Skip to main content
ORNL DAAC HomeNASA Home

DAAC Home > Get Data > Field Campaigns > Large Scale Biosphere-Atmosphere Experiment (LBA-ECO) > Data Set Documentation
LBA-ECO LC-22 Field Validation of MODIS Deforestation Detection, Brazil, 2003-2004

Get Data

Revision date: December 10, 2014

Summary:

This data set contains field observations, corresponding GPS points, and point and polygons of deforested areas in the state of Mato Grosso, Brazil, for the period August 2003 to July 2004. The field observations were conducted in the forested areas between Nova Mutum and Sinop, MT. These data were part of a study to validate Moderate Resolution Imaging Spectroradiometer (MODIS) data at 250-m resolution for the detection of deforested areas.

There are 16 data files with this data set. This includes 10 shapefiles (.shp) and six comma-separated files (.csv).

DATA QUALITY STATEMENT: The Data Center has determined that there are questions about the quality of the data reported in this data set. The data set has missing or incomplete data, metadata, or other documentation that diminishes the usability of the products.

KNOWN PROBLEMS: The definitions of the fields in the shapefiles and related *.csv files were not clearly defined. Many of the fields are in common (point and polygon IDs), but the exact connection of observations in the *.csv files to the shapefiles is left up to the users. How random subsets were defined and how other subsetting criteria were applied were not provided. These are the field observations used to validate MODIS land cover classification as described in Morton et al., 2005.

Data Citation:

Cite this data set as follows:

DeFries, R., Y.E. Shimabukuro, D.C. Morton, L.O. Anderson, and F. del Bon Espirito-Santo. 2014. LBA-ECO LC-22 Field Validation of MODIS Deforestation Detection, Brazil, 2003-2004. Data set. Available on-line (http://daac.ornl.gov) from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, Tennessee, USA. http://dx.doi.org/10.3334/ORNLDAAC/1262

Implementation of the LBA Data and Publication Policy by Data Users:

The LBA Data and Publication Policy [http://daac.ornl.gov/LBA/lba_data_policy.html] is in effect for a period of five (5) years from the date of archiving and should be followed by data users who have obtained LBA data sets from the ORNL DAAC. Users who download LBA data in the five years after data have been archived must contact the investigators who collected the data, per provisions 6 and 7 in the Policy.

This data set was archived in December 2014. Users who download the data between December 2014 and November 2019 must comply with the LBA Data and Publication Policy.

Data users should use the Investigator contact information in this document to communicate with the data provider.

Data users should use the Data Set Citation and other applicable references provided in this document to acknowledge use of the data.

Table of Contents:

1. Data Set Overview:

Project: LBA (Large-Scale Biosphere-Atmosphere Experiment in the Amazon)

Activity: LBA-ECO

LBA Science Component: Land Use and Land Cover

Team ID: LC-22 (DeFries / Shimabukuro)

The investigators were DeFries-Bajpai, Ruth; Shimabukuro, Yosio Edemir; Anderson, Liana Oighenstein; Coura, Samuel Martins da Costa; Espirito-Santo, Fernando Del Bon; Hansen, Matthew C.; Jasinski, Ellen W.; Latorre, Marcelo Lopes; Lima, Andre; Morton, Douglas; Piromal, Rodrigo Sbravatti and DeFries, Ruth Sarah. You may contact Morton, Douglas C. (morton@geog.umd.edu).

LBA Data Set Inventory ID: LC22_MODIS_Field_Val_2004

This data set contains field observations, corresponding GPS points, and point and polygons of deforested areas in the state of Mato Grosso, Brazil, for the period August 2003 to July 2004. The field observations were conducted in the forested areas between Nova Mutum and Sinop, MT. These data were part of a study to validate Moderate Resolution Imaging Spectroradiometer (MODIS) data at 250-m resolution for the detection of deforested areas.

Related data set:

LBA-ECO LC-22 Field Validation of MODIS Deforestation Detection, Brazil, 2005. MODIS field validation data collected in Mato Grosso from March 17-24, 2005.

2. Data Characteristics:

There are 10 shapefiles (.shp) and six comma-separated (.csv) files with this data set. The data provide field observations and corresponding GPS points, preliminary data sets for MODIS deforestation detection in the state of Mato Grosso, Brazil for the period August 2003 to July 2004, and point and polygon samples with corresponding site descriptions.

