Reflectance relations in row crop scenes

Abstract

Models that relate composite reflectance to its components are useful for inferring crop growth information from measured scene reflectance. Radiation measurements in Thematic Mapper bands (TM1, TM2, TM3, and TM4) were made from cotton, soybean, sunflower and grain sorghum at three stages of growth and used to evaluate three reflectance models. Two models, AIRM1 and AIRM2, assumed that scene components contribute in an additive independent manner to composite reflectance. The third model, TRIM, assumes that radiation transmitted through the canopy interacts with bare soil in two scene components. The AIRM2 and TRIM models divide the composite reflectance into canopy, bare soil, and shadow components, but AIRM1 considers only canopy and bare soil. Ranking of models in order of decreasing accuracy for predicting composite reflectance in bands TM3 and TM4 was AIRM2, TRIM, and AIRM1. The AIRM1 and AIRM2 models estimated average TM3 reflectance at full plant cover between 1 and 4 per cent for all crops. Their estimations in band TM4 were 60 per cent for cotton, soybean, and sunflower with grain sorghum being 50 percent.

Measured canopy and composite reflectances were graphically compared at the lowest and highest levels of canopy cover observed in each crop. Measured band TM3 canopy reflectance did not change with solar zenith angle, composite reflectance decreased with increasing zenith angle at the lowest canopy cover levels but was invariant at the highest canopy cover levels. Measured band TM4 canopy reflectance increased linearly with increasing solar zenith angle in all crops, but for composite reflectance this pattern was observed only at the highest canopy cover levels of cotton and soybean. The absence of a uniform pattern between band TM4 composite reflectance and solar zenith angle in grain sorghum is presumably due to large horizontal leaf angles and in sunflower to long vertical spacings between leaves. Predicted compared to measured band TM3 and TM4 composite reflectances of the AIRM1 and AIRM2 models were insensitive but the TRIM model was overly sensitive to zenith angle.     

Remote sensing of leaf water content in the near infrared

A stochastic leaf radiation model was used to predict leaf spectral reflectance as a function of leaf water content for a dicot leaf. Simulated spectral reflectances were analyzed to quantify reflectance differences between different equivalent water thicknesses. Simulated results coupled with consideration of atmoshperic transmission properties and the incident solar spectral irradiance at the earth's surface resulted in the conclusion that the 1.55–1.75 μm region was the best-suited wavelength interval for satellite—platform remote sensing of plant canopy water status in the 0.7–2.5 μm region of the spectrum.     

Measurement of leaf relative water content by infrared reflectance

From basic considerations and Beer's law, a leaf water content index incorporating reflectances of wavelengths from 0.76 to 0.90 μm and from 1.55 to 1.75 μm (Landsat Thematic Mapper Bands TM4 and TM5, respectively) was developed that relates leaf reflectance to leaf relative water content. For the leaf succulent, Agave deserti, the leaf water content index was not significantly different from the relative water content for either individual leaves or an entire plant. Also, the relative water contents of intact plants of Encelia farinosa and Hilaria rigida in the field were estimated by the leaf water content index; variations in the proportion of living to dead leaf area could cause large errors in the estimate of relative water content. Thus, the leaf water content index may be able to estimate average relative water content of canopies when TM4 and TM5 are measured at a known relative water content and fraction of dead leaf material.     

Determination of land surface spectral reflectances using Meteosat and NOAAA/AVHRR Shortwave Channel Data

Abstract

A method to derive surface spectral reflectances from currently available Meteosat geostationary and NOAA/AVHRR polar orbiting satellite data is described. Broadband reflectance was derived from Meteosat measurements while NOAA/AVHRR vegetation index provided a spectral weighting which enabled the spectral reflectances on either side of 0-7 μm to be estimated. The method takes into account satellite calibrations, viewing geometry, and correction of some atmospheric effects. Conversion from narrow-band to broadband reflectances is discussed. The method was applied to a month of data to obtain the surface spectral reflectances of Africa which are compared with some data sets used by climate modellers, in order to assess them and to monitor their seasonal and interannual changes on a global scale.     

