Conference PaperPDF Available

TERRABRASILIS: A SPATIAL DATA INFRASTRUCTURE FOR DISSEMINATING DEFORESTATION DATA FROM BRAZIL

Authors:

Abstract and Figures

DETER and PRODES projects have been pioneerers at monitoring large deforestation mapping areas in Brazil. They have proved with quality-assured data their commitment to bridge the gap between society and better management of resources. Such environmental programs complexity, however, requires the designing, implementation and deployment of a spatial data infrastructure able to easily disseminate those data. TerraBrasilis thus plays an important role by supporting storage and query web services using modularized and virtualized environments to enable high availability and performance. The key idea behind TerraBrasilis is to facilitate the access and analysis of deforestation data generated by PRODES and DETER, and perhaps further thematic mapping projects. In this paper, we demonstrated TerraBrasilis within GIS and analytics environments with deforestation data of the Cerrado biome.
Content may be subject to copyright.
TERRABRASILIS: A SPATIAL DATA INFRASTRUCTURE FOR
DISSEMINATING DEFORESTATION DATA FROM BRAZIL
Luiz Fernando Ferreira Gomes de Assis1, Karine Reis Ferreira.1, Lúbia Vinhas1, Luis Maurano1, Cláudio Aparecido de
Almeida1, Jether Rodrigues Nascimento1, André Fernandes Araújo de Carvalho1, Claudinei Camargo1, Adeline Marinho
Maciel1
1Image Processing Division, National Institute for Space Research (INPE),
Astronautas Avenue 1758, São José dos Campos, São Paulo, 12227-001, Brazil
{luizffga, jetherrodrigues, claudineicam}@gmail.com, {karine.ferreira, lubia.vinhas, luis.maurano, claudio.almeida,
andre.carvalho, adeline.maciel}@inpe.br
ABSTRACT
DETER and PRODES projects have been pioneerers at
monitoring large deforestation mapping areas in Brazil. They
have proved with quality-assured data their commitment
to bridge the gap between society and better management
of resources. Such environmental programs complexity,
however, requires the designing, implementation and
deployment of a spatial data infrastructure able to easily
disseminate those data. TerraBrasilis thus plays an
important role by supporting storage and query web services
using modularized and virtualized environments to enable
high availability and performance. The key idea behind
TerraBrasilis is to facilitate the access and analysis of
deforestation data generated by PRODES and DETER, and
perhaps further thematic mapping projects. In this paper,
we demonstrated TerraBrasilis within GIS and analytics
environments with deforestation data of the Cerrado biome.
Key words – Spatial Data Infrastructure, Deforestation
Data, PRODES, DETER, Cerrado Biome
1. INTRODUCTION
Internationally acclaimed projects such as the projects
of deforestation monitoring (PRODES) and near-real time
deforestation detection (DETER), conducted by the Brazilian
National Institute for Space Research (INPE) have provided
quality-assured, up-to-date, and spatially extensive official
deforestation data. They together have improved the
socio-economic and environmental processes linked to
the implementation of public policies for the sustainable
management of natural resources in Brazil [1–3]. Based
on GIScience techniques and satellite imagery analysis of
different resolutions, interpreters monitor critical areas to
conceive annual clear-cut deforestation rates and increments,
as well as alerts in near-real time of potentially dangerous
forest suppression circumstances. DETER and PRODES
projects also try to mitigate the loss of information caused
by seasonal cloud cover and canopy shadows, specific issues
related to tropical areas. As a result of the complex nature of
spatio-temporal data combined with a demand for processing
and visualization capabilities of large mapping areas, new
possibilities have been opened up to leverage emergent
technologies that enable their easy and flexible manipulation.
That is exactly where Spatial Data Infrastructure (SDI)
plays an important role. It aims to facilitate the discovery of
spatial data, by using metadata base catalogue and standard
web services (e.g., Web Mapping Service (WMS), Web
Feature Service (WFS) and Web Coverage Service (WCS)).
SDIs take advantage of large data storage connected to spatial
service plug-ins, for instance, object-relational database
such as PostgreSQL combined with GIS extensions (e.g.,
PostGIS) can keep vectorial data following Open Geospatial
Consortium (OGC) specifications. These tools are essential
to manage the high variability of thematic mapping projects
specially in large areas such as Cerrado biome in Brazil.
Cerrado is the second largest biome in Brazil, covering a
fourth of its territory. Over the last few years it has lost almost
24% of its original coverage due to the agriculture expansion
(e.g., soybean, cotton, and corn production), supressed
vegetation and pasture cattle. Its biodiversity richness
has been highlighted for the maintenance of environmental
resources and prevention against vanishing of thousand of
annimals and plants [4]. Cerrado’s degree of destruction has
reached such alarming rates that if it continues it will be
difficult to recover its biodiversity. With that in mind, much of
the attention that has flowed towards Amazon Forest over the
last few years while other biomes stayed in the background,
has cloven to Cerrado now.
