MODIS MOD12 Land Cover and Land Cover Dynamics Products User Guide
MOD12Q1 Land Cover Product
The MOD12Q1 Land Cover Product (MODIS/Terra Land Cover 96 Day L3 Global
1 km ISIN Grid) supplies an IGBP land cover classification
map (Belward et al., 1999; Scepan, 1999) of the globe
along with an assessment of the quality or confidence that is placed in
that classification. For ease of use by the community, a number of other
classification schemes (and their assessments) are also provided. They
include the 14 class UMD scheme (Hansen et al., 2000), the BGC Biome scheme
(Running et al., 1994) and the LAI/fPAR Biome scheme (Myneni et al., 1997).
The classification was produced using a supervised approach. Training
sites where developed by analyzing high resolution (e.g., Landsat TM) imagery
in conjunction with ancillary data (Muchoney et al., 1999). The classification was produced
using a decision tree classification algorithm (C4.5 [Quinlan
1993]) in conjunction with a technique for improving classification accuracies
that has received considerable attention in the machine learning and statistics
communities known as boosting [Freund 1995]. Boosting improves classification
accuracies by iteratively estimating classifiers using a base learning
algorithm (e.g., a decision tree) while systematically varying the training
sample. At each iteration, the training sample is modified
to focus the classification algorithm on examples that are difficult
to classify correctly. This modification is performed by providing
a weight for each training example. The importance of misclassified
training examples is increased and the classification algorithm focuses
on learning these examples. The boosted classifier's prediction is
then based upon an accuracy-weighted vote across the estimated classifiers.
A number of different boosting algorithms have been developed.
The implementation used here is Adaboost.M1 [Freund and Schapire
97], which is the simplest multi-class boosting method.
Recently, boosting has been shown to be a form of additive logistic
regression [Friedman et al. 2000]. As a result, probabilities of
class membership can be obtained from boosting. These probabilities
provide a means of assessing the confidence of the classification results
as well as a means of incorporating ancillary information in the
form of prior probabilities to improved discrimination of cover types
that are difficult to separate in the remote sensing feature space.
In addition to the classifications and assessments, the MOD12Q1
Land Cover Product also provides mandatory quality information, information
on whether each pixel has been newly classified or is depending on a persisted
value, and an embedded land/water mask as bit flags in the 8 bit
Land_Cover_Type_QC
parameter.
IMPORTANT: See the Production
Notes page for information about each run by nominal data day and version. The
nominal data day appears in the file name of each distributed released product
as a string of the form YYYYDDD, where YYYY is the year, i.e.: 2001, and DDD
is the day of the year, i.e.: 001 for January 1st. The version number appears in the file name of each distributed released product
as a string of the form Vvvv, a leading "V" followed by three numerical digits.
References Cited
-
Belward, A. S., J. E. Estes, and K. D. Kline, The IGBP-DIS Global
1-km Land-Cover Data Set DISCover: A Project Overview,
Photogram. Eng. Remote Sens., 65, 1013-1020, 1999.
-
Freund, Y., "Boosting a weak learning algorithm by majority," Information
and Computation, 121(2), 256-285, 1995.
-
Freund, Y. and R. E. Schapire, "A decision-theoretic generalization of
on-line learning and an application to boosting," Journal of Computer and
System Sciences, 5(1), 119-139, 1997.
-
Friedman, J., T. Hastie, and R. Tibshirani, "Additive logistic regression:
A statistical view of boosting," The Annals of Statistics, 28(2), 337-374,
2000.
- Hansen, M. C., R. S. DeFries and J. R. G. Townshend and R.
Sohlberg, Global land cover classification at the 1km spatial
resolution using a classification tree approach, Int. J. Remote
Sens., 21, 1331-1364, 2000.
- 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.
- Myneni, R.B., R.R. Nemani, and S.W. Running, Estimation of global leaf area index and absorbed PAR using radiative transfer model, IEEE Trans. Geosci. Remote Sens., 35, 1380-1393, 1997.
