The detection of plant water deficit based on hyperspectral technology is the current research hotspot. Fescue is one of the major herbaceous plants which have the maximum usage in northern China, and its growth has a large demand for water, and it will have a series of changes in physical characteristics (color, texture, shape, etc. ) and physiological characteristics under the condition of water deficit. By studying the establishment of plant water content detection model based on hyperspectral technology, the rapid non-destructive monitoring and assessment for the plant water deficit can be achieved, and the plant water status can be diagnosed comprehensively and reliably. The research can provide important basis for predicting the physiological response and change of common herbaceous plants in the North under future climate change. Fescue was sampled to carry out pot control simulation research under constant temperature and humidity conditions. The experiment involves two variables of CO2 concentration (CX) and soil water holding capacity (WX). Two CO2 gradients were set, 400 and 700 μmol·mol-1, respectively. Three water holding capacity treatments were carried out at each CO2 gradient, 100%, 40% and 20% of water holding capacity respectively. And then an ASD Field spec Hand Held spectrometer was used to measure the spectral reflection parameters of the fescue at 10:00—14:00 per day, including Spectral Reflectance (Ri), First Derivative Spectrum (Dλi), Red Ddge Magnitude (Dλr), Red Edge Position (λr), Red Valley Absorption Depth (D), Red Edge Area (Sr), Photochemical Reflectance Index (PRI), Chlorophyll Index (Rch), Normalized Difference Vegetation Index (NDVI), Ratio Vegetation Index (RVI), Normalized Difference Spectral Index (NDSI), Ratio Spectrum Index (RSI), Fractal Dimension (Fd), etc. Regression analysis and statistical model building were applied to analyze the quantitative relationships between spectral parameters and physiological parameters. The complex relationships between spectral characteristic parameters and Fescue leaf water content were analyzed using statistical methods to extract optimal spectral characteristic parameters and subsequently to establish the estimation models of spectral characteristics and water deficit. The results showed that Normalized Difference Vegetation Index (NDVI), Chlorophyll Index (Rch), and Fractal Dimension (Fd) were significantly correlated with leaf water content at the 99% confidence level. So, we reckoned that the three spectral characteristic indicators are the most effective parameters for monitoring water deficit of Fescue. There are good linearities between leaf water content and spectral characteristic parameters, and the formula of detection model is, Y=-0.125XRch+1.714XNDVI-0.023XFd+0.018, and the model test is significant at the 99% confidence level. The results can provide the technical supports for rapid non-destructive monitoring of the drought degree of Fescue, and provide a scientific guidance for large scale irrigation and management of Fescue.