Wheat is one of the most important crops

and its grain protein content varies from 9% to 18%.Grain protein content is one of the important quality index of winter wheat product.The winter wheat can be divided into weak-gluten （≤1.5%）

medium-gluten（

>

11.5% and

<

15%）and strong-gluten types（≤15%）according to Chinese of Stantard grain protein content（GB/T 17892-1999 and GB/T17893-1999）.Therefore

it is valuble to monitor grain protein content from remotely sensed image for winter wheat quality division and food processing at large scale.Almost all the previous re- searches of monitoring grains protein have made use of content empirical statistical model between grain protein content and the canopy or image spectral information.These empirical statistical models were limited frequently by spatial and temporal condition of the experiments.If the grain protein content of winter wheat is monitored by using the empirical statistical model

serious predicting error often occurs due to the variations of the external experiment conditions

such as tempera- ture

sunlight

soil condition

fertilization and irrigation.The crop growth model possesses the traits of better continuity and mechanism

which can simulate winter wheat growth status dynamically for all growth stages

and can interpret the difference resulted from temporal or environmental changes.However

it is difficult to predict grains protein content of winter wheat at large region by the crop growth model

because the spatial distribution of the crop growth model’s parame- ters

such as LAI

above-ground biomass

canopy the temperature and soil moisture cannot be collected without remote sensing technique.Some researches result indicated that from quantitative remote sensing data we can retrieve the spatial distribution information for the growth model parameters.Therefore

it is necessary to monitor grain protein content of win- ter wheat dynamically by assimilating remote sensing data and crop growth model.The field experiments were carried out in Jiangsu and Henan province during 2003 to 2005.The dynamic relations between vegetation index of remote sensing and growth condition of winter wheat were analyzed

and the nitrogen transfer model from plant to grain was built based on the experiment data in Jiangsu province in 2005.As a result

the predicting model for grain protein content of winter wheat was developed based on the normalized difference vegetation index（NDVI）of TM image of the landsat satellite and grain nitrogen accumulation.According to the predicting model

the plant nitrogen accumulation

leaf area index（LAI） and dry above-ground biomass were firstly regressed from NDVI at anthesis

then

the grain nitrogen accumulation was predicted by using the plant nitrogen transfer model from plant to grain.The model was validated using the data sets of dif- ferent ecological regions in Henan province in 2003 and in Jiangsu province in 2004.The root mean square error（RMSE） of the predicted grain protein content varied from 0.47% to 0.59%

and the RMSE of the nitrogen accumulation varied from 4.75 kg·hm

-2

to 8.76 kg·hm

-2

.Therefore

the model was accurate and applicable for predicting grain protein content of winter wheat under various conditions

which made it possible to predict grain protein content at large region from remotely sensed images.However

the formation of protein content in winter wheat grain is very complex

which is not only influenced by genotype

but also by environmental conditions

such as air temperature

solar radiation

soil con- ditions

fertilization and irrigation.Besides

remote sensing image is influenced by the function sensor

the atmospheric conditions

and imaging angle.Therefore

the model presented in this paper still need to be tested by remote sensing exper- imental data under different environments.