-
Quinlan, J. R., C4.5: Programs for Machine Learning, Morgan Kauffman, 1993.
- 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
-
Scepan, J., Thematic Validataion of High-Resolution Global
Land-Cover Data Sets, Photogram. Eng. Remote Sens., 65,
1051-1060, 1999.
Science Data Sets
While the full MOD12Q1 specification
should be consulted for the most current description, the product includes
sixteen Science Data Sets (SDS) for each pixel in the tile. The long_names
of these SDSes are:
Land_Cover_Type_1 (IGBP)
Land_Cover_Type_2 (UMD)
Land_Cover_Type_3 (LAI/fPAR Biomes)
Land_Cover_Type_4 (BGC Biomes)
Land_Cover_Type_5 (TBD)
Land_Cover_Type_1_Assessment (IGBP)
Land_Cover_Type_2_Assessment (UMD)
Land_Cover_Type_3_Assessment (LAI/fPAR Biomes)
Land_Cover_Type_4_Assessment (BGC Biomes)
Land_Cover_Type_5_Assessment (TBD)
Land_Cover_Type_QC
Land_Cover_Type_1_Secondary
Land_Cover_Type_1_Secondary_Percent
LC_Property_1
LC_Property_2
LC_Property_3
The Land_Cover_Type_1 SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_1 ("YDim", "XDim")
Description: land cover types (IGBP)
water 0
evergreen needleleaf forest 1
evergreen broadleaf forest 2
deciduous needleleaf forest 3
deciduous broadleaf forest 4
mixed forests 5
closed shrubland 6
open shrublands 7
woody savannas 8
savannas 9
grasslands 10
permanent wetlands 11
croplands 12
urban and built-up 13
cropland/natural vegetation mosaic 14
snow and ice 15
barren or sparsely vegetated 16
unclassified 254
The Land_Cover_Type_2 SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_2 ("YDim", "XDim")
Description: land cover types (UMD)
water 0
evergreen needleleaf forest 1
evergreen broadleaf forest 2
deciduous needleleaf forest 3
deciduous broadleaf forest 4
mixed forests 5
closed shrubland 6
open shrublands 7
woody savannas 8
savannas 9
grasslands 10
croplands 12
urban and built-up 13
barren or sparsely vegetated 16
unclassified 254
The Land_Cover_Type_3 SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_3 ("YDim","XDim")
Description: land cover types (LAI/FPAR Biomes)
water 0
Grasses/Cereal Crops 1
Shrubs 2
Broadleaf crops 3
Savannah 4
Broadleaf forest 5
Needleleaf forest 6
Unvegetated 7
Urban 8
unclassified 254
The Land_Cover_Type_4 SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_4 ("YDim","XDim")
Description: land cover types (BGC Biomes)
water 0
Evergreen Needleleaf Vegetation 1
Evergreen Broadleaf Vegetation 2
Deciduous Needleleaf Vegetation 3
Deciduous Broadleaf Vegetation 4
Annual Broadleaf Vegetation 5
Annual Grass Vegetation 6
Non-vegetated Land 7
Urban 8
unclassified 254
The Land_Cover_Type_5 SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_5 ("YDim","XDim")
Description: land cover types (TBD)
The Land_Cover_Type_1_Assessment SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_1_Assessment ("YDim", "XDim")
Description: Land Cover Assessment (Confidences).
The Land_Cover_Type_2_Assessment SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_2_Assessment ("YDim", "XDim")
Description: Land Cover Assessment (Confidences).
The Land_Cover_Type_3_Assessment SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_3_Assessment ("YDim", "XDim")
Description: Land Cover Assessment (Confidences).
The Land_Cover_Type_4_Assessment SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_4_Assessment ("YDim", "XDim")
Description: Land Cover Assessment (Confidences).
The Land_Cover_Type_5_Assessment SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_5_Assessment ("YDim", "XDim")
Description: Land Cover Assessment (Confidences).
