Accurate forecast of power production and remaining fatigue life of a wind turbine require accurate estimation of actual wind loading. Lidar are used to monitor wind conditions during testing of new prototypes. However, their cost is prohibitive to deploying such measurement systems for any installed wind turbine. This work explores the idea of estimating wind loading locally exerted on a single wind turbine using soft-sensing. The 5 MW NREL open-design wind turbine is used as a case study. Applying the Augmented Extended Kalman Filter to estimate wind speed (input) and wind turbine response (output) based on few acceleration measurements taken from the blade tips based on a multibody dynamical model. The study is entirely numerical, that is, wind and sensing data is synthetic. The accuracy and the computational costs of the input-state estimates are examined.