A multi-angle spectrometer for automatic measurement of plant canopy reflectance spectra

The paper describes the design and operation of a multi-angle spectrometer (MAS) for automatic measurement of near-field spectral reflectances of plant canopies at hourly intervals. A novel feature of the instrument is a rotating periscope connected to a spectrometer via a fiber optic cable. Canopy reflectances are calculated for multiple view azimuths, at a single zenith angle from measurements of spectrometer dark current, incoming solar irradiance and reflected radiances. Spectral measurements are made between 300 and 1150 nm wavelength at a band-to-band spacing of 3 nm, and a bandwidth (full-width, half maximum) of 10 nm. Preliminary data analysis showed that the canopy reflectance model of Kuusk [Kuusk, A. (1995). A fast, invertible canopy reflectance model. Remote Sensing of Environment 51, 342-350] reproduced the observed large differences in visible and near-infrared (NIR) reflectances, but the model was unable to predict quantitatively the observed variations in the measured reflectance spectra with azimuth, particularly in the NIR. Discrepancies between model and measurements are likely due to the inhomogeneous nature of the forest canopy in contrast to the assumption of a uniformly absorbing turbid medium in the model. Measurements using the MAS can be used to investigate directional dependences of reflectance indices and for testing BRDF models used to separate geometrical and plant physiological contributions to the reflectance signals. The MAS provides continuous sampling of reflectance indices which can be compared with canopy properties such as chlorophyll content and photosynthetic capacity.     

Foliar spectral properties following leaf clipping and implications for handling techniques

After leaves are clipped their reflectance properties change over time at variable rates. Spectral change can in part be attributed to the changing water content of the leaf, which affects absorption in the VIS, NIR and the SWIR. Maintaining water volume within samples has been the motivation behind many leaf handling techniques. This study has assessed the time constraints between leaf collection and spectral measurement. Specifically the relationship between leaf water content and foliar spectra (350-2500 nm) was examined over time for five tropical trees (common guava (Psidium guajava), purple guava (Psidium littorale), weeping fig (Ficus benjamina), floss silk (Chorisia speciosa), and coffee (Coffea arabica)). This investigation was carried for leaves wrapped with moist gauze around their petiole (treatment leaves) and leaves with no treatment. Spectral measurements and mass measurements were repeated for each leaf once every hour for the first 12 h, then every 4-6 h for 18 h, followed by one measurement after 12 h, and finally once a day until the control samples became air-dry. Foliar reflectance in the visible spectrum was not immediately responsive to water content changes and did not change until wilting of the leaf was observed. The NIR and SWIR wavelength regions were affected immediately by small changes in water content. Thus, by the time wilting was first observed the NIR and SWIR foliar reflectance differed considerably from corresponding fresh leaf reflectance. No common time limit could be observed for leaf clipping and reflectance measurement. Leaves have a variety of water contents and dehydration rates hence measurement time constraints are dependent on the properties of the leaf or species. Rather than using a time limit it is recommended that leaf handling techniques be based upon managing leaf water content and leaf structure. The results of this study indicate that leaves with petioles wrapped in moist paper towel and placed within plastic bags will maintain leaf reflectance longer than equivalent leaves without treatment; samples tested here lasted a minimum of 7 days. θ and D indices (“angle difference” and “root mean square difference”, respectively) revealed a stronger relationship between leaf water content and spectral shape than between leaf water and raw reflectance magnitude. The ratio of 1187/1096 nm, when compared with θ and D indices and individual reflectance bands, showed the highest coefficient of determination with leaf water content (r² = 0.952).     

