Alan H. Strahler

Global land cover mapping from MODIS: algorithms and early results

Until recently, advanced very high-resolution radiometer (AVHRR) observations were the only viable source of data for global land cover mapping. While many useful insights have been gained from analyses based on AVHRR data, the availability of moderate resolution imaging spectroradiometer (MODIS) data with greatly improved spectral, spatial, geometric, and radiometric attributes provides significant new opportunities and challenges for remote sensing-based land cover mapping research. In this paper, we describe the algorithms and databases being used to produce the MODIS global land cover product. This product provides maps of global land cover at 1-km spatial resolution using several classification systems, principally that of the IGBP. To generate these maps, a supervised classification methodology is used that exploits a global database of training sites interpreted from high-resolution imagery in association with ancillary data. In addition to the IGBP class at each pixel, the MODIS land cover product provides several other parameters including estimates for the classification confidence associated with the IGBP label, a prediction for the most likely alternative class, and class labels for several other classification schemes that are used by the global modeling community. Initial results based on 5 months of MODIS data are encouraging. At global scales, the distribution of vegetation and land cover types is qualitatively realistic. At regional scales, comparisons among heritage AVHRR products, Landsat TM data, and results from MODIS show that the algorithm is performing well. As a longer time series of data is added to the processing stream and the representation of global land cover in the site database is refined, the quality of the MODIS land cover product will improve accordingly.

First operational BRDF, albedo nadir reflectance products from MODIS

With the launch of NASA’s Terra satellite and the MODerate Resolution Imaging Spectroradiometer (MODIS), operational Bidirectional Reflectance Distribution Function (BRDF) and albedo products are now being made available to the scientific community. The MODIS BRDF/Albedo algorithm makes use of a semiempirical kernel-driven bidirectional reflectance model and multidate, multispectral data to provide global 1-km gridded and tiled products of the land surface every 16 days. These products include directional hemispherical albedo (black-sky albedo), bihemispherical albedo (white-sky albedo), Nadir BRDF-Adjusted surface Reflectances (NBAR), model parameters describing the BRDF, and extensive quality assurance information. The algorithm has been consistently producing albedo and NBAR for the public since July 2000. Initial evaluations indicate a stable BRDF/Albedo Product, where, for example, the spatial and temporal progression of phenological characteristics is easily detected in the NBAR and albedo results. These early beta and provisional products auger well for the routine production of stable MODIS-derived BRDF parameters, nadir reflectances, and albedos for use by the global observation and modeling communities.

Monitoring vegetation phenology using MODIS

Abstract Accurate measurements of regional to global scale vegetation dynamics (phenology) are required to improve models and understanding of inter-annual variability in terrestrial ecosystem carbon exchange and climate–biosphere interactions. Since the mid-1980s, satellite data have been used to study these processes. In this paper, a new methodology to monitor global vegetation phenology from time series of satellite data is presented. The method uses series of piecewise logistic functions, which are fit to remotely sensed vegetation index (VI) data, to represent intra-annual vegetation dynamics. Using this approach, transition dates for vegetation activity within annual time series of VI data can be determined from satellite data. The method allows vegetation dynamics to be monitored at large scales in a fashion that it is ecologically meaningful and does not require pre-smoothing of data or the use of user-defined thresholds. Preliminary results based on an annual time series of Moderate Resolution Imaging Spectroradiometer (MODIS) data for the northeastern United States demonstrate that the method is able to monitor vegetation phenology with good success.

The Moderate Resolution Imaging Spectroradiometer (MODIS): land remote sensing for global change research

The first Moderate Resolution Imaging Spectroradiometer (MODIS) instrument is planned for launch by NASA in 1998. This instrument will provide a new and improved capability for terrestrial satellite remote sensing aimed at meeting the needs of global change research. The MODIS standard products will provide new and improved tools for moderate resolution land surface monitoring. These higher order data products have been designed to remove the burden of certain common types of data processing from the user community and meet the more general needs of global-to-regional monitoring, modeling, and assessment. The near-daily coverage of moderate resolution data from MODIS, coupled with the planned increase in high-resolution sampling from Landsat 7, will provide a powerful combination of observations. The full potential of MODIS will be realized once a stable and well-calibrated time-series of multispectral data has been established. In this paper the proposed MODIS standard products for land applications are described along with the current plans for data quality assessment and product validation.

The factor of scale in remote sensing

Abstract Thanks to such second- and third-generation sensor systems as Thematic Mapper, SPOT, and AVHRR, a user of digital satellite imagery for remote sensing of the earths surface now has a choice of image scales ranging from 10 m to 1 km. The choice of an appropriate scale, or spatial resolution, for a particular application depends on several factors. These include the information desired about the ground scene, the analysis methods to be used to extract the information, and the spatial structure of the scene itself. A graph showing how the local variance of a digital image for a scene changes as the resolution-cell size changes can help in selecting an appropriate image scale. Such graphs are obtained by imaging the scene at fine resolution and then collapsing the image to successively coarser resolutions while calculating a measure of local variance. The local variance/resolution graphs for the forested, agricultural, and urban/suburban environments examined in this paper reveal the spatial structure of each type of scene, which is a function of the sizes and spatial relationships of the objects the scene contains. At the spatial resolutions of SPOT and Thematic Mapper imagery, local image variance is relatively high for forested and urban/suburban environments, suggesting that information-extracting techniques utilizing texture, context, and mixture modeling are appropriate for these sensor systems. In agricultural environments, local variance is low, and the more traditional classifiers are appropriate.

