In this work, a systematic approach based on system observability analysis and state estimation is developed to estimate the soil moisture inside an agro-hydrological system where measurements are not easily available. A discrete-time state-space model based on Richards equation is used to describe the agro-hydrological system that considers water dynamics inside soil-plant and atmosphere systems. The nonlinear agro-hydrological system is linearized every sampling time and the observability of the overall system is determined for the locally linearized model every sampling time. Based on the linearized models, we investigate how the number and location of output measurements affect the degree of observability of the system. To demonstrate the efficiency of the proposed approach, state estimation is performed using the extended Kalman filter on both simulated and real field data. The parameters of the model are estimated using prediction error method based on historical output measurements. (C) 2018 Elsevier Ltd. All rights reserved.