Abstract: Abstract: Rapid and non-destructive estimation of leaf salt ion concentrations in halophytes can provide valuable information for plant growth monitoring, selection of salt-tolerant plants and soil salinity monitoring. In this study, the canopy reflectance (350-2 500 nm) and the leaf salt ion (K+, Na+, Ca2+, Mg2+) concentration in the halophytes were measured in the Ebinur Lake Protection Zones, Xinjiang, China. Data collected includes hyperspectral data and leaf salt ion data, and the relationships between the leaf ion concentrations and the selected spectral indices were analyzed. K+ sensitive wave bands on the photosynthetic effective radiation area of the 400- 700 nm (photosynthetically available radiation, PAR), and focused on the red and yellow areas without differential transform; The sensitive bands of Na+ are concentrated in the near infrared region of 949- 1 355 nm. Ca2+ sensitive bands were concentrated in the visible red and near-infrared regions of 665-672 and 919-1 283 nm. Mg2+ sensitive bands were mainly concentrated in 384, 651- 669 nm, mainly in the visible red light region. There was a certain correlation with the ultraviolet region band, but the correlation was generally small. The correlation between the original spectrum and K+ and Na+ was relatively high, reaching a significant level. Spectral transformation increased the correlation between the contents of Ca2+ and Mg2+ and the spectrum, so that modeling bands could be selected according to the standard. Spectral transformation could improve the correlation between the content of salt ions and the spectrum. There were 64 samples in total, and the proportion of samples used for modeling and verification was 3:1. R2 and root mean squared error (RMSE) were used as accuracy evaluation criteria. A Geographically Weighted Regression (GWR) model and a back propagation (BP) model were constructed for estimating leaf salt ion concentrations with the spectral transform and the spectral indices as ratio vegetation index (RVI), difference vegetation index (DVI) and normalized difference vegetation index, and achieved a promising accuracy. The GWR estimation was the best in the bands in the red light region selected by the reciprocal logarithm of reciprocal of reflectance. The characteristic bands of Na+ were concentrated in the short-wave infrared region under the spectral transformation, and the two-dimensional vegetation index was concentrated in the near-infrared region, short-wave near-infrared region, yellow, orange and red region. The short-wave infrared band selected under first order of square root for Ca2+ content had the best estimation effect through GWR model. Mg2+ content was best estimated in the characteristic bands in the red light region selected by DVI, but the GWR model was not as accurate as BP model in estimating Mg2+ content. Based on the GWR salt ion model, the estimation of ions with higher content was better, and the accuracy of K+ and Na+ models was better than that of Ca2+ and Mg2+. When the GWR model was used to estimate the salt ion content in plant leaves, the characteristic bands all pointed to red and short-wave infrared bands. The model based on logarithms of reciprocal of reflectance and GWR for estimated K+ produced the superior performance (R2=0.930, RMSE=0.018 mg/kg). The optimal GWR model with the highest R2 and lowest RMSE was estimation model on Na+ (R2=0.984, RMSE=0.041 mg/kg) via processing. For the estimation model on Ca2+, the model produced reasonable outcome using first order of square root of reflectance-GWR strategy. Moreover, compared with BP model, the GWR model had insufficient estimation for Mg2+ whereas DVI scheme contributed to improve accuracy of the BP estimated model. By comparison, the GWR model yielded better results in higher-content ion models. In conclusion, our study showed GWR model was effective for estimating leaf salt ions through vegetation spectral information. Sensitive bands for salt ions were prominent in the red bands and short-wave infrared bands, which were consistent with the response of vegetation spectral mechanism.