An algorithm for the retrieval of albedo from space using semiempirical BRDF models

Spectral albedo may be derived from atmospherically corrected, cloud-cleared multiangular reflectance observations through the inversion of a bidirectional reflectance distribution function (BRDF) model and angular integration. This paper outlines an algorithm suitable for this task that makes use of kernel-based BRDF models. Intrinsic land surface albedos are derived, which may be used to derive actual albedo by taking into account the prevailing distribution of diffuse skylight. Spectral-to-broadband conversion is achieved using band-dependent weighting factors. The validation of a suitable BRDF model, the semiempirical Ross-Li (reciprocal RossThick-LiSparse) model and its performance under conditions of sparse angular sampling and noisy reflectances are discussed, showing that the retrievals obtained are generally reliable. The solar-zenith angle dependence of albedo may be parameterized by a simple polynomial that makes it unnecessary for the user to be familiar with the underlying BRDF model. The algorithm given is that used for the production of a BRDF/albedo standard data product from NASAs EOS-MODIS sensor, for which an at-launch status is provided. Finally, the algorithm is demonstrated on combined AVHRR and GOES observations acquired over New England, from which solar zenith angle-dependent albedo maps with a nominal spatial resolution of 1 km are derived in the visible band. The algorithm presented may be employed to derive albedo from space-based multiangular measurements and also serves as a guide for the use of the MODIS BRDF/albedo product.

Geometric-Optical Modeling of a Conifer Forest Canopy

A geometric-optical forest canopy model that treats conifers as cones casting shadows on a contrasting background can explain the major portion of the variance in a remotely sensed image of a forest stand. The model is driven by interpixel variance generated from three sources: 1) the number of crowns in the pixel; 2) the size of individual crowns; and 3) overlapping of crowns and shadows. The model uses parallel-ray geometry to describe the illumination of a three-dimensional cone and the shadow it casts. Cones are assumed to be randomly placed and may overlap freely. Cone size (height) is distributed lognormally, and cone form, described by the apex angle of the cone, is-fixed in the model but allowed to vary in its application. The model can also be inverted to provide estimates of the size, shape, and spacing of the conifers as cones using remote imagery and a minimum of ground measurements. Field tests using both 10-and 80-m multispectral imagery of two test conifer stands in northeastern California produced reasonable estimates for these parameters. The model appears to be sufficiently general and robust for application to other geometric shapes and mixtures of simple shapes. Thus it has wide potential use not only in remote sensing of vegetation, but also in other remote sensing situations in which discrete objects are imaged at resolutions sufficiently coarje that they canot be resolved individually.

On the nature of models in remote sensing

Abstract An explicit framework can provide a better understanding of remote sensing models and their interrelationships. This framework distinguishes between the scene, which is real and exists on the ground, and the image, which is a collection of spatially arranged measurements drawn from the scene. The scene model generalizes and parameterizes the essential qualities of the scene. Scene models may be discrete, in which the scene model consists of discrete elements with boundaries, or continuous, in which matter and energy flows are taken to be continuous and there are no clear or sharp boundaries in the scene. In the discrete case, there are two possibilities for models: H- and L- resolution . In the H- resolution case, the resolution cells of the image are smaller than the elements, and thus the elements may be individually resolved. In the L- resolution case, the resolution cells are larger than the elements and cannot be resolved. Most canopy models are L- resolution , deterministic, and noninvertible in nature; image processing models, however, tend to be H- resolution , empirical, and invertible. This taxonomy helps add insight to the development of remote sensing theory and point the way to new, productive areas of research.

Terrestrial remote sensing science and algorithms planned for EOS/MODIS

Abstract The Moderate Resolution Imaging Spectroradiometer (MODIS) will be the primary daily global monitoring sensor on the NASA Earth Observing System (EOS) satellites, scheduled for launch on the EOS-AM platform in June 1998 and the EOS-PM platform in December 2000. MODIS is a 36 channel radiometer covering 0·415-14·235 μm wavelengths, with spatial resolution from 250 m to 1 km at nadir. MODIS will be the primary EOS sensor for providing data on terrestrial biospheric dynamics and process activity. This paper presents the suite of global land products currently planned for EOSDIS implementation, to be developed by the authors of this paper, the MODIS land team (MODLAND). These include spectral albedo, land cover, spectral vegetation indices, snow and ice cover, surface temperature and fire, and a number of biophysical variables that will allow computation of global carbon cycles, hydrologic balances and biogeochemistry of critical greenhouse gases. Additionally, the regular global coverage of these var...

The use of prior probabilities in maximum likelihood classification of remotely sensed data

Abstract The expected distribution of classes in a final classification map can be used to improve classification accuracies. Prior information is incorporated through the use of prior probabilities—that is, probabilities of occurrence of classes which are based on separate, independent knowledge concerning the area to be classified. The use of prior probabilities in a classification system is sufficiently versatile to allow (1) prior weighting of output classes based on their anticipated sizes; (2) the merging of continuously varying measurements (multispectral signatures) with discrete collateral information datasets (e.g., rock type, soil type); and (3) the construction of time-sequential classification systems in which an earlier classification modifies the outcome of a later one. The prior probabilities are incorporated by modifying the maximum likelihood decision rule employed in a Bayesian-type classifier to calculate a posteriori probabilities of class membership which are based not only on the resemblance of a pixel to the class signature, but also on the weight of the class which is estimated for the final output classification. In the merging of discrete collateral information with continuous spectral values into a single classification, a set of prior probabilities (weights) is estimated for each value which the discrete collateral variable may assume (e.g., each rock type or soil type). When maximum likelihood calculations are performed, the prior probabilities appropriate to the particular pixel are used in classification. For time-sequential classification, the prior classification of a pixel indexes a set of appropriate conditional probabilities reflecting either the confidence of the investigator in the prior classification or the extent to which the prior class identified is likely to change during the time period of interest.