Crop yield is strongly correlated to soil moisture and its variation with respect to phenological stage of the plant. For this reason spatial and temporal distribution of soil moisture is substantial for agricultural planning, irrigation and water resource management. A soil profile estimation model where agro-meteorological and remote sensing data sets are fused by wavelet neural network is developed for large scale multi sensor agricultural monitoring network in Turkey (TARBIL). Temporal rain, parcel based evapotranspiration, amount of irrigation together with fractional vegetation cover have been used as inputs of multiple-input time-delay wavelet neural network for spatiotemporal multi-depth soil moisture profile estimation. An error source in soil moisture estimation is discontinuity in soil property change. Meanwhile additional estimation on z-axis for profile data is also effective in complexity of the problem. In this study, we have investigated effect of some physical and chemical property changes and proposed possible improvements on the soil profile estimation model.