Rice is a staple grain and accounts for over 40 % of the grain protein production in China. It is not only the main food in China but also in other highly populated countries in Asia. For the security of food production and quality, the estimation of agronomic parameters is an important task for decision support in rice cultivation. Methods based on remote sensing can support related analyses to attain this goal. The objective of this contribution is to monitor the rice (Oryza sativa L., irrigated lowland rice) growth parameters during the growing seasons from 2007 to 2009. Therefore, spectral and agronomic multitemporal data were collected.
The study area is Qixing farm which is located in the northeastern part of Heilongjiang Province (China) close to the Russian border. It belongs to the Sanjiang Plain and to the northernmost rice cropping system in China. Hyperspectral data were collected using field spectroradiometers (ASD Inc.: QualitySpec and FieldSpec3) in wavelength ranges from 350 nm to 2500 nm. The field measurements were carried out together with agronomic data collection (e.g. biomass, N concentration, number of tillers). Two field experiments were set up at different sites of the farm, each with 7 N fertilizer rates: 0, 60, 75, 90, 105, 120, and 150 kg N ha-1. Each N-management had three replications. Additionally, farmers' fields (12 ha to 27 ha) were selected in order to evaluate the results on a field scale under farmers' management practices. The annual measuring campaign started during tillering stage and ended when heading stage was reached.
Multispectral and hyperspectral vegetation indices (VIs) were calculated using sensitive wavelength and best band combinations. The performance of the standard and narrow band VIs was evaluated in terms of several statistical measures: coefficient of determination, root mean square error and relative error for every growth stage, over several stages and as well as several years. High correlations were found for the VIs based on the reflectance and absorption peaks in the VIS, NIR and SWIR. Additional analysis methods like multiple linear regression and partial least square models were applied to establish regression models based on the data of the field experiments. These models were validated using data from the farmers' fields. The results suggest that hyperspectral-based methods can estimate rice agronomic parameters with a high accuracy.
In the context of precision agriculture, the findings are useful for future development of new hyperspectral devices such as scanners or cameras which could be fixed on tractors or unmanned aerial vehicles (UAVs).

Keywords: Agriculture, Crop, Hyper spectral, Estimation, Infrared, Multitemporal.