The observations of the traditional parameters as wind, temperature, and precipitation amount are performed fully automatic by KNMI for more than a decade. Observers enter the so-called visual observations of visibility, clouds, and present weather manually at designated stations. KNMI works on the automation of these visual observations. Presently KNMI operates three fully automated stations which make hourly reports including the visual observations. One automated station is a new visual station and forms an extension of the operational network. The other 2 stations are existing manned stations that make in addition to the operational report a fully automated report for internal use only. The results of these stations are used for testing and optimising the automated system. Furthermore, sensor measurements for other stations are processed off-line and compared with routine visual observations.

KNMI uses the Vaisala FD12P forward scatterometer, a so-called present weather sensors, to measure visibility, precipitation amount and type. The measurements are used in combination with measurements from other sensors such as rain gauge, anemometer, ceilometer and a lightning detection network, to generate nearly all WMO synoptic weather codes for automated stations. The manned and automated weather codes differ and cannot easily be compared. Therefore, the comparison is performed on precipitation type, which is readily available from the present weather sensor and is derived from routine hourly SYNOP reports made by observers. The results of this comparison will be presented. A comparison of visibility and lightning reports will also be considered in this presentation. A detailed comparison of cloud reports is the topic of an accompanying paper.

The table shows an example of the comparison for visibility at De Bilt in 2000. The coloured parts in the table indicate the observations where sensor and observer agree that fog is present or not. The contingency table shows that the sensor fog reports have a POD=68% and a FAR=41%. The bias is 1.17 the sensor reporting more hours with fog than the observer. Differences are mainly the result of differences in location (sensor at 1,5m versus observer at 15m) and reporting rules (sensor reports 10-minute average versus the lowest observed value around).