For example, PRODES Cerrado project have generated
more than 2.5 million features from 2000 to 2017. In 2004,
more than 500 thousand features were created with about
60000 km2. Large features reach 213 km2, while small ones
reach meters scales. If one wants to analyze more regionally
information, further geoprocessing and adjusts are required
and that takes time. SDI thus must support both visualization
and downloading of pre-filtered spatial data, which is one of
the main goals proposed by TerraBrasilis [5]. TerraBrasilis
helps to organize, access and use spatial data produced by
INPE’s environmental monitoring programs, but throughout
a web portal, it makes possible based on customized views
to aggregate other types of spatial data. Rather than just
relying on geoservices, it uses ubiquitous clear and simple
APIs accross a cluster of virtualized machines to make spatial
data analysis easier. Furthermore, TerraBrasilis enables the
management of dynamic environments such as those found
in DETER project that produces daily data. That means, it
becomes reasonable to trace forest degradation and fire scars
areas every day even before they are deforestated.
Dealing with all these issues is not a trivial task,
since account must be taken of the following: (1) the
influence of regional governamental policies to increase the
resilience of Cerrado biome and to preserve its biodiversity
(2) the concern for handling the integrated and adaptive
management of historical and near-real time deforestation-
related rates, increments and alerts; (3) the expensiveness to
afford constantly the technology innovation transformations
that often follow SDI evolution; and (4) the degree
of SDI modularity with benefit of generic and flexible
implementations to other biomes .
Thus, the main contributions made by this work are the
following:
1. Engineering requirements, designing, implementing,
and evaluating an open-source SDI to organize
and disseminate deforestation data obtained from
consolidate thematic mapping projects such as DETER
and PRODES;
2. Learning lessons from the application of the proposed
approach in a real-world deforestation scenario that
has called attention for its fast natural anthropological
conversion, complex formation and high correlation to
soybean cultivation in Cerrado biome, Brazil.
2. TERRABRASILIS PLATFORM
TerraBrasilis aims to transform GIS society into large
data analysts by means of Geoinformatics and visual
analytics technologies. It leverages modularized processing
components of traditional SDI architectures to increase data
availability and interoperability with OGC standards, and
flexible and rapid API designs, developments and delivery.
2.1. TerraBrasilis - Combining Web Services for Maps
and Dashboards
Traditional SDIs such as TerraBrasilis require servers
for sharing geospatial data (e.g., Geoserver1), and robust
metadata search and catalogues (e.g., Geonetwork2)
following the ISO standard 19115:2003 Geographic
Metadata structure recommended by the Brazilian SDI
(INDE – Infraestrutura Nacional de Dados Espaciais)
initiative3. They also support heterogeneous databases
for dealing with spatio-temporal, thematic and business
data (e.g., PostgreSQL4, MongoDB5and Redis6). They
provide Web GIS libraries to build simple and responsive
interactive maps (e.g., Leaflet Javascript7), to visualize high
performance components (e.g., d3 Javascript8) and to filter
large multivariate datasets (e.g., crossfilter Javascript9) based
on modern web browsers’ standards (see PRODES Cerrado
deforestation in Figure 1 and 2).
The services of TerraBrasilis are modularized accross
a cluster of virtualized machines to facilitate future
deployments in different information technology (IT)
1http://geoserver.org/
2https://geonetwork-opensource.org/
3http://www.inde.gov.br/
4https://www.postgresql.org/
5https://www.mongodb.com/
6https://redis.io/
7https://leafletjs.com/
8https://d3js.org/
9http://square.github.io/crossfilter/
Figure 1: TerraBrasilis - Deforestation Layers.
Figure 2: TerraBrasilis - Deforestation Rates Charts.
infrastructure. All of them are packaged up using a
lightweight and standalone manner 10 so they can be wrapped
with all their dependencies in order to improve performance
and reliability of the computing environment. An http server
and a reverse proxy11 are also used to balance the web
requests.
2.2. Spatial Data Infrastructure: Improving GIS
Interoperability
A SDI aims to integrate a set of technologies, policies,
and standards for a more efficient exploitation, access and
dissemination of spatial data. It harmonizes and catalogues
those data so that any client who has access to the Internet
can manipulate them without specialized knowledge. In this
sense, TerraBrasilis offers data based on WMS, WFS and
WCS to monitor and coordinate environmental programs. It
allows users to open layers such as deforestated features using
any consolidate GIS platform (e.g., QGIS and ArcGIS). Users
just need to add a new layer, for instance, from a WMS Server
and TerraBrasilis data is there at their’s desktop. A simple
way using QGIS is demonstrated in Figure 3.
