MODIS MOD12 Land Cover and Land Cover Dynamics Products User Guide
Introduction
Land cover and the human or natural alteration of land cover
play a major role in global scale patterns of climate and the
biogeochemistry cucle of the earth system. Although the oceans are the
major driving force for the Earth's physical climatology, the
land surface has considerable control on the planet's
biogeochemical cycles, which in turn significantly influence the
climate system through the radiative properties of greenhouse
gases and reactive species. Further, variations in topography,
albedo, vegetation cover, and other physical characteristics of
the land surface generate variations of weather and climate by
forcing atmospheric circulation patterns that are driven by
surface-atmosphere matter and energy fluxes and the momentum of
the Earth's rotation.
In this context, an important application of accurate global
land cover information is the inference of parameters that
influence biophysical processes and energy exchanges between the
atmosphere and the land surface as required by regional and
global scale climate and ecosystem process models (Townshend, et
al., 1991). Examples of such parameters for climate modeling
include leaf area index (LAI), roughness length, surface
resistance to evapotranspiration, canopy greenness fraction,
vegetation density, root distribution, and fraction of
photosynthetically-active radiation absorbed (fPAR) (Sellers,
1991a, 1991b). These serve as input variables that control
surface energy and mass balances. Examples of ecosystem process
model parameters for which land cover type may serve as a
surrogate include leaf photosynthetic capacity, canopy
conductance, type of photosynthetic system, and maximum
photosynthetic rate (Running and Coughlan, 1988).
Most of these inferences are based on the structural character
of the vegetation cover, which is sensible by remote sensing instruments. The
objective of the Land Cover Product is to identify a suite of
land cover types amenable to such parameterization by exploiting
the information content of MODIS data in the spectral, temporal, spatial,
directional, and quantization domains. The objective of the Land Cover
Dynamics Product is to detect and quantify the intra-annual phenological
dynamics of global vegetation cover and persistent interannuaul changes in
land cover type over time. The natural and anthropogenic processes that
bring these changes about may be detected by global and regional
models through changes in annual and inter-annual cycles.
The MODerate resolution Imaging
Spectroradiometer (MODIS) Land
Cover Science Data Product (MOD12Q1)
provides a suite of Land Cover classifications with the primary
classification in the International Geosphere-Biosphere Programme (IGBP)
scheme (Running, et al., 1994, Belward, et al., 1995). Each of these
classification schemes is accompanied by assessments of its quality or
confidence. Additional information about the IGBP classification (the
second most probable class label and percent) are also provided as are
overall quality flags and an embedded land/water mask.
The MODIS Land Cover Dynamics Product
(MOD12Q2) will provide information
on land cover change vectors and on global vegetation phenological attributes.
This product is not yet operational.
NOTE: The most recent MOD12Q1 product delivery, nominal data day 2000289, version 3, is the first result using a full annual time series, with the exception of June 2001, missing due to instrument down time. All IGBP classes are included in this delivery except Urban and Built-Up, Class 13 in SDS Land_Cover_Type_1, which is an
overlay of a rasterized version of the Digital Chart of the World from the
EDC GLCC team. The first two MOD12Q1 product deliveries, nominal data day 2000289, version 1, and 2001001, version 1, were based on less
than one year of data (July 2000 to December 2000) and the product algorithm
presumes a complete global annual cycle as input for the classification.
Furthermore, input data prior to 31 October 2000 had not been corrected for
aerosols at the time these products were generated. We anticipate improvements in accuracy and confidence as reprocessed
data from 2000 and 2001 is available to complete an annual cycle. Despite
these drawbacks, the performance of the beta Land Cover classification
algorithm was very encouraging. For nominal data day
2000289, version 1, Permanent Wetlands, Class 11 in SDS Land_Cover_Type_1, were not
included, and Urban and Built-Up, Class 13 in SDS Land_Cover_Type_1, is an
overlay of a rasterized version of the Digital Chart of the World from the
EDC GLCC team.
Every 96 days, the operational MODIS Land Cover algorithm makes use of the
last 12 months of Nadir BRDF Adjusted surface Reflectance data
(NBAR - MOD43B4), Enhanced Vegetation Index data (EVI - MOD13A2), Land Surface
Temperature data (MOD11A2), BRDF information (MOD43B1), and surface texture
information (MODAGTEX), in conjunction with a global set of training data
(Muchoney, et al., 1999) to provide global classifications of the land
surface. These one kilometer gridded classifications are in several
classification schemes in addition to IGBP, inclunding the UMd modification of
the IGBP scheme, the MODIS LAI/fPAR (MOD15) scheme, and the MODIS NPP (MOD17)
scheme.
When insufficient input information is available, the preexisting
classification label for an affected pixel will be carried forward for the
IGBP and UMd data layers. Thus even though new data may be unavailable due to
persistent cloud cover or other data loss, each pixel will be assigned a
consistent land cover classification.
Note: Beta and provisional runs of the product will carry forward values
from the at-launch product produced by the Laboratory for Global Remote Sensing
Studies, housed within the Geography Department at the University of
Maryland, College Park. The at-launch product primary layer, in the IGBP
scheme, is the USGS EDC DAAC Global Land Cover Characteristics Version 1
database, available at
http://edcdaac.usgs.gov/glcc/glcc.html.
