The impact of land variables on temperature forecasts in atmospheric cycling is often underestimated or overlooked. This oversight primarily occurs due to the abundance of meteorological measurements available for assimilation and partly because soil states are assumed to be quickly reset by atmospheric forcing, such as precipitation, justifying no spin-ups or no updates of soil states during cycling. In this study, by updating soil moisture every 6 hours using different analysis datasets for May 2019, considerable discrepancies were found, highlighting large uncertainties in soil moisture analysis. Different soil moisture analyses produced systematically different temperature forecasts, with errors growing over cycles to be comparable to a typical error magnitude of 2-m temperature observations (~2ºK). This study demonstrates that temperature forecasts are significantly influenced by whether and how soil moisture is updated, not only near the surface but also up to the low-mid troposphere and throughout the cycles.