The Land_Cover_Type_QC SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_QC ("YDim", "XDim")
Description: Quality flags for Land Cover Type.
00-01 Mandatory QA
0 = processed, good quality
1 = processed, see other QA
2 = not processed due to cloud effects
3 = not processed due to other effects
02-03 Quarters since updated
0 = 1 quarter
1 = 2 quarters
2 = 3 quarters
3 = 4 quarters
04-07 land/water
0 = Shallow ocean
1 = Land (Nothing else but land)
2 = Ocean coastlines and lake shores
3 = Shallow inland water
4 = Ephemeral water
5 = Deep inland water
6 = Moderate or continental ocean
7 = Deep ocean
The Land_Cover_Type_1_Secondary SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_1_Secondary ("YDim", "XDim")
Description: Secondary class.
The Land_Cover_Type_1_Secondary_Percent SDS is specified as:
Data Field Name: UINT8 Land_Cover_Type_1_Secondary_Percent ("YDim", "XDim")
Description: Percent of secondary class.
The LC_Property_1 SDS is specified as:
Data Field Name: UINT8 LC_Property_1 ("YDim", "XDim")
Description: (TBD).
The LC_Property_2 SDS is specified as:
Data Field Name: UINT8 LC_Property_2 ("YDim", "XDim")
Description: (TBD).
The LC_Property_3 SDS is specified as:
Data Field Name: UINT8 LC_Property_3 ("YDim", "XDim")
Description: (TBD).
Local Attributes
In addition to the actual SDS data values produced at each pixel in a tile,
each SDS is associated with a number of standard Local Attributes that
apply to the data. For all the Land_Cover_Type_* SDSes they include:
| Name: |
Type: |
Value: |
| units |
HDF-STRING |
"class number" |
| valid_range |
HDF-uint8 |
0, 254 |
| _FillValue |
HDF-uint8 |
255 |
For all the Land_Cover_Type_*_Assessment SDSes, they include:
| Name: |
Type: |
Value: |
| units |
HDF-STRING |
"flags" |
| valid_range |
HDF-uint8 |
0, 254 |
| _FillValue |
HDF-uint8 |
255 |
For the Land_Cover_Type_QC SDS, they include:
| Name: |
Type: |
Value: |
| units |
HDF-STRING |
"concatenated flags" |
| valid_range |
HDF-uint8 |
0, 254 |
| _FillValue |
HDF-uint8 |
255 |
For the Land_Cover_Type_1_Secondary SDS, they include:
| Name: |
Type: |
Value: |
| units |
HDF-STRING |
"class number" |
| valid_range |
HDF-uint8 |
0, 254 |
| _FillValue |
HDF-uint8 |
255 |
For the Land_Cover_Type_1_Secondary_Percent SDS, they include:
| Name: |
Type: |
Value: |
| units |
HDF-STRING |
"percent in integers" |
| valid_range |
HDF-uint8 |
0, 254 |
| _FillValue |
HDF-uint8 |
255 |
For all the LC_Property_*, they include:
| Name: |
Type: |
Value: |
| units |
HDF-STRING |
"flags" |
| valid_range |
HDF-uint8 |
0, 254 |
| _FillValue |
HDF-uint8 |
255 |
Global Attributes (Metadata)
In addition to the albedo and quality information (SDSs and Local Attributes)
that is provided at a per-pixel level, each tile of the Level 3 MOD12Q1
Land Cover Product also includes three types of Global Attributes or Metadata
which summarize the tile. These are Core, Archive, and Structural Metadata.
Of particular interest to the user community are the Core Metadata QAFlags
of SCIENCEQUALITYFLAG and SCIENCEQUALITYFLAGEXPLANATION and some of the
AdditionalAttributes which identify the tile number or which summarize
the quality of the product over that entire tile (see the full MOD12Q1
specification
for
the complete listing).
|
