IEEE Geoscience and Remote Sensing Magazine - September 2019 - 112

GP-BASED CLASSIFIER
The GPC module (Figure 4) allows the training of a binary
classification model using remote sensing images as input
(previously proposed in [15]). The user must already have
downloaded the input images through, for example, the
DAM module. In the Processing tab, the user must provide a
shapefile of the region of interest (ROI) and also a shapefile
containing information about the positive class of interest
(COI). The negative class may be easily inferred by subtracting the COI region from the entire ROI. After providing the
necessary inputs, the GPC module selects the number of
points that will be used in the training, validation, and test
steps. The GPC preprocesses the input data (as described in

DAM
BE

TSR
GPC

Uploads
Data

[15]), computes the distance matrices, and, finally, trains the
GP model. After the GP model is trained, the user may view
the model's balanced accuracy and the syntax-tree representation of the GP individuals in the Results tab.
BFAST EXPLORER
In Figure 5, we present a functional overview of the BE module. A map is displayed in the Map tab of the module, and the
user can specify points of interest to be analyzed. The user
may optionally upload shapefiles to help localize and select
those points. After choosing the points, the user must specify
which Landsat missions to download (missions 5, 7, and/or 8
are currently available). After data are acquired through the
download module, the Analysis tab
becomes available. In this tab, the user
may visualize the time series of several
spectral bands from the surface reflectance product and some precomputed
vegetation indexes. In addition, it is
possible to run some BFAST-based algorithms to detect structural breaks in
Provides
Data
the time-series data.

Input Tab

User

MATERIALS AND METHODS
STUDY AREAS
In this section, we present the study areas
used as input to our Tucumã toolbox.

Exploratory
Analysis
Exploratory Tab
Provides Reference Image

Retrieval
Task
Retrieval Tab

FIGURE 3. A functional overview of the TSR module.

DAM
BE

TSR
ROI and COI
Definition

Processing
Tab

Preprocesses
Data
Computes
Distance
Matrices

User

GPC

View Accuracies
and Syntax Trees of
Results Tab
Trained GP
FIGURE 4. A functional overview of the GPC module.

112

Runs
GP

EMAS NATIONAL PARK
To demonstrate the usefulness of the
DAM and TSR modules, we focused
on a region inside Emas National
Park (see Figure 6). Emas National
Park is an ecological reserve with approximately 133,000 ha of extension,
located in the Brazilian Central Plateau (Cerrado core region) between
Goiás and Mato Grosso do Sul [17].
Emas National Park is the largest
continuous preserved piece of the
Cerrado region in a protected area,
and it is covered predominantly by
grasslands, with small patches of
cerrado (sensu stricto, a savanna vegetation type), marshes, and gallery
forests [18].
Approximately 1,500 mm of rain
falls during the wet season from September to May [19]. During the dry
season, from June to August, there is
practically no rainfall in this region
[18]. The temporal patterns of vegetation growth and reproduction are
strongly related to climatic seasonality in tropical savannas, as is the case

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

SEPTEMBER 2019

IEEE Geoscience and Remote Sensing Magazine - September 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2019