Streamflow and sediment load exporting from catchments usually show high inter-variability caused by climate variability and land surface changes. It is challenge to predict monthly or daily streamflow and sediment load using simply statistical methods. The state-space model is an effective tool for quantifying localized variations, and it only needs some easily obtained climate and land surface variables. However, there is no state-space model developed to predict the long-term variations of monthly streamflow, sediment load and suspended sediment concentration. In this study, fourteen main catchments in the Loess Plateau of China were chosen, and these catchments experienced abrupt reductions of annual streamflow (Q), specific sediment yield (SSY) and suspended sediment concentration (SSC) during the past five decades. The state-space model was developed to predict monthly Q, SSY and SSC during 1982–2011 by precipitation, potential evapotranspiration and normalized difference vegetation index. The results indicated that the monthly Q, SSY and SSC in summer season decreased significantly in nearly all the catchments (p < 0.01). The periodicity of monthly Q, SSY and SSC reduced, and especially the periodicity was not detected after 2003. The precipitation and previous Q were the two main state variables to estimate monthly Q with transition coefficients of 0.56 and 0.25, respectively, whereas Q and previous SSC were the most important state variables to estimate monthly SSY and SSC with transition coefficients of 0.50 and 0.34, respectively. The state-space model was satisfactory to simulate monthly Q, SSY and SSC with Nash-Sutcliffe model efficiency (NSE) over 0.85 in model calibration and validation. Furthermore, the model performance was best for predicting monthly Q, followed by predicting monthly SSY and SSC with the average NSE of 0.90, 0.89 and 0.87, respectively. The reductions of monthly streamflow and sediment load were mainly derived by soil and water conservation measures (SWCMs) and especially check-dams. The catchment runoff and sediment coefficient decreased significantly and linearly with the percentages of area affected by SWCMs (p < 0.01). This study indicates that the state-space model is a useful tool to predict streamflow and sediment load time series with the hydrological and land surface informations.