All shapefiles have the following projection parameters:

Geographic Coordinate System:

GCS_WGS_1984 Datum: D_WGS_1984

Prime Meridian: Greenwich

Angular Unit: Degree

Table 1. Data file names and descriptions:

File Name Description
br163_sample_points.csv This file contains field and image classification values for 100 random roadside samples between Sinop and Cuiaba, MT. These points correspond to the point shapefile in br163_sample_points.zip.

interp_gridcode (image classification) and field_code (field observation) value definitions:
0=forest, 1=ag, 2=pasture, 3=mixed, 4=change, 5=cerrado.

Values are a preliminary estimate of map accuracy for an initial version of the 2002-2003 Mato Grosso land cover classification based on NDVI and EVI time series.
br163_sample_points.zip When expanded, this .zip file contains a shapefile with point geometry and is comprised of seven files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, *.xml, and *.shx). The shapefile consists of 98 randomly selected points along the BR163 between Sinop and Cuiaba, Mato Grosso.
mt_road_intersect_mod13_deforestation_polygons.zip When expanded, this .zip file contains a shapefile with polygon geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile contains potential deforestation polygons derived from MOD13 red reflectance data from composite 161-176 (julian days) 2004.

Only polygons that intersect dirt or asphalt roads (coverage from FEMA, 2002) are shown (n=513)
.

The subset of polygons that were visited in the field (n=120) are found in the file 120_field_visited_mod13_polygons.zip.
mt_offroad_mod13_deforestation_polygons.zip When expanded, this .zip file contains a shapefile with polygon geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile contains potential deforestation polygons derived from MOD13 red reflectance data from composite 161-176 (julian days) 2004.

It includes only polygons that do not intersect roads in the FEMA 2002 coverage.
Some field-visited polygons were drawn from this sample of potential deforestation.
polygon_fieldobs_all.csv This file provides the polygon ID, observation notes, observation date, and clearing status for 143 polygons observed during the field campaign.  ~120 were observed to be deforested. About 90% were correctly classified by MOD13 imagery.

The spatial locations for each observed deforestation polygon are found in the corresponding shapefile 120_field_visited_mod13_polygons.shp.
120_field_visited_mod13_polygons.zip When expanded, this .zip file contains a shapefile with polygon geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile is a subset of all potential deforestation polygons identified in MOD13 composite 161-177 (Julian days) 2004 that were observed in the field.
central_mt_mod13_points_on_road.zip When expanded, this .zip file contains a shapefile with point geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile is a sample of all MOD13 250-m resolution pixel center points within the forest mask that lie within a 100-m buffer of roads in Mato Grosso.

Sample points are further classified as either forest or deforested for field validation purposes. A subsample of these points was visited in the field for validation of classification accuracy on a pixel-by-pixel basis (n=482).
points_fieldobs_all.csv This file provides observations, dates, and agreement between observational teams for each MODIS pixel visited in the field. Two teams visited 199 of the 482 pixels and classification agreement was 95%. Or, disagreement between observers was 5 percent (10 points).
field_visited_482_points_with_observations.zip When expanded, this .zip file contains a shapefile with point geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile is a subsample of all MOD13 gridded product pixel center points that fall within 100-m of a road in Mato Grosso (FEMA 2002 road coverage) and were observed during the field campaign.

Forested (0) or deforested (1) status and date of observation are noted in the file points_fieldobs_all.csv.
forest_disagree_points.zip When expanded, this .zip file contains a shapefile with  point geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile is a subset of all visited MODIS pixel center points where field and classification results differed.

In all cases where points classified as forest were observed as deforested in the field, pixels contained some subpixel fraction of deforestation. Stated differently, these pixels lie at the forest/non-forest edge.
gps_track.zip When expanded, this .zip file contains a shapefile with point geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile maps the areas visited during the field campaign (7-14 to 7-22).

The focal area for the campaign was the transition forest region of central Mato Grosso and field observations concentrated on the forested areas between Nova Mutum and central Mato Grosso.
omission_waypoints.csv This file provides latitude, longitude, waypoint, date, and comments pertaining to the deforestation areas that were not detected in the MODIS classification results.

Omission of deforestation detections was difficult to estimate in a statistically rigorous fashion. However, the team did record the size, condition, and location of deforestation areas that were not detected in the MODIS classification results. These observations are listed in this data file.
obs1_waypoint_descriptions.csv Latitude, longitude, waypoint, date, and descriptions made by Doug Morton during field observation trips. The data correspond to the shapefile obs1_waypoints.shp.
obs1_waypoints.zip This shapefile has point geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). This shapefile contains other waypoints of interest in addition to the condition and location of new deforestation events that were noted during the field campaign.
obs2_waypoint_fieldobs.csv This file provides latitude, longitude, waypoint, date, and comments made by Ruth DeFries during field observation trips. The data correspond to the shapefile in the .zip file obs2_gps_points.zip.
obs2_gps_points.zip When expanded, this .zip file contains a shapefile with point geometry and is comprised of six files (*.dbf, *.prj, *.sbn, *.sbx, *.shp, and *.shx). The shapefile contains other waypoints of interest that mark boundaries of visited polygons, locations of specific land cover classes, or other relevant information from the field campaign.