The MODIS Land Cover and Land Dynamics Science Data Products will
be provided in an Integerized Sinusoidal Grid (ISG) projection
with standard tiles representing 1200 by 1200 pixels on the earth.
While the projection becomes increasingly sheared with
distance from the Equator and the Greenwich meridian, the equal area properties
of ISG mean that it is a good data storage format and it is
possible to convert each tile to other, more common projections
through the use of any one of a number of commercial or public
software packages. These Level-3 one-kilometer MODIS Land
(MODLAND) products are being released in Hierarchical Data Format - Earth
Observing System (HDF-EOS)
for each of the 328 land tiles on the globe (see HDF-EOS
FAQ).
Each operational product is associated with extensive quality
assessment information so that users can evaluate the quality
of the land cover classification or the change vector and
phenology parameters.
At a minimum, all MODIS Land products supply a per-pixel
quality flag indicating whether the algorithm produced results or
not for that pixel and if so, whether the result is of the
highest quality and can be used without reservation or whether
(due to some uncertainties in the processing) the user should
check the extensive additional product-specific quality assurance
to make sure the output is appropriate for their application. In
addition, each tile of data is accompanied by extensive metadata
that provide similar quality assessments of the entire tile.
Note that the per-pixel data and the quality information are
computed for all land and coastal areas and for shallow water
regions (pixels that are within 5 km of land OR are less than
50 meters deep). The products and quality flags are not computed
for moderate or deep water regions (pixels greater than 5km from
land and with water depths greater than 50m). The EOS land-water
mask (which is static at a 1km resolution for Level 1B products)
is passed along through the production chain with the
reprojection and aggregation of the reflectance data to Level 3
and is stored for the user's convenience as bit flags in the
per-pixel quality information associated with each MOD12Q1
product.
The Level 2G Surface Reflectance Product (MOD09) for MODIS
provides daily, cloud-cleared, atmospherically-corrected surface
reflectances. The data from channels 1-7 are aggregated together
in to a one kilometer resolution and stored in Level 3 ISG tiles.
This aggregation and binning occurs on a daily basis and results
in the interim Level 3 product MODAGAGG. In the higher
latitudes, as many as four of these aggregated observations will
be retained at each pixel for each day. The data from sixteen
days worth of MODAGAGG are then used as the primary input for the
MOD43B BRDF/Albedo Product(MOD43B) The algorithm fits a BRDF model to
these directional surface reflectances and the parameters of the
model (RossThick-LiSparseReciprocal) are used to
compute the Nadir BRDF Adjusted Reflectance (NBAR)
values provided in MOD43B4 (Lucht et al., 2000). NBARs
constitute the primary input values to the Land Cover Classifiers
and to the Change Vector Products. The MODAGAGG is also
used to generate the 16-day EVI product (MOD13A2) which is another
key input to the classification algorithm and surface texture
information (MODAGTEX, based on the 250meter channel 1 surface reflectance
values). Eight-day Land Surface
Temperature (MOD11A2) is also considered as are additional
BRDF information (MOD43B1). All of these input data are
assembled on a 32 day basis (MOD12M1). Once three of these
global 32-day data assemblages are complete, they (along with the
previous nine 32 day assemblages) are submitted to a
training algorithm. This extracts training data from specific
locations and uses it to train the classification algorithm.
Once the classification is training, the resulting classification
tree is applied to the global data set and global land cover
maps and confidence assessments are produced. Finally
the data are compared with the global classification from
the previous 96-day period and any missing pixels are
assigned classification values by persisting the earlier
product.
References
Belward, A. S., and Loveland, T., 1995, The IGBP-DIS 1km Land Cover
Project, Remote Sensing in Action, Proceedings of the 21st Annual Conference
of the Remote Sensing Society, held in Southampton, UK 11-14 Sept., 1995
(RSS:Nottingham)pp 1099-1106
Lucht, W., Schaaf, C. B., and Strahler, A. H., 2000, An algorithm for the
retrieval of albedo from space using semiempirical BRDF models, IEEE
Trans. Geosci. Remote Sens., vol. 38, no. 2, pp. 977-998.
Muchoney, D., Strahler, A., Hodges, J., and LoCastro, J., 1999, The IGBP
DISCover Confidence Sites and the System for Terrestrial Ecosystem
Parameterization: Tools for Validating Global Land Cover Data.
Photogrammetric Engineering & Remote Sensing, v65 n9, 1999.
Running, S. W., Loveland, T. R., and Pierce, L. L., 1994, A vegetation
classification logic based on remote sensing for use in global scale
biogeochemical models, Ambio, 23, 77-81
Running, S. and J. C. Coughlan, 1988. A general model of forest ecosystem
processes for regional applications: I. Hydrologic balance, canopy gas
exchange and primary production processes. Ecological Modelling 42:125 154.
Sellers, P., 1991a. Canopy reflectance, photosynthesis and transpiration.
International Journal of Remote Sensing 6:1335 1372.
Sellers, P., 1991b. Modeling and observing land-surface-atmosphere
interactions on large scales. Surveys of Geophysics 12:85 114.
Townshend, J. R. G., C. Justice, W. Li, C. Gurney, and J. McManus, 1991.
Global land cover classification by remote sensing: Present capabilities and
future possibilities. Remote Sensing of Environment 35:243 255.
|
