A coupled land-atmosphere climate model is examined for evidence of climate
drift in the land surface state variable of soil moisture. The drift is ch
aracterized as pathological error growth in two different ways. First is th
e systematic error that is evident over seasonal timescales, dominated by t
he error modes with the largest saturated amplitude: systematic drift. Seco
nd is the fast-growing modes that are present in the first few days after e
ither initialization or a data assimilation increment: incremental drift. W
hen the drifts are robust across many ensemble members and from year to yea
r, they suggest a source of drift internal to the coupled system. This sour
ce may be due to problems in either component model or in the coupling betw
een them. Evidence is presented for both systematic and incremental drift.
The relationship between the two types of drift at any given point is shown
to be an indication of the type and strength of feedbacks within the coupl
ed system. Methods for elucidating potential sources of the drift are propo
sed.