Coordinated waterbird counts have been carried out in major Kenyan wetlands annually over the last decade. The lakes of Naivasha, Elmenteita, Nakuru (counted since 1991) and Bogoria are close to each other in the southern Rift Valley and hold the bulk of both resident and migrant waterbirds counted each year. Levels of the four lakes fluctuate substantially and all but Naivasha are saline. By contrast, Dandora Ponds near Nairobi is a site with stable water levels. We compared trends in numbers of birds, categorized by family and by migratory status, with an earlier analysis in 1996 and between the fluctuating Rift Valley lakes and Dandora Ponds. Flamingo numbers remained highly variable at all sites, showing no long-term trends. In 1996, significant declines were found in many piscivorous waterbird groups, including grebes, pelicans, cormorants, storks, terns and gulls, mainly associated with falling lake levels at Lake Nakuru; by 2000, numbers had recovered. Consistent declines in rallids, birds of prey and kingfishers were evident in 1996, mainly due to falling numbers at Lake Naivasha. By 2000, kingfisher numbers had fully, and raptor numbers had partially, recovered. Rallids, mainly Red-knobbed Coot (Fulica cristata), continued to decline sharply through 1998, but recovered slightly to 1996 levels by 2000. After notably high numbers in 1996-1999, Palaearctic ducks declined dramatically in 2000 to a ten-year low. With the exception of rallids, none of the groups that had declined from 1991-1996 showed any trend in numbers over the extended time period. At Dandora, numbers of most waterbird groups fluctuated substantially despite constant conditions. The fluctuations showed little synchrony with between species or with those at the southern Rift Valley lakes, suggesting waterbirds were responding to the availability of suitable wetlands over a large area. The declines from 1991-1996 in the southern Rift Valley were temporary; recoveries in waterbird numbers were not attributable to improved environmental management (human pressures on these wetlands intensified from 1997-2000), but to higher lake levels in 1997/98. The ten-year data set covers a full cycle of low and high lake levels, which the six-year dataset did not, and may now form an adequate baseline for monitoring.