Upscaling ground observations of vegetation water content, canopy height, and leaf area index during SMEX02 using aircraft and Landsat imagery

Microwave-based remote sensing algorithms for mapping soil moisture are sensitive to water contained in surface vegetation at moderate levels of canopy cover. Correction schemes require spatially distributed estimates of vegetation water content at scales comparable to that of the microwave sensor footprint (10¹ to 10⁴ m). This study compares the relative utility of high-resolution (1.5 m) aircraft and coarser-resolution (30 m) Landsat imagery in upscaling an extensive set of ground-based measurements of canopy biophysical properties collected during the Soil Moisture Experiment of 2002 (SMEX02) within the Walnut Creek Watershed. The upscaling was accomplished using expolinear relationships developed between spectral vegetation indices and measurements of leaf area index, canopy height, and vegetation water content. Of the various indices examined, a Normalized Difference Water Index (NDWI), derived from near- and shortwave-infrared reflectances, was found to be least susceptible to saturation at high levels of leaf area index. With the aircraft data set, which did not include a short-wave infrared water absorption band, the Optimized Soil Adjusted Vegetation Index (OSAVI) yielded best correlations with observations and highest saturation levels. At the observation scale (10 m), LAI was retrieved from both NDWI and OSAVI imagery with an accuracy of 0.6, vegetation water content at 0.7 kg m⁻², and canopy height to within 0.2 m. Both indices were used to estimate field-scale mean canopy properties and variability for each of the intensive soil-moisture-sampling sites within the watershed study area. Results regarding scale invariance over the SMEX02 study area in transformations from band reflectance and vegetation indices to canopy biophysical properties are also presented.     

The effect of specular reflectance on the relationship between reflectance and vegetation amount

Abstract

The specular reflectance of a leaf is unrelated to wavelength or leaf content. However, a vegetation canopy is not a large leaf and specular reflectance is likely to be related to wavelength and vegetation amount because of the correlation between canopy geometry and vegetation amount. It was hypothesised that if the specular component were removed from the total (specular and diffuse) reflectance of a canopy then the strength of the correlation between diffuse reflectance and vegetation amount would decrease in near-infrared wavelengths and increase in visible wavelengths.

Field based measurements of grassland using a polarising radiometer verified this hypothesis. It was recommended that where possible the specular component of the total reflectance be determined, at least in visible wavelengths, prior to the estimation of vegetation amount.     

Influence of wind on crop canopy reflectance measurements

Remote sensing techniques of measuring red and far-red crop canopy reflectance are frequently used to estimate crop canopy characteristics. The variability introduced in reflectance data from nonvegetative factors such as wind decreases the usefulness of the techniques. The objective of this study was to quantify and minimize the variability from wind on spectral reflectances. Red and far-red reflectances were acquired above wheat, barley, and alfalfa canopies throughout days of changing wind conditions. Periods of 312 s with little changes in irradiance values were used for the analysis. Wind had negligible effect on reflectances of a short canopy such as cut alfalfa, while it had a significant effect on reflectances from canopies with a higher vertical structure, particularly during gusty conditions. Within the windy and calm periods, extreme values of spectral reflectance differed by 60% and 12%, respectively, in the red, and by 40% and 8% in the far-red for the barley canopy. For the compact and dense canopy structure of alfalfa, these differences reached a maximum of 10% under windy conditions in both spectral regions. The plant canopy architecture, the wind conditions, and the spectral regions all affected the magnitude of the influence of wind on crop canopy spectral reflectances. The mean reflectance of a canopy overestimated the true reflectance by 2–4% while the use of the median reduced this overestimation. Sampling requirements for this sensor are evaluated, and the possibility of decreasing either the sampling rate or the sampling period is discussed.     

