This dissertation describes efforts to overcome the challenges in designing in situ soil moisture observation network. The surface soil moisture data collected at two spatial scales in a working field in Iowa throughout the growing seasons in 2004 to 2008 were used to describe the spatio-temporal characteristics of soil moisture at field scale. The rank stability analysis was used to identify the locations on the ground to represent the mean soil moisture at the field scale across five different growing seasons. Optimal sampling locations (OSLs), giving accurate estimates of field mean soil moisture for each season, were selected using the rank stability analysis. The results indicated that there were OSLs in the field for each growing season and for the compiled five-season, and these locations were different from each sampling season to the next, which suggested that it is not sufficient to use only one year or a few years' data to identify the soil moisture rank stable behavior using rank stability analysis. The spatial patterns of soil moisture exhibited certain consistency across multiple seasons. The OSLs all tended to be located at those locations with higher elevation. Therefore, multiple linear regression was used to predict recurring soil moisture patterns with topographic indices at optimal resolutions. A genetic algorithm was developed to select the input independent variables over a range of resolutions for multiple linear regression models. Using this approach, not only were the primary influential topographic indices to soil moisture patterns uncovered, but the most appropriate resolutions for each influential index was identified. The recurring patterns at field scale were well predicted by the combination of static topographic indices at optimal resolutions. Although the studies included in the dissertation contributed knowledge to in situ soil moisture network design, more work is required to obtain a complete scheme for implementing a ground-based observation network effectively.