Abstract:
Abstract: Normalized difference vegetation index (NDVI) is one of the vegetation indices (VI) most widely used. It is related to quite a few indices indicating crop growth, such as leaf area index (LAI), leaf nitrogen accumulation (LNA), chlorophyll content and so on. Several researchers have developed various field crop NDVI measuring instruments to acquire NDVIs, which contains passive sensors and active sensors. Greenseeker, Crop Circle, CropSpec, RapidScan are in common use, and numerous researches are based on them. However, few domestic devices have been developed and corresponding researches haven't, either. This study assessed the reliability of a novel field crop NDVI measuring instrument based on CropSense crop growth parameter meter (CropSense) together with spectral reflectance sensor (SRS) in varied illumination condition. Field crop NDVI measuring instruments are susceptible to illumination conditions, the height above the canopy and other factors, and it was supposed to detect the performance of sensors through radiometric calibration, which contains laboratory calibration and on-orbit calibration. Integrating sphere was mostly employed for traditional radiometric calibration on field crop NDVI measuring instruments. What was used as the calibration model was the functional relationship between the radiance of integrating sphere and the DN digital number of each channel fitted into the field crop NDVI measuring instruments. Nevertheless, it was found that deviations tended to increase after calibrated by the model built via indoor calibration in the outdoor measurement, with a decline of R2 from 0.997 1 at least of all the calibration model to 0.698 0 for CropSense and 0.877 3 for SRS. In order to deal with such problems, cross calibration method was proposed in this study for the field crop NDVI measuring instruments in the outdoors, with CropSense together with SRS to be calibrated and ASD FieldSpec 4 hyperspectral radiometer (ASD) as reference which can all gather NDVIs. NDVIs were obtained on several objects, referring to diffuse white reference panel, gray cloth, green cloth, blue cloth, cement pavement, naked soil and wheat in varied test periods from 5:30 to 18:30 in sunny days and from 5:30 to 15:00 in cloudy days. The functional relationship between NDVIs acquired through the sensor to be calibrated and ones measured by the reference sensor was taken as an outdoor calibration model to adjust the testing data of CropSense and SRS. Comparing the R2, RMSE and MAE got between field crop NDVI instruments and reference sensor before calibrated, the accuracy and the most suitable conditions for the field crop NDVI measuring instruments were determined. The results showed that RMSEs of NDVIs between CropSense and ASD were 0.074 5-0.104 9 and that between SRS and ASD were 0.026 8-0.054 3 during 6:00-18:00 in sunny days with light intensity from 6.4 to 92.3 klx, and 0.094 2-0.117 9 for CropSense and 0.029 9- 0.070 3 for SRS during 6:00-15:00 in cloudy days with light intensity from 5.2 to 10.8 klx. After calibrated by the model built through outdoor calibration, RMSEs were decreased by 37.1%-41.4% for CropSense and by 10.7%-31.5% for SRS in sunny days; RMSEs were decreased by 33.8%-48.3% for CropSense and by 48.7%-62.6% for SRS in cloudy days. At the same time, the light intensity and solar elevation angle have certain impacts on the accuracy of field crop NDVI measuring instruments. Within the conditions allowed for the instruments to be employed, the performance of the field crop NDVI measuring instruments improves with the increasement of the light intensity; and there is a slight fluctuation with the change of the solar elevation angle. However, the both factors, generally speaking, have no significant influence on the accuracy of the instruments, as long as they are utilized in the illumination conditions more than 5.2 klx. This research results provide a scientific calibration method and momentous reference for passive instruments of vegetation indices measurement.