Abstract. The bias in atmospheric variables and that in model computation are
two major causes of failures in discharge estimation. Attributing the bias in
discharge estimation becomes difficult if the forcing bias cannot be
evaluated and excluded in advance in places lacking qualified meteorological
observations, especially in cold and mountainous areas (e.g., the upper
Tarim Basin). In this study, we proposed an Organizing Carbon and Hydrology In Dynamic
EcosystEms (ORCHIDEE)-Budyko framework which
helps identify the bias range from the two sources (i.e., forcing and model
structure) with a set of analytical approaches. The latest version of the land
surface model ORCHIDEE was used to provide reliable discharge simulations
based on the most improved forcing inputs. The Budyko approach was then
introduced to attribute the discharge bias to two sources with prescribed
assumptions. Results show that, as the forcing biases, the water inputs
(rainfall, snowfall or glacier melt) are very likely underestimated for the
Tarim headwater catchments (−43.2 % to 21.0 %). Meanwhile, the
potential evapotranspiration is unrealistically high over the upper Yarkand
and the upper Hotan River (1240.4 and 1153.7 mm yr−1, respectively). Determined by the model structure, the bias in actual
evapotranspiration is possible but not the only contributor to the discharge
underestimation (overestimated by up to 105.8 % for the upper Aksu River). Based
on a simple scaling approach, we estimated the water consumption by human
intervention ranging from 213.50×108 to
300.58×108 m3 yr−1 at the Alar gauge station, which is
another bias source in the current version of ORCHIDEE. This study succeeded
in retrospecting the bias from the discharge estimation to multiple bias
sources of the atmospheric variables and the model structure. The framework
provides a unique method for evaluating the regional water cycle and its
biases with our current knowledge of observational uncertainties.