Search

Crop analysis aided by mathematical transformation that optimizes perspective of six-dimensional, six-band, spectral data taken from spacecraft or aircraft. Transformation applied to any temperatureclimate vegetated scene, providing direct view of regions of data concentration resulting from band correlations and fundamental reflectance properties of scene classes. Almost all of data variability captured in three spectral features, thus reducing by factor of 2 number of spectral features carried, incurring minimal loss of important information. Three-dimensional representation with two principal planes retains about 95 percent of six-dimensional spectral data used to distinguish among scenes containing green plants and bare soil with varying degrees of moisture.

The fundamental concepts on which the Tasseled Cap transformations of MSS and TM data are based - particularly the identification of inherent data structures - are explained and discussed. Emphasis on the structures present in data from any given sensor, which are themselves the expression of physical characteristics of scene classes, provides a number of advantages, including (a) reduction in data volume with minimal information loss; (b) spectral features which can be applied, without re-definition or adjustment, to any data set for a given sensor; (c) spectral features which can be directly associated with important physical parameters; and (d) easier integration of data from multiple sensors.

The launch of LANDSAT-4 in July 1982 represents a continuation in the remote sensing of earth resources. The 80-m spatial resolution provided by the Multispectral Scanner (MSS) on board the satellite is fine enough to resolve many natural features and land-use details in both rural and urban settings. The second sensor of the spacecraft, the Thematic Mapper (TM), introduce a new era of sensing with refined spatial resolution (30 m) and expanded spectral coverage (7 bands). This paper describes results from engineering studies of the characteristics of digital image data from the two LANDSAT-4 sensors are described. These studies form a part of the LANDSAT-4 Image Data Quality Analysis program (LIDQA). The image data were generally found to be of high quality and the TM provided several improvements over the MSS, in its spatial and spectral characteristics.

A transformation of TM-waveband reflectance-factor data which provides features related as directly as possible to the corresponding TM Tasseled Cap brightness, greennes, and wetness features is presented. The reflectance factor transformation is based on spectrometer data integrated over the prelaunch composite-detector response functions of the Landsat-4 Thematic Mapper. A description, in general terms, of the approach for adjusting the transformation matrix to other types of reflectance factor data (different instrument or band response) is given.

The effects of atmospheric haze on Thematic Mapper data, transformed to TM Tasseled Cap features, are illustrated by means of simulation. A spectral feature by which the amount of atmospheric haze may be inferred is derived and described for both simulated and actual TM data. Results presented for two actual TM scenes illustrate the diagnostic feature's sensitivity to changes in haze level as well as its insensitivity to scene-class-related variability. The method by which such a diagnostic feature might be incorporated into a haze normalization procedure is also discussed.

A simulated LANDSAT 4 TM and MSS data set, representing three crops over three growing seasons and a wide variety of soil types, was used to evaluate the structure of TM data and to compare its characteristics to those of MSS data. TM bands 2, 3, and 4, transformed to tasseled cap-like coordinates, provide an equivalent data space to MSS tasseled cap data, with greater dynamic range and no apparent loss of information resulting from the exclusion of the 0.9 to 1.1 micron region. Data from the six reflective TM bands (excluding the thermal band) primarily occupy two planes and a transition zone between them. The plane of vegetation is comparable to the MSS tasseled cap plane, while the plane of soils and transition zone provide a new dimension of information unavailable from the MSS. This added dimension offers promise of improved ability to determine the relative mix of vegetation and soil in the sensor field of view and to estimate soil moisture status. The improvement in spectral characteristics of the TM over the MSS, not to mention the greater spatial resolution, have resulted in a significant increase in the information content of the data.

An extension of the TM tasseled cap transformation to reflectance factor data is presented, and the basic concepts underlying the tasseled cap transformations are described. The ratio of TM bands 5 and 7, and TM tasseled cap wetness, are both shown to offer promise of direct detection of available soil moisture. Some effects of organic matter and other soil characteristics or constituents on TM tasseled cap spectral response are also considered.

