The multispectral reflectance of shortwave radiation by agricultural crops in relation with their morphological and optical properties

Relations between morphological properties of uniform canopies. optical properties of the leaves and reflection of shortwave radiation, in the visible light region and the near infrared, by crops are the subject of this thesis. The aim of the study was a further investigation of potential applications of multispectral scanning for agricultural purposes by a fundamental approach. Multispectral scanning is a modem aerial survey technique, based on the simultaneous measurement of radiation reflected by the earth's surface within a discrete number of wavelength bands. These wavelength bands are distributed in the wavelength interval encompassing the visible light and the near reflectance infrared. Applications of thermal infrared radiation and reflected or emitted microwave radiation are left out of consideration. Chapter 1 gives a general introduction to the relation between aerial photography, remote sensing and multispectral scanning in particular. The objectives of the research are described and interrelated with potential applications in agriculture. In chapter 2 attention is first paid to the physical relations dealing with the remote detection of radiation reflected by the earth's surface. Current knowledge concerning reflection, transmission and absorption of shortwave radiation by single leaves is summarized. With this the optical behaviour is related to leaf anatomy, the present pigments and the leaf water. After mentioning the predominating factors determining reflectance of bare soils, reflectance of homogeneous crops is treated. Utilizing a description of crop structure and the optical properties of leaves as most important canopy components, several mathematical models of canopy reflectance are discussed. A deterministic model, published by G. H. SUITS in 1972, is adopted for this study and further elaborated. A comparison with a numerical model according to GOUDRIAAN (1977) showed that a drastic simplification of canopy structure introduced initially led to similar results as, when the realistic numerical model was simplified afterwards. The actual leaf angle distribution is characterized by one parameter only in Suits' model. The reflectance as function of canopy parameters and the directions of incoming radiation and observation is described by an analytical expression. This also holds for the reprocity relations connected with and found earlier by Goudriaan. By means of Suits' model, a sensitivity analysis is performed in chapter 3 for the relation between reflectance as a function of wavelength and canopy variables, like leaf area index, leaf angle distribution, leaf colour and soil reflectance. Information distributions calculated from own measurements were confirmed by model predictions. Taking account of selective absorption of radiation by the atmosphere, spectral bands with a high sensitivity of reflectance to variation in crop parameters have been chosen. Reflectance in the green at 550 nm and in the red at 670 nm in the visible light region and in the near reflectance infrared plateau at 870 run and in the water absorption region at 1650 nm and 2200 nm were selected for a more detailed study. Canopy reflectance at these wavelengths has a direct relation with leaf pigments, leaf morphology and the water content of the leaf. The sensitivity of the measured canopy reflectance to variations in canopy structure dependent on spectral bandwidth was investigated for the green, the red and the near infrared plateau. In the red with 670 nm as centre wavelength, a bandwidth of 20 nm appeared to be acceptable, in relation to the loss of variance, as well as to the increased covariance with the spectral band in the green with equal width. In the near infrared, bandwidth is allowed to be larger, at least 100 nm since the small spectral variation of the optical properties of the leaves in the infrared plateau. Chapter 4 gives a description of the in situ canopy reflectance measurements during growth performed at Wageningen, with a specially designed spectrometer. By means of a comparative method canopy reflectance is measured at 153 wavelength values between 361 and 2360 nm The optical system and the principle of the spectrometer used is discussed. Examples of canopy variables measured, like soil cover percentage, leaf area index and biomass are presented. Finally, the principle of multispectral scanning is explained in this chapter. Some attention is paid also to relations concerning ground resolution, spectral resolution, the radiation level, detector properties and limitations due to detector noise. Chapter 2 shows the complex relation between reflectance of radiation by crops and the optical properties of canopy components and structure. In chapter 5 new spectral parameters are found by combining reflectance values of homogeneous green canopies of different spectral bands which show an optimum sensitivity to a canopy parameter under detection, with a high invariability to other parameters. The systematic coherence between the reflectance at different wavelengths. caused by canopy structure and correlations between the optical properties of leaves and of soil is investigated. For 2 combinations of the reflectance in the green, the red and the near infrared plateau, a useful relation with the soil cover percentage has been found. This combination is insensitive to variations in the solar angle and in soil moisture content. Out of the same reflectance values combinations are made which, assuming an incomplete soil cover. are sensitive to detection of soil moisture variations only or, with a high soil cover. sensitive to actual differences in the leaf angle distribution. By observing a crop under an angle of 75° relative to the vertical, the influence of the soil background on the reflectance in the visible light may be neglected in most crops. For the ratio of reflectance in the red and the green it is concluded that variation in this parameter during growth may be attributed uniquely to colour variation in the canopy components. By means of reflectance in the three wavelength bands mentioned, a structure parameter, a soil characteristic and a colour indication for the canopy components may be monitored during the growing season and employed for applications in agriculture. Making use of the analytical expression for plant canopy reflectance it has been demonstrated that for detection in the canopy hot spot (directions of incoming radiation and of observation coincide), the relations between multispectral reflectance and canopy variables are simplified considerably. Detection perpendicular to the earth's surface or under a zenith angle of 51.8°, eliminates the influence of leaf angle distribution function on reflectance as a function of the apparent soil cover percentage. Combinations of reflectance values in the green. the red and the near infrared produce a well-defined relation with soil cover, the leaf area index and differences in soil moisture content. The ratio of reflectance in the red and the green wavelength band for these view angles is independent of canopy structure and is a function of leaf colour only. Conical scanning under an angle of 51.8° using an active radiant source and appropriate sensors, from an airplane offers the facility to collect crop data under cloudy weather conditions, while the disturbing influence of the atmosphere can be eliminated considerably. In chapter 6 an analysis is made on which and how many spectral bands. well-chosen within the atmospheric windows available, are needed at a minimum to discriminate between green crops of different structure, leaf colour and soil reflectance. In this a multi-regression analysis is used with the restriction that by deletion of redundant spectral bands not more than 1 % of the information available is lost. Starting with the already chosen bands in the green. red and near infrared plateau, it appeared for both passive detection perpendicular to the earth's surface and active detection into the hot spot under 51.8° that addition of a fourth band at 1650 rim in the water absorption region produced more than 99% of the information available. The preferred spectral bands were confirmed by the same analysis applied on field reflectance and MSS data. The interclass variations for passive detection in the nadir appeared to be wider than compared to active detection in the hot spot under an oblique angle of 51.8°. This is related with the elimination of the shadow observed and the weaker contrast influence of the bounding soil brought about by the hot spot conditions mentioned. The possibility of discriminating between crops on the basis of differences between spectral reflectance by means of a decision rule, with a small probability of misclassification, depends on the stochastic behaviour of the reflectance values measured, the number and the selection of spectral bands. A criterion for the probability of misclassification, using statistical parameters has been applied to illustrate the separability between potatoes and sugarbeets during the growing season. Field spectrometer data in four spectral bands were used. It appeared that the spectral band in the water absorption region at 1650 nm is a good discriminator for sugarbeets, because of the significant difference in the water content of the single leaf.