SUMMARY In this paper we analyse a 3-yr long time-series of activity at Merapi volcano, Indonesia, which was recorded using a Doppler radar system. Between November 2001 and July 2004 about 57 000 events associated with dome instabilities have been recorded by the radar system. Because of the huge amount of data an automatic classification system has been developed, which identifies different types of instabilities at the lava dome (i.e. sliding material, gravitational break-offs and explosive outbursts) as well as rainfall. Comparing the record of detected instabilities to rockfall measurements deduced from the seismic network of the Volcanological Technology Research Center (VTRC), we are able to demonstrate the high potential of Doppler radar measurements for monitoring of dome activity. When aligned to the most active area at the dome, the radar detects significantly more events than the seismic system. In combination with seismic measurements, the Doppler radar allows for identifying changes in the location of the most active area at the dome. This is particularly helpful since the dome of Merapi is covered by clouds about 50 per cent of the time. Additionally, the radar data can help to distinguish between rockfalls and multiphase events in the seismic measurements. The different types of instabilities have been analysed for their occurrence frequency in terms of size, volume and mean velocity. Significant temporal changes in the relative occurrence frequencies or the characteristic properties of the different types of instabilities have not been observed. However, during the period of fastest activity decrease, explosive instability events with a gravitational precursor decreased faster than events starting explosively. Because the radar system is able to observe rainfall, the data has been analysed for a link between rain and dome activity. Throughout the observation period no such link could be found, however, this might be due to the fact that the activity has been almost continuously decreasing.