Thematic Mapper data, simulated from field and laboratory spectrometer measurements of a variety of agricultural crops and a wide range of soils, are analyzed to determine their dispersion in the six space defined by the reflective TM bands (i.e., excluding the thermal band). While similar analyses of MSS data from agricultural scenes show that the vast majority of the MSS data occupy a single plane, the simulated TM data primarily occupy three dimensions, defining two intersecting planes and a zone of transition between the two. Viewing the plane of Vegetation head on provides a projection comparable to the single plane of MSS data. The Plane of Soils and transition zone represent new information made available largely as a result of the longer infrared bands included in the Thematic Mapper. A transformation, named the Thematic Mapper Tasseled Cap, is presented which rotates the TM data such that the described data structure is most readily accessible to view.

In an extension of previous simulation studies, a transformation of actual TM data in the six reflective bands is described which achieves three objectives: a fundamental view of TM data structures is presented, the vast majority of data variability is concentrated in a few (three) features, and the defined features can be directly associated with physical scene characteristics. The underlying TM data structure, based on three TM scenes as well as simulated data, is described, as are the general spectral characteristics of agricultural crops and other scene classes in the transformed data space.

The launch of Landsat-4 in July 1982 represents a continuation in the remote sensing of earth resources. The 80-m spatial resolution provided by the Multispectral Scanner (MSS) on board the satellite is fine enough to resolve many natural features and land-use details in both rural and urban settings. The second sensor of the spacecraft, the Thematic Mapper (TM), introduces a new era of sensing with refined spatial resolution (30 m) and expanded spectral coverage (7 bands). This paper describes results from engineering studies of the characteristics of digital image data from the two Landsat-4 sensors. These studies form a part of the Landsat-4 Image Data Quality Analysis program (LIDQA). The image data were generally found to be of high quality and the TM provided several improvements over the MSS, in its spatial and spectral characteristics.

Thematic Mapper signal counts in the six reflective bands (i.e., excluding the thermal band) are simulated using field and laboratory spectrometer measurements of a variety of crops, crop conditions, and soil types. The Dave atmospheric model and prelaunch sensor characteristics comprise the other components of the simulation. The simulated data are found to occupy essentially three dimensions, two of which are equivalent to the MSS Tasseled Cap Greennes and Brightness features, and a third which is substantially influenced by the mid-infrared bands of the TM. This new dimension is primarily related to soil characteristics, including soil moisture. The nature and characteristics of each dimension are discussed, as are some of the expected information gains (over MSS data) resulting from the additional dimensionality of the data.

The typical patterns of spectral development (profiles) for corn and soybeans are presented, based on field-collected reflectance data transformed to correspond to Landsat-MSS Tasseled Cap coordinates. Reasonable variations in field conditions and cultural practices are shown to significantly influence profile features. The separability of the two crops is determined to be primarily related to the maximum value of the reflectance equivalent of Greenness, and to the plateau effect seen in corn Greenness profiles. The impact of changes in conditions on separability is described. In addition, association is made between profile features and stages of development for corn and soybeans. Corn is shown to peak at a stage well before tasseling or maximum LAI, while the characteristics of the soybean profile are shown to be unrelated to any particular stage of development.

The spectral characteristics of the Landsat-4 MSS and TM are compared using signal counts simulated from field-measured reflectance spectra. The Dave atmospheric model and prelaunch sensor characteristics are also used in the simulation. Comparison is made based on individual bands and Tasseled Cap transformed data. TM bands 2, 3, and 4 are shown to contain virtually all the information contained in the four MSS bands. Tasseled Cap transformed data from the six reflective bands of the TM are shown to contain at least one and possibly two additional dimensions of information beyond the two dimensions in MSS data, while still providing a direct association with the MSS Tasseled Cap features.

Landsat-4 TM and MSS data, simulated from field reflectance spectra, are used to determine the dimensionality and structure of TM data (excluding the thermal band), demonstrate the relationships between the two sensors, and derive a Tasseled Cap Transformation for TM data. The simulated TM data are found to primarily occupy three dimensions, and to be concentrated into two adjoining and orthogonal planes and a transition zone between the two. The 'Plane of Vegetation' is shown to be comparable to the MSS Tasseled Cap plane, while the 'Plane of Soils' represents new information. The potential for improved spectral estimation of the relative mix of vegegation and soil in the field of view, and for improved monitoring of soil moisture status, is demonstrated.