An algorithm for snow and ice detection using AVHRR data An extension to the APOLLO software package

Abstract

The AVHRR (Advanced Very High Resolution Radiometer) Processing scheme Over Land, cLbud and Ocean (APOLLO) is used to extract surface and cloud parameters from satellite data. Before these parameters can be computed, it is necessary to distinguish between land and ocean surfaces and to apply algorithms for the detection of partially cloudy and cloud-filled pixels. In regions with seasonal or permanent snow and ice coverage the separation of clouds becomes much more difficult or often impossible. For this reason, and to find cloud-free and partly cloudy snow and ice pixels,- a day-time algorithm has been developed which uses all five AVHRR channels as follows: The threshold testing of the reflected part of channel-3 radiance leads to a definite distinction between snow/ice and water clouds due, to the clouds much higher reflectivity at 3.7 μm. The detection; of sea ice is based on threshold tests of visible reflectances and, in particular, of the temperature difference between channels-4 and -5. Snow is identified if a high visible reflectance is combined with a low reflectance in channel-3 and with a ratio of channel-2 to channel-1 reflectances similar to that of a cloud. The latter criterion is also mostly suitable to distinguish between snow-covered and snow-free ice areas. Some tests of this algorithm applied to AVHRR data from the 1987 Baltic Sea ice season have shown reasonable classification results with the exception of a few areas with ice clouds or with ice topped water clouds.     

The derivation of water volume reflectances from airborne MSS data using in situ water volume reflectances,and a combined optimization technique and radiative transfer model

Abstract

Water volume reflectance images of three water cooling reservoirs were derived from airborne Daedalus DS-1260 multispectral scanner (MSS) data using a radiative transfer model to eliminate atmospheric effects and to derive downwell-ing irradiances. Sixty in situ water volume reflectances and the associated sensor and sun geometries were input into the radiative transfer model. Using the radiative transfer model to generate water volume reflectance values at each field site, an optimization procedure minimized the difference between modelled and in situ water volume reflectances resulting in optimized sensor brightness value to radiance conversion gain and bias factors, and optimized surface visual range and mean terrain background reflectance. Subsequently water volume reflectance images were derived at six wavelengths encompassing the blue to near-infrared using the optimized atmospheric parameters and conversion factors as inputs into the radiative transfer model. Modelled reflectance images were evaluated for accuracy by statistical comparison to the in situ reflectances, and for improved contrast by subjective comparison to the original images. Daedalus DS-1260 MSS bands 3, 4, 5, 7 and 8 modelled reflectances explained 19, 78, 86, 89 and 82 per cent, respectively of the in situ reflectance variances, while band 2 correlation was not significant (p<0.05). All generated reflectance image histograms showed dramatic improvement in contrast when compared with the histograms depicting variance within the original images.     

Seasonality of finely-resolved spatial structure of NDVI and its component reflectances in tallgrass prairie

Grass canopies exhibit distinct seasonality in reflectance and spatial patterns in reflectance result from landscape heterogeneity. To investigate the spatial structure of NDVI and its two component reflectances at fine resolution in burned and unburned tallgrass prairie, we use range and relative nugget effect two indicators of spatial structure derived from semivariograms. Results show that spatial dependence of the three spectral measures are similar in the early season when the litter layer has been removed by burning and again in the later season when the canopy is senescent. However, as the canopy develops, the spatial dependence of NDVI deviates from that of its component reflectances. In the unburned canopy, red reflectance appears to strongly influence the spatial pattern of NDVI. In the burned canopy, neither component reflectance strictly determines the spatial pattern of NDVI. Maximum divergence in spatial pattern between NDVI and its components coincides with minimum available moisture, suggesting a relation between moisture stress and spatio-spectral heterogeneity.     

Variation in spectral response of soybeans with respect to illumination,view and canopy geometry

Comparisons of the spectral response for incomplete (well-defined row structure) and complete (overlapping row structure) canopies indicated that there was a greater dependence on Sun and view geometry for the incomplete canopies. This effect was more pronounced for the highly absorptive red (0.6-0.7 μm) wavelength band than for the near-infrared (IR) (0.8-1.1 μm) based on relative reflectance factor changes.

