A new approach to estimate weathering indices (WIs) in soil was developed using proximally-sensed spectral reflectance over visible to shortwave-infrared (vis-NIR) and mid-infrared (MIR) region of electromagnetic spectrum. Partial-least-squares regression (PLSR) analysis was used to develop spectral algorithms for estimating twelve different WIs commonly used in geochemical literature. For each index, two different models were calibrated: 1) based on all the features in spectra and 2) based on the features obtained by variable importance projection, which we denote by vis-NIRvip, MIRvip, and (vis-NIR+MIR)vip. The residual prediction deviation (RPD) was used for checking the robustness of spectral models. Results showed that the MIR reflectance data provided superior estimation capability for all WIs compared with the vis-NIR reflectance data with the best possible prediction obtained for index of laterization (IOL; RPD=5.86) in the MIR and Mg Index (MgI; RPD=2.43) in the vis-NIR approach. The highest RPD values of 3.12, 4.13, 3.78, 6.13, and 5.08 were obtained for chemical index of alteration (CIA), MgI, mafic index of alteration (MIA(O)), IOL, and weathering index of Parker (WIP), respectively, when the PLSR model was based on (vis-NIR+MIR)vip. Best predictions were obtained when vis-NIR and MIR were combined and important spectral features were selected through variable importance projection (VIP) approach. Both vis-NIR and MIR technologies are available in the form of portable devices and are amenable for remote sensing mode of operation. This study demonstrates for the first time that the WI values in soil may be estimated in a rapid and non-destructive way in situ. © 2015 Elsevier B.V.