309 
HYPERSPECTRAL DATA ANALYSIS OF N-FERTILISATION EFFECTS ON WINTER 
WHEAT: A CASE STUDY OF HUIMIN COUNTY, NORTH CHINA PLAIN 
M. L. Gnyp a ’ *, F. Li b ’ c , S.D. Hennig 3 , W. Koppe 3 , L. Jia b ’ d , X. Chen b , F. Zhang b , R. Laudien 3 , G. Bareth 3 
a Department of Geography, University of Cologne, 50923 Cologne, Germany - (mgnypl, simon.hennig, r.laudien, 
g.bareth)@uni-koeln.de, wolfgang.koppe@web.de 
b College of Natural Resources and Environment Sciences, China Agricultural University, 100094 Beijing, China - 
(jiall, chenxp, zhang)@cau.edu.cn , cau_lifei@hotmail.com 
c College of Ecology and Environment Sciences, Inner Mongolia Agricultural University, 010019 Hohhot, China 
d Institute of Agriculture Resource and Environment, Hebei Academy of Agricultural and Forestry Sciences, 050051 
Shijianzhuang, China 
Commission VII, WG VII/3 
KEY WORDS: Hyperspectral Remote Sensing, Classification, Image Understanding, Data Integration, Crop Management, 
Hyper-Spectral Instrument 
ABSTRACT: 
High nitrogen fertiliser applications are very common in many densely populated areas as in the North China Plain (NCP). A still 
growing population and an ongoing economic progress enforce that process. Remote sensing methods can help to optimize the 
N-management for diagnosing crop N status on a field and regional level. For that purpose, spectral and agronomic data were 
collected during the vegetation period of winter wheat (Triticum aestivum L.) in 2006 and in 2007 in the study area of Huimin 
County. In a post-processing step, a hyperspectral and agronomic data library was established which enables the analysis of spectra 
in dependence of agronomic data. The investigated parameters include biomass, N-uptake, LAI (Leaf Area Index), plant height and 
others. Finally, a model was developed which could be applied for extrapolation of the regional knowledge by using hyperspectral 
air or satellite borne remote sensing data. 
This paper shows that the collection and postprocessing of spectral and agronomic data of winter wheat in combination with GIS 
(Geographic Information Systems) and RS (Remote Sensing) analysis help to identify over-fertilised and undersupplied 
managements for different phenological stages from shooting to heading. 
1. INTRODUCTION 
The agricultural area in China declines in many densely 
populated areas due to a fast urbanization, an ongoing growth of 
population and continuing economic progress. Further negative 
effects are caused by desertification and erosion in sparsely 
populated areas. In order to compensate these losses of 
agricultural land, high fertiliser, pesticide, fungicide and other 
measures are applied to increase yields. As a result of the green 
revolution the fertiliser input increased and quintupled from 
6 bn. t in 1977 to 32 bn. t. in 2006 in China (FAO, 2004). 
Consequently, the environmental pollution from agricultural use 
is a severe issue in China and especially in the North China 
Plain (NCP), which is one of the most important areas for crop 
production in the country. Traditionally, the yearly N-input per 
ha amounts to approximately 600 kg to 800 kg by conventional 
methods in the NCP (Boning-Zilkens, 2004; Jia et al., 2004) for 
usually 2 harvests of a winter wheat/ summer maize rotation. 
An optimal N-fertilisation is very important for a profitable 
harvest. Chen et al. (2006) show that N min -based fertilisation 
management results in moderate harvest. Time-consuming 
N-soil and N-crop analyses are required for evaluation during 
and after the growth stage (GS). Remote sensing methods have 
been used in order to estimate crop N-status in wheat 
(Flowers et al., 2001; Serrano et al., 2000; Wright et al., 2004). 
The analysis helps to derive significant parameters which can 
be used to fulfil the requirements for environmental problems in 
large areas (Haboudane et al., 2002). The combination of 
spectral reflectance and agronomic parameters show linear 
correlations for winter wheat crops at the shooting and heading 
stage. Based on a spectral and agronomic library many 
combinations can be tested and adequate models can be 
developed. 
Different N-treatments were detected by spectral measurements 
during a field campaign as the N-status affects the chlorophyll 
content in the plants. Deficient N-supply causes earlier 
chlorophyll degradation in winter wheat crops compared to an 
optimal supply (Biiker, 1992; Schellberg, 1990). The economic 
situation on the world market and the doubling of nutritional 
needs in the forthcoming decades (Meng and Chang, 2004) 
illustrate this problem. Hence, recent research projects in 
precision agriculture are focusing on the relationship between 
N application rate, yield and environmental pollution 
(Serrano et al., 2000), the workflow of this study is based on 
three steps: (1) spectral and agronomic data collection, (2) 
establishment of a hyperspectral and agronomic data library and 
(3) model development for knowledge extrapolation. 
* Corresponding author.