Reliable estimation of rainfall distribution in mountainous regions poses a great challenge not only due to complex relationships between irregular topography and precipitation, but also due to non-availability of abundant rainfall measurement points. This study presents a comparison of univariate and multivariate methods for spatial interpolation of rainfall in complex terrain of Indian Himalayas in Uttarakhand region. To understand the role of topography in explaining the spatial variability of rainfall, elevation and slope calculated from DEM are incorporated as covariables in multivariate interpolation methods. Further, the study area is divided into three different regions based on variation in the elevation of topography and the performance of different spatial interpolation methods is analyzed for subdivided regions. The normal annual rainfall data of 80 rain gauge stations spread over 53, 484 km2 is used to generate rainfall map using different interpolation methods. Cross validation is used to compare the prediction performances of the three multivariate geostatistical interpolation algorithms: simple kriging with varying local means (SKlm), ordinary cokriging (OCK), regression kriging (RK) and two univariate techniques: inverse distance weighted (IDW) and ordinary kriging (OK). The inclusion of elevation or slope as auxiliary variables improves the prediction of rainfall in mountainous region. SKlm performs better than OCK in all the cases. Residual kriging outperforms SKlm in Shivalik and Lesser Himalayas region where the correlation between rainfall and elevation is high (rshivalik=0. 79, rlesserhimalayas=0. 5). Also, OCK proves to be a better choice than OK only when the correlation between the rainfall and secondary variable (elevation or slope in this case) is good. When the data of Shivalik region (r=0. 79) is considered, the RMSE of OCK is 363 and that of OK is 368. The prediction using OK and OCK yields similar result when the correlation between rainfall and elevation is moderate (r < 0. 4). For all the cases, residual kriging outperforms OCK. © 2014, Capital Publishing Company.