Sustainable use of hot spring resources depends on affective management actions, including resource investigation, exploitation control and utilization monitoring. Water quantity, water quality, and water temperature are three major factors concerned in the management of hot spring resources. Moreover, those three factors are determined by environmental variable in different scales. The geophysical investigation, such as controlled source audio-frequency magnetotelluric (CSAMT), is cost consuming, though it can provide a whole sketch of underground resources. Under the assumption that the private hot spring wells were economically constructed to gain the largest outflows from the shallowest wells, this study adopted six variables, temperature, conductivity, pH value, and oxidation reduction potential, flow rate per unit cross area of a well, and well depth, from 20 private hot spring wells in the Chihpen Hot Spring Area coupled with geostatistics to determine the optimal sites for monitoring wells. Multivariate factorial kriging (MFK) is different from conventional dimension reduction methods, such as principal component analysis (PCA) and correspondence analysis (CA), because it counts on spatial components and deals with scale problems. The MFK were applied to decompose the lumped variation into componential variations at regional and local scales. The principal components at different scales were inspected to clarify the causation of variations. For different management purposes, different principal components at different scales coupled with variation reduction analysis (VRA) were used to decide the optimal sites for hot spring monitoring wells to reduce uncertainty in management by objectives.