HYNIM - Hyperspectral derived nitrogen indicators for maize crop

U bent hier

Context and objectives

Nowadays, agricultural activities are no longer limited to food production. Even if it is still its major role, agriculture also has to reduce its impacts on the environment, to protect the rural territory and to insure its durability. The fertilizer surpluses not used by plants are leached and can then pollute surface and groundwaters. This lost of N also constitute an additional cost for the farmers who can not get the best return of it. On the other hand, a nitrogen deficit can limit yields and compromise the soil fertility in the long term. Hyperspectral remote sensing, due to the high flexibility in the selection of wavelengths, can play a major role in the detection of parcels fertilized in a inappropriate way. This research consists in setting up agri-environmental indicators to detect parcels of maize that show an abnormal reflectance which may be explained by either an excess or a lack of N fertilization.

Project outcome

Expected scientific results

Simple correlations and multiple regressions (stepwise, principal component and partial least square) were used to select spectral bands which can explain most differences in maize nitrogen content. Only five bands in green (558 nm), red (695nm) and NIR (756, 896 and 1099 nm) were retained. Several combinations of these narrow bands (normalized difference, simple ratio) and other specific indicators (TCARI, GNDVI, Red-edge) were also tested in order to estimate their ability to react differently according to variable nitrogen content. Indicators derived from narrow bands were more accurate than those derived from larger bands in the same spectral zone and analyses of variance show that fertilization level but not variety has a significant influence ( ) on the variation of indicators. These results are encouraging and suggest that use of hyperspectral imagery is a promising tool to detect unsuited fertilization parcels.