Currently the application of fractional differential (FD) algorithm for field spectra and the optimal band mechanism of spectra inversion salt in saline soil has not been seen. In view of the problem that traditional integer-order differential algorithm causes fractional-order spectral information lose and model accuracy decrease the salinity and field spectra of saline soils in Fukang City of Xinjiang, China, were taken as data sources and original spectrum and common four transforms were subjected to a total of 21-order FDbetween 0～2 orders (interval is 0.1), in order to explore the mechanism improvement of pretreatment precision for saline soil spectra. Results showed that: (1) FD could accurately display the details of spectral transformation in the derivation process due to continuous orders, and improve the resolution between peaks of the spectra. Italso gradually changed peak shape and removed peaking operationdue to the increase of orders, which resulted in the gradual change of FD curve of saline soil to the slope of the curve, namely, a detailed description of subtle differences from 0-order to slope and between slope and curvature. (2) Correlation coefficient (CC) between FD value and salt content of five spectral transforms was tested by 0.01 significance level, and max CC absolute larger than integer order (1-order, 2-order) was mainly concentrated at 1.3, 1.4, 1.5 orders. Among them, CC of 1.4 order 1/lgR and 1.3 order 1/R had the largest increase percentage, which were 12.78% and 13.03%, respectively. (3) Regardless of the spectral transformation, the corresponding bands for max CC between salt content and all FD value appeared at 598 nm (1/R) and 618 nm (R, R, 1/lgR and lgR), and they were all in 1.3 or 1.4 order. (4) Na+ accounted for 65.74% of total cations, and its correlation with total salt was 0.738. The 589.3 nm spectrum was the main reason why the bands with the best correlation between various spectral transforms and soil salinity were located at 598 and 618 nm.