Red and near-IR reflectance for the incomplete canopy decreased as solar zenith angle increased for a nadir view angle until the soil between the plant rows was completely shaded. Thereafter for increasing solar zenith angle the red reflectance levelled off and the near-IR reflectance increased. A ‘hot-spot’ effect was evident for the red and near-IR reflectance factors, especially when the Sun-sensor view directions were perpendicular to the rows. The ‘hot-spot’ effect was more pronounced for the red band based on relative reflectance value changes. The effect of Sun angle was more pronounced for view angles perpendicular to the row direction.

An analysis of the ratios of off-nadir- to nadir-acquired data revealed that off-nadir red band reflectance factors more closely approximated straight-down measurements for time periods away from solar noon. Near-IR and greenness responses showed a similar behaviour. Normalized difference generally approximated straight-down measurements during the middle portion of the day. An exception occurred near solar noon when sunlit bare soil was present in the scene.     

Reflectance spectra of minerals and their discrimination using Thematic Mapper,IRS and SPOT multi-spectral data

Abstract

Minerals and rocks show varying spectral reflectances under different spectral ranges of the electromagnetic spectrum and such spectral responses vary because of colour, texture, crystal structure, specific gravity and other physical and optical properties. Hence in order to optimise the spectral ranges which may be used in the recognition of the minerals and the rocks, spectral reflectance measurements were carried out for 29 minerals under the visible and near-infrared ranges of the electromagnetic spectrum covered by the TM bands 1, 2, 3 and 4, IRS bands 1, 2, 3 and 4 and SPOT bands 1, 2 and 3. The analysis of the data identifies the optimum spectral bands for distinguishing the different minerals and the mineral aggregates/rocks     

A field experiment in Saharan Algeria for the calibration of optical satellite sensors

A field experiment took place in February-March 1993 to characterize the reflectance properties of four Saharan desert sites, identified from satellite imagery as having remarkably stable optical properties. The objective of the experiment was to measure reflectances so that they can be used as references for the calibration of optical satellite sensors. Bidirectional measurements of the surfaces in reflectance (and also in polarization) were collected in different planes, in the visible, near infrared, and shortwave infrared. Particular attention was given to instrumental calibration, with independent calibration experiments in the laboratory and in the field. The surface reflectance measurements were then adjusted against an empirical model of bidirectional reflectance, and converted to reflectances at the top of the atmosphere (TOA) using an atmospheric radiative transfer model. The angular dependence of these TOA reflectances, named reference reflectances, was tested against those seen by AVHRR on the four desert sites. This comparison shows that multiangular calibration of optical sensors using reference reflectances can be achieved with an accuracy better than 1 per cent.     

The Pasterze Glacier,Austria

The remote sensing of foliar biochemical concentration assumes that leaf biochemical absorption features will be manifest in canopy reflectance. This is a reasonable assumption providing the effect of a given change in foliar biochemical concentration has a similar effect on both leaf and canopy reflectance. A comparison between canopy and leaf reflectance was made to determine if canopy effects (composite of leaf area index, biomass, structure, multiple scattering and shadow) could alter the leaf biochemical information in canopy reflectance spectra. Differences in leaf biochemical concentrations and leaf biomass were induced by the application of fertilisers to large plots of slash pine (Pinus elliottii var elliottii) in Florida, U.S.A. The reflectance of plot canopies was measured using the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). The reflectance of samples of leaves drawn from each plot were measured using a laboratory spectrometer. The differences between airborne and laboratory reflectance ratios (fertilised/control spectra) were used to isolate the effects of the canopy in AVIRIS reflectance spectra. From this study it was concluded that the canopy influenced leaf reflectance substantially at wavelengths beyond the water absorption feature at 1400nrn and leaf biochemical information was transmitted virtually unchanged from the leaf to the canopy in near-infrared wavelengths.     

