The reed leaves in Ebinur Lake Wetland Nature Reserve were sampled and water content was measured. The spectrum of reed leaves was measured by FieldSpec3 portable spectrometer. The original spectrum of reed leaves was smoothed and derivative transformation was carried out. The response characteristics of the first derivative of reed leaves to water content change under different differential window scales were analyzed， and the characteristics were extracted. The water content of reed leaves was characterized by characteristic parameters. The results show that when w = 1－15， the noise of the first-order spectral curve is difficult to be effectively reduced， which makes the contour of the spectral curve difficult to distinguish， and the spectral response characteristics caused by the change of leaf water content are not easy to judge. When w = 16－30， the first derivative spectrum noise obtained can be effectively removed. Compared to the correlation coefficient between the first derivative spectrum of phragmites reed leaves and the water content in the 1374－1534 nm band， the correlation coefficient in 1460－1500 nm band demonstrates small fluctuations， small standard deviations， and stable performance， the response feature between the whole first derivative spectrum within 1460 nm and 1500 nm， and water content variation has more consistency and significantly response. Take the average value of the first derivative spectral value in the 1460－1500 nm band as the characteristic parameter. Under the differential range of 1－30， compare the characteristic parameter with other differential ranges， it is more suitable to indicate the change of reed leaf water content， which provides a new way for quantitative detection of reed leaf moisture by short-wave infrared reflectance spectroscopy in remote sensing monitoring of reed water in arid area.