The TM Tasseled Cap transformation, which provides both a 50% reduction in data volume with little or no loss of important information and spectral features with direct physical association, is presented and discussed. Using both simulated and actual TM data, some important characteristics of vegetation and soils in this feature space are described, as are the effects of solar elevation angle and atmospheric haze. A preliminary spectral haze diagnostic feature, based on only simulated data, is also examined. The characteristics of the TM thermal band are discussed, as is a demonstration of the use of TM data in energy balance studies. Some characteristics of AVHRR data are described, as are the sensitivities to scene content of several LANDSAT-MSS preprocessing techniques.

Previous characterization of scan-related low-frequency noise was extended and refined through detailed analysis of shutter calibration data on CCT-ADDS tapes and reflective-band data from nighttime acquisitions. A recommended correction procedure was identified that uses calibration shutter data both as a diagnostic and to obtain correction values. Through comparison of coincident TM and MSS data, illustrations of the added information content of TM data for agricultural applications were developed. The capability of improved spatial resolution to better define boundaries and to resolve spatial details is shown. Spectral analysis of tasseled-cap transformations of TM and MSS data shows high correlation between greenness features, greater signal range for TM, and indications that a subset of TM bands could accurately simulate MSS data, if required.

A transformation is described which rotates Thematic Mapper data (excluding the thermal band) in a manner analogous to that used in the MSS Tasseled Cap Transformation, thus providing a direct view of the planes of data dispersion and a direct association of spectral features with physical scene characteristics. This TM Tasseled Cap Transformation includes MSS-equivalent Grenness and Brightness features, as well as at least one additional important feature. The new feature, tentatively termed 'Wetness', offers promise of enhanced ability to assess soil conditions, monitor vegetative development, and delineate cover classes. Relationships between scene characteristics and spectral variation in the transformed data space are discussed based on both simulated and actual TM data.

Research supporting the use of remote sensing for inventory and assessment of agricultural commodities is summarized. Three task areas are described: (1) corn and soybean crop spectral/temporal signature characterization; (2) efficient area estimation technology development; and (3) advanced satellite and sensor system definition. Studies include an assessment of alternative green measures from MSS variables; the evaluation of alternative methods for identifying, labeling or classification targets in an automobile procedural context; a comparison of MSS, the advanced very high resolution radiometer and the coastal zone color scanner, as well as a critical assessment of thematic mapper dimensionally and spectral structure.

The typical patterns of spectral development (profiles) for corn and soybeans are presented, based on field-collected reflectance data transformed to correspond to LANDSAT-MSS Tasseled Cap coordinates. Reasonable variations in field conditions and cultural practices are shown to significantly influence profile features. The separability of the two crops is determined to be primarily related to the maximum value of the reflectance equivalent of Greenness, and to the plateau effect seen in corn Greenness profiles. The impact of changes in conditions on separability is described. In addition, association is made between profile features and stages of development for corn and soybeans. Corn is shown to peak at a stage well before tasseling or maximum LAI, while the characteristics of the soybean profile are shown to be unrelated to any particular stage of development.

Research activities in support of AgRISTARS Inventory Technology Development Project in the use of aerospace remote sensing for agricultural inventory described include: (1) corn and soybean crop spectral temporal signature characterization; (2) efficient area estimation techniques development; and (3) advanced satellite and sensor system definition. Studies include a statistical evaluation of the impact of cultural and environmental factors on crop spectral profiles, the development and evaluation of an automatic crop area estimation procedure, and the joint use of SEASAT-SAR and LANDSAT MSS for crop inventory.

Association is made between the development stages of corn as defined by Hanway and the temporal-spectral development pattern of corn in a transformed data space derived from Landsat-MSS band reflectance values, using field-collected reflectance and associated data. Results indicate that the spectral vegetation index used (a reflectance equivalent of Tasseled Cap Greenness) reaches a maximum well before the stage at which corn is expected to achieve its peak leaf area index. Possible physiological and canopy geometry related causes for this and other results are presented.