Inversion of vegetation canopy reflectance models for estimating agronomic variables. I. Problem definition and initial results using the suits model

An important but relatively uninvestigated problem in remote sensing is the inversion of vegetative canopy reflectance models to obtain agrophysical parameters, given measured reflectances. The problem is here formally defined and its solution outlined. Numerical nonlinear optimization techniques are used to implement this inversion to obtain the leaf area index using Suits' canopy reflectance model. The results for a variety of cases indicate that this can be done successfully using infrared reflectances at different views or azimuth angles or a combination thereof. The other parameters of the model must be known, although reasonable measurement errors can be tolerated without seriously degrading the accuracy of the inversion. The application of the technique to ground based remote-sensing experiments is potentially useful, but is limited to the degree to which the canopy reflectance model can accurately predict observed reflectances.     

Panicle contribution to bidirectional reflectance factors of a wheat canopy

Bidirectional reflectance factors (BRFs) of crop stands are strongly influenced by canopy architecture. In wheat, as well as in many other crops, canopy architecture changes dramatically with the phenological development of the plant community.

A ground-based experiment was performed to examine the effect of panicles of winter wheat (Triticum aestivum L.) at the flowering stage on canopy BRFs. Reflectance factors were measured in the field with a portable radiometer in the red (0-63-0-69 μm) and near-infrared (0-76-0-90 μm) wavelength intervals. Observations were made at three viewing angles and 14 solar zenith angles during two consecutive days on a control target and on a target where panicles had been removed.

Panicles did not contribute significantly to the red nor to the near-infrared (NIR) reflectance factors computed from nadir observations. Off-nadir NIR reflectance was also not altered by the presence of panicles, but was moderately sensitive to illumination angle. Off-nadir red reflectance in the backscattcring direction was higher in the canopy with panicles than in the canopy without panicles: at a solar zenith angle of about 50° the difference in the reflectance of the two targets reached a maximum of about 39 per cent.

These findings imply a potential to identify crops and their phenological development by more fully exploiting reflectance at several different viewing and solar angles.     

Spectral reflectance of dehydrating leaves: Measurements and modelling

Remotely sensed measurements at optical wavelengths may provide information on crop water status and increase the accuracy of crop production forecasts. Previous research has shown that canopy spectral response to water stress is attributable to change in leaf water content, canopy structure and soil moisture. This experiment was designed to study leaf spectral response resulting from changes in leaf water content and to evaluate the use of a radiative transfer model for predicting the spectral behaviour of the leaf. The difference between measured and modelled reflectance increased as leaf water content decreased and it was hypothesized that this may be due to a change in leaf internal structure that was unaccounted for by the model.     

Evaluating temporal variability in the spectral reflectance response of annual ryegrass to changes in nitrogen applications and leaching fractions

A two‐year study was conducted in 2002 and 2003 at the University of Nevada, Las Vegas's center for urban water conservation to assess canopy spectral response of annual ryegrass (Lolium multiflorum Lam.) grown under various combinations of N and irrigation (based on leaching fraction: LF) treatments. Multispectral measurements were acquired using a ground‐based spectroradiometer (200–1100 nm) on a biweekly basis during the growing season (October–May) in 2002 and 2003. Multispectral parameters were correlated with soil–plant parameters and temporal variability was investigated. Results showed that the normalized difference vegetation index (NDVI), stress index (SI), photochemical reflectance index (PRI) and canopy reflectance at 693 nm, were highly correlated with tissue N concentration (TN), tissue moisture content (TM), TN×TM and canopy colour, as influenced by N and LF treatment combinations. Coefficients of determination ranged from 0.50 to 0.79 (P<0.001) based on single‐day correlations and correlations established over the entire growing period in 2002 and in 2003. TN was mainly predicted from wavelengths in the VIS portion of the spectrum, while TM was predicted from wavelengths in the VIS and NIR. Correlations were inconsistent between spectral parameters and physiological parameters throughout the study confirming the problem of temporal variation associated with spectral signatures of turfgrass species. However, spectral reflectance showed significant potential for monitoring turfgrass N and moisture status, and was able to capture temporal variability over the same growing period and from one year to another. The results provide a sound basis for future validation of ground‐based remote sensing for turfgrass management on golf courses.