Three land-surface models with land-data assimilation scheme (DA) were evaluated for one growing season using in situ observations obtained across Europe. To avoid drifts in the land-surface state in the models, soil moisture corrections are derived from errors in screen-level atmospheric quantities. With the in situ data it is assessed whether these land-surface schemes produce adequate results regarding the annual range of the soil water content, the monthly mean soil moisture content in the root zone and evaporative fraction (the ratio of evapotranspiration to energy available at the surface). DA considerably reduced bias in net precipitation, while slightly reducing RMSE as well. Evaporative fraction was improved in dry conditions but was hardly affected in moist conditions. The amplitude of soil moisture variations tended to be underestimated. The impact of improved land-surface properties like Leaf Area Index, water holding capacity and rooting depth may be as large as corrections of the DA systems. Because soil moisture memorizes errors in the hydrological cycle of the models, DA will remain necessary in forecast mode. Model improvements should be balanced against improvements of DA per se. Model bias appearing from persistent analysis increments arising from DA systems should be addressed by model improvements.