A technique for evaluating crop temporal-spectral development patterns is described and applied to the analysis of cropping practices and environmental conditions as they affect reflectance characteristics of corn and soybean canopies. Typical variations in field conditions are shown to exert significant influences on the spectral development patterns, and thereby to affect the separability of the two crops.

An overall approach to crop spectral understanding is presented which serves to maintain a strong link between actual plant responses and characteristics and spectral observations from ground based and spaceborne sensors. A specific technique for evaluating field reflectance data, as a part of the overall approach, is also described. Results of the application of this technique to corn and soybeans reflectance data collected by and at Purdue/LARS indicate that a number of common cultural and environmental factors can significantly affect the temporal spectral development patterns of these crops in tasseled cap greenness (a transformed variable of LANDSAT MSS signals).

A labeling technique is described which seeks to associate a sampling entity with a particular crop or crop group based on similarity of growing season and temporal-spectral patterns of development. Human analyst provide contextual information, after which labeling decisions are made automatically. Results of a test of the technique on a large, multi-year data set are reported. Grain labeling accuracies are similar to those achieved by human analysis techniques, while non-grain accuracies are lower. Recommendations for improvments and implications of the test results are discussed.

Classification and technology development for area estimation of corn, soybeans, wheat, barley, and sunflowers are outlined. Supporting research for corn and soybean foreign commodity production forecasting is highlighted. Graphs profiling the greenness and brightness of the crops are presented.

There are no author-identified significant results in this report.

This paper presents a method for describing the overall continuous pattern of crop spectral development from a set of discrete Landsat observations, using mathematical representations termed profiles. The agronomic basis for the approach is described along with model forms and techniques for estimation of the model parameters. In addition, several current or potential applications of the technology are described.

There are no author-identified significant results in this report.

This paper presents relationships among Landsat MSS bands and selected transformations of them, with emphasis on the Tasseled-Cap Transformation and its Brightness, Greenness variables. It also discusses relationships between reflectance measurements made in the Landsat spectral bands and actual Landsat data. Agronomically oriented analyses of reflectance measurements of wheat throughout a growing season are presented, with a comparison of various green measures, correlation with crop development stage, and examination of the effects of moisture stress. The final example addresses the use of transformed variables in a newly developed approach to forestry change detection.

There are no author-identified significant results in this report.

There are no author-identified significant results in this report.

The author has identified the following significant results. An outgrowth of research and development activities in support of LACIE was a multicrop area estimation procedure, Procedure M. This procedure was a flexible, modular system that could be operated within the LACIE framework. Its distinctive features were refined preprocessing (including spatially varying correction for atmospheric haze), definition of field like spatial features for labeling, spectral stratification, and unbiased selection of samples to label and crop area estimation without conventional maximum likelihood classification.

The Suits bidirectional reflectance model (1972) for vegetation canopies was used to calculate reflectance values for the four Landsat spectral bands under a variety of canopy and observation conditions. Relationships between characteristics of developing wheat fields and associated reflectances were studied. Coincident measurements of crop reflectance and Landsat signals were compared using regression analysis techniques. The model was also applied to forest situations both to help establish the extent to which Landsat may provide direct information about conditions occurring beneath forest overstories and to help develop methods for reducing the effects of terrain slope and aspect in the processing of Landsat data. Finally, a relationship based on model calculations was used to adjust Landsat signal values and was found to reduce topography-related variability and improve subsequent classification performance.

The author has identified the following significant results. Effects of terrain topography in mountainous forested regions on LANDSAT signals and classifier training were found to be significant. The aspect of sloping terrain relative to the sun's azimuth was the major cause of variability. A relative insolation factor could be defined which, in a single variable, represents the joint effects of slope and aspect and solar geometry on irradiance. Forest canopy reflectances were bound, both through simulation, and empirically, to have nondiffuse reflectance characteristics. Training procedures could be improved by stratifying in the space of ancillary variables and training in each stratum. Application of the Tasselled-Cap transformation for LANDSAT data acquired over forested terrain could provide a viable technique for data compression and convenient physical interpretations.