Crop harvest index (HI) is a key parameter in grain yield simulation from indirect biomass estimates and varies with crop growth conditions. How to access the HI still remains an issue to be addressed. Several HI estimation techniques for wheat have been proposed based on the fraction of water transpired after anthesis (theta(E) or theta(Ee)), the fraction of biomass accumulation after anthesis (f(G)) and the ratio of the mean normalized difference vegetation index (NDVI) after anthesis to that of the pre-anthesis period (NDVI(post)/NDVI(pre)). In this paper, efforts were made to test the performances of the above HI estimation methods based on field measurements of winter wheat at different nitrogen (N) levels. Results showed that the HI changed significantly from low-N to high-N treatments, while the theta(E) or theta(Ee) was not sensitive to the N deficiency. In addition, no consistent effect of N fertilizer on f(G) was observed. Using the canopy NDVI to estimate HI seemed workable since strong correlations were found between HI, the term 'NDVI(post)/NDVI(pre)' and the fraction of cumulative NDVI after anthesis. Further analysis suggested a good estimation of HI can be obtained just using the NDVI dataset from anthesis to maturity, especially via binary regression relations of HI, NDVI at anthesis and cumulative NDVI after anthesis. The proposed method in this study may be well oriented to practical application as its short-term dataset is relatively easy to access from satellite platforms. (C) 2011 IAgrE. Published by Elsevier Ltd. All rights reserved.