Relative discrimination of planophile and erectophile wheat [Triticum aestivum] types using multi-temporal spectrum measurements

Over-luxuriant winter wheat can always be seen when pre-winter accumulated temperature is high or too much fertilizer is applied before sowing. Over-luxuriant erectophile type wheat shows a similar canopy spectral feature to that of the planophile type. This special canopy spectral feature influences N fertilizer application decision making when using remote sensing technologies to detect N status. The objective of this study was to relatively discriminate erectophile type wheat from the planophile type and then further detect over-luxuriant canopy using seasonal changes in canopy spectral reference. Two groups of winter wheat cultivars with different canopy structures were investigated. The wheat leaf distribution feature was combined with the ratio of canopy spectral reflectance of 790 nm at elongation stage (R790sub(e)) to that at rising stage (R790sub(r)), i.e. R790sub(e)/R790sub(r). Based on the values of both normalized R790sub(e) (NR790sub(e)) and normalized R790sub(e)/R790sub(r), (N(R790sub(e)/R790sub(r))), a two-dimensional spectral coordinate system charactering seasonal change in canopy spectral reflectance was constructed. According to the distribution features, cultivars with N(R790sub(e)/R790sub(r))=1 were discriminated as the planophile type in which cultivars with NR790sub(e) =1 were normal and over-luxuriant (or dense) planophile type, and those with NR790sub(e) 1 were considered the sparse planophile type. On the contrary, cultivars with N(R790sub(e)/R790sub(r)) 1 were grouped in the erectophile type in which cultivars with NR790sub(e) 1 were considered normal and sparse erectophile type, and those with NR790sub(e)=1 were attributed to the over-luxuriant erectophile type. Thus, N(R790sub(e)/R790sub(r) could be used to relatively discriminate wheat canopy structure, and NR790sub(e) to judge wheat canopy density. The larger the NR790sub(e) value, the higher the canopy density. The average discrimination accuracy of our method was 84.4%. Nevertheless, this method could provide the relative positions only within the investigated cultivars since both NR790sub(c) and N(R790sub(e)/R790sub(r)) were relative ratios to their average.