing values in climatological time series is an important
task. In order to find an appropriate method, we examined six methods for e
stimating missing climatological data (daily maximum temperature, minimum t
emperature, air temperature, water vapour pressure, wind speed and precipit
ation) for different time scales at six German weather stations and three B
avarian forest climate stations. The multiple regression analysis (using th
e five closest weather stations) with least absolute deviations criteria (R
EG) predominantly gave the best estimation for daily, weekly, biweekly, and
monthly maximum temperature, minimum temperature, mean temperature, water
vapour pressure, wind speed, under different topographical conditions (vall
ey, alpine foothills and mountain sites). The six methods gave similar esti
mates for the averaged precipitation amount. The mean absolute errors (MAE)
of estimating climatological data using different techniques are of simila
r magnitude at the weather stations, but they are significantly different a
t the forest climate stations. For the same climatological variable (i.e.,
air temperature) for different time scales, mean absolute errors of estimat
ed data are larger for shorter time scales (e.g., a day) than for longer on
es (e.g., a month). For the different climatological variables, the most ac
curately estimated variables are maximum temperatures, mean temperatures an
d water vapour pressure, followed by minimum temperature and wind speed. Th
e poorest results were obtained for precipitation data. (C) 1999 Elsevier S
cience B.V. All rights reserved.