Site boundaries: (All latitude and longitude given in decimal degrees)

Site (Region) Westernmost Longitude Easternmost Longitude Northernmost Latitude Southernmost Latitude Geodetic Datum
Mato Grosso - Sinop (Mato Grosso) -58.25-54-10.96-15.65World Geodetic System, 1984 (WGS-84)

Time period:

Platform/Sensor/Parameters measured include:

3. Data Application and Derivation:

These data were collected to serve as validation information for classification of remotely sensed data. Field information and observations describe the ground conditions at particular MODIS 250-m pixels and polygons of new deforestation that can be used to train or validate algorithms for deforestation detection.

From Morton et al. (2005), field validation of MODIS classification results in central Mato Grosso generated two important measures of classification accuracy. The method proved highly accurate for identifying deforestation events, with 100% accuracy for observed clusters larger than two pixels. Detection accuracy was lower on a per-pixel basis because of subpixel deforestation edge effects. All pixels that were classified as forest, but were observed to contain some new deforestation, were located at the forest/deforestation edge. Future analyses will focus on the subpixel fraction of deforestation within these edge pixels.

Study design for field validation made omission errors difficult to quantify. Pixel- and cluster-based approaches were stratified by accessibility with a vector layer of Mato Grosso roads. All pixels and clusters that intersected a road were considered part of the validation sample. Field observers were directed to forest and deforested samples based on classification results, not to likely areas for omission. As a result, characterization of omission was only possible through anecdotal evidence from chance observations. Future validation campaigns should more accurately quantify omission errors. Field validation was made possible through quality equipment and highly accessible terrain. The combination of near-coincident Landsat TM and MODIS data with real-time GPS proved excellent for evaluating MODIS pixels and clusters in the field. The unpaved road network in central Mato Grosso is extensive, which facilitated access to a variety of sampling conditions. Ground-based field validation without high-resolution imagery or roaded access would yield significantly fewer sample observations.

4. Quality Assessment:

Locational accuracy of GPS point observations was +/- 15 m. Deforestation classification disagreement between observers was five percent.

Problems identified by the Oak Ridge National Laboratory (ORNL) Distributed Active Archive Center (DAAC):

The definitions of the fields in the shapefiles and related *.csv files were not clearly defined. Many of the fields are in common (point and polygon IDs), but the exact connection of observations in the *.csv files to the shapefiles is left up to the users. How random subsets were defined and how other subsetting criteria were applied were not provided.

5. Data Acquisition Materials and Methods:

The goal of this study was to develop, test, and validate simple and efficient methods for the annual or more frequent monitoring of deforestation in the Brazilian Amazon as a basis for prioritizing high-resolution analyses and characterizing recent deforestation dynamics. The focal area for the campaign was the transition forest region of central Mato Grosso and field observations were conducted in the forested areas between Nova Mutum and Sinop, MT.

Based on performance in a test scene analysis, the single-date threshold method with MOD13 red reflectance was chosen for validation with field observations in central Mato Grosso, particularly on the forested areas between Nova Mutum and Sinop, MT in July 2004 (Morton et al., 2005). Ground-based teams of observers assessed per-pixel and per-cluster accuracy in the following manner:

Field observations and records were kept using a Garmin 76s GPS unit. Locational accuracy of GPS point observations was +/- 15 m.

PRODES deforestation data, including remaining forest cover, for the Brazilian Amazon can be downloaded in shapefile or Spring format at: http://www.obt.inpe.br/prodes. Please see related journal citations for relevant pre-processing methodology and related results.

6. Data Access:

This data set is available through the Oak Ridge National Laboratory (ORNL) Distributed Active Archive Center (DAAC).

Data Archive Center:

E-mail: uso@daac.ornl.gov
Telephone: +1 (865) 241-3952

7. References:

Morton, D.C., R.S. DeFries, Y.E. Shimabukuro, L.O. Anderson, F. Del Bon Espirito-Santo, M. Hansen, and M. Carroll. 2005. Rapid Assessment of Annual Deforestation in the Brazilian Amazon Using MODIS Data. Earth Interactions 9(8):1-22.

Related Publication