AVHRR Global Land Cover Classification
Over the past several years, researchers have increasingly turned to remotely sensed data to improve the accuracy of
data sets that describe the geographic distribution of land cover at regional and global scales. To develop improved
methodologies for global land cover classifications as well as to provide global land cover products for immediate use
in global change research, we have employed the NASA/NOAA Pathfinder Land (PAL) data set with a spatial resolution of
1km. This data set has a length of record of 14 years (1981-1994), providing the ability to test the stability of
classification algorithms. Furthermore, this data set includes red, infrared, and thermal bands in addition to the
Normalized Difference Vegetation Index (NDVI). Inclusion of these additional bands improves discrimination between cover
types. We aim through this study to 1) develop methodologies for global land cover classifications that are objective,
reproducible, and feasible to implement on data from additional years and 2) produce a global land cover classification
at 1 km spatial resolution accessible to the global change research community.
Reliable, geographically-referenced data on global vegetative cover is an important requirement for global models of
earth system. Satellite data provide the only truly synoptic view of the earth, and may potentially increase the
quality, internal consistency, and reproducibility of global land cover information.
This project initially aimed to develop a coarse resolution, global land cover data set from satellite data for use
in climate models. To this end, AVHRR data were resampled to a spatial resolution of one by one degree and used to carry
out a conventional, supervised classification of global land cover. Classifications have also proceeded at a finer
spatial resolution of 8km at a continental scale. In addition to describing vegetative cover according to topological
schemes, the project has explored methodologies to represent vegetative cover more realistically as gradients and
mosaics of cover types.
Most recently we have worked with colleagues in the Geography Department at the University of Maryland to develop
land cover characterizations for net primary productivity models. Supervised classifications at finer spatial
resolutions are underway, drawing particularly on the Pathfinder 1-km and 8-km data sets. The current project aim is to
develop and validate global land cover data sets and to develop advanced methodologies for more realistically describing
the vegetative land surface based on satellite data.
Procedure Training Data
To identify the pixels to be used for training of the 1 km AVHRR Pathfinder data, we collected a total of over 200
high resolution scenes of which we were confident of which cover type occurs. Most of the scenes used were acquired by
the Landsat Multispectral Scanner System (MSS), and a few by Landsat Thematic Mapper and the LISS (Linear Imaging
Of the initial 200 scenes, we considered 156 suitable for interpretation. Scenes were considered unsuitable if haze
or poor quality data obscured the scene or if the cover types in the scene could not be visually distinguished. For most
scenes, we aimed to identify only one cover type within the scene. It was possible, however, to identify more than one
cover type in some scenes if croplands were visually identifiable based on the spatial patterns of fields or if
vegetation maps showed the presence of clearly identifiable cover types.
These training data provide the basis for carrying out a global land cover classification. They also provide data for
validating other land cover classification products. The methodology and Landsat images used for deriving these training
data for classification of AVHRR data at 8km resolution can also be applied to 1km AVHRR data and, in the future, MODIS
data at 250m and 500m resolution.
Deriving the Classification Product
The second part of the study involved deriving a global land cover classification product. The product was derived by
testing several metrics that describe the temporal dynamics of vegetation over an annual cycle. These metrics have the
potential to be used as input variables to a global land cover classification. The tested metrics are based on 1) the
ratio between surface temperature and NDVI, 2) seasonal metrics derived from the NDVI temporal profile such as length of
growing season, 3) a rule-based approach that determines cover type through a series of hierarchical trees based on
surface temperature and NDVI values, and 4) annual mean, maximum, minimum, and amplitude values for all optical and
thermal channels in the AVHRR Pathfinder Land (PAL) data. These metrics were applied to 1984 PAL data at 8km resolution
to derive a global land cover classification product using a decision tree classifier. The method described in this
website can be applied to AVHRR data from other years and to data at higher spatial resolutions.