2.3. Transforming GIS Experts into Data Science
Analysts
To increase the data availability, we consider not only users’
GIS expertise, but also by means of the designing and
development of an API, we leverage their empowerment
with scripts’ language such as Rfor further data analysis.
This API follows a design pattern in which the write and
read information are different, used specially in scenarios
10https://docs.docker.com/engine/reference/builder/
11https://www.nginx.com/
Figure 3: Adding new Layer from a TerraBrasilis WMS Server
in QGIS.
where multiple representations are needed12. This approach
mitigates the need for creating additional columns or use a
multi-table approach when dealing with diverse data. Users
are then able to check layers information for a plenty of
local of interests by the seggregation of updates and queries
responsabilities. This helps to improve the representation of
a fraction or a portion of the overall data. The Listing 1
represents the resulting API paths operations to get all the
available data, its additional information such as their local of
interests (e.g., states, municipalities, conservation units, and
indigeneous areas) as well as a queryable data, by defining
classes (e.g., pasture), local of interest names (e.g., Mato
Grosso), and periods (e.g., from 2000 to 2017).
Listing 1: TerraBrasilis API Calls
GET api_path/ list_available_data
GET a pi _ p a t h / c o nf i g / l i s t _ a v a i l a b l e _ l o i n a m e
GET a p i _ p a t h / d a t a / d e f i n e d _ c l a s s / d e f i n e d _ l o i n a m e
GET a p i _ p a t h / q u er y ? c l a s s = . . . & l o in a m e = . . . & t i m e = . . .
In order to ensure the API interoperability with analytics
environments as well, an Rscript demonstration with
deforestation data can be seen in Figure 4. At first, it is
necessary to manipulate handlers with customized headers as
inputs to curl requests to list all the available data. With those
information, it is possible then to list the available local of
interests associated with that data. Those server responses are
casted to JSON format and finally a simple analysis such as a
linear estimation method to check a regression for each state
can be performed (see Figure 5). Without an automatized
manner, it would be difficult to see that Piaui is the only
state in which its deforestation rates has a positive coefficient
although its absolute values only represents 5%.
3. RESULTS AND DISCUSSIONS
We deployed an alpha version of TerraBrasilis so users
can access the Cerrado view of annual deforestation
rates (PRODES) and deforestation alerts (DETER) layers.
Throughout a nicely designed querying and visualization
front-end Web GIS, we demonstrated how to take advantage
and observe the deforestation dynamics in Cerrado.
Furthermore, it is highlighted how spatial database operations
were optimized and the fast web app loading performance.
12https://martinfowler.com/bliki/CQRS.html
Figure 4: Using the TerraBrasilis API within an analytics
environment (R Jupyter).
3.1. PRODES and DETER Cerrado Projects Data
Handling using TerraBrasilis: Lessons Learned from
Deforestation Data in Cerrado Biome
Although the complexity of deforestation dynamics hampers
its controlling and handling, SDI architectures such as
TerraBrasilis make easier the access and use of historical and
near-real time data. Together, they can improve planning
actions and mitigate deforestation by realeasing an up-to-
date overview to any interested citizen. There, PRODES
and DETER projects are fundamentally important since they
represent a surveillance strategy more suitable to dynamic
environments such as those found in Cerrado (see Figure 6).
Without an adequate management, it would be hard to depict
the dynamics of such large areas in a long-term period.
3.2. Performance Analysis of Web App Loading
Measuring performance of web pages is essential to
determine how much of our audience is satisfied. In this
experiment, we gathered the time metrics acquired over ten
Figure 5: Linear smooth of deforestation data for all the
Brazilian Cerrado States.
Figure 6: PRODES Cerrado Dynamics from 2000 to 2017
without 2001, 2003, 2005, 2007, 2009 and 2011.
requests repeatedly to emulate the behaviour of web browser
to fetch all the web pages files (e.g., css, javascript, images).
We highlighted that TerraBrasilis loads very quickly although
the experiment has no statistical significance. We also
emphasized the small amount of time users spent waiting the
scripting, rendering and painting of those files (see Table 1).
Table 1: Time in ms Comparison.
Mean Standard Deviation
Loading 78.69 20.44
Scripting 2388.03 205.09
Rendering 298.56 28.87
Painting 65.52 12.59
3.3. “SubDivide” and Conquer: Tunning Spatial
Database Operations for Query Performance
Optimization
In computer science, a well-known technique (divide and
conquer) to solve big problems aims to break them down
into small parts so each of the remaining part can be simply
answered. In order to process thousand of features faster and
extract area values according to specific local of interests, we
took advantage of the “subdivide” operation13. It converted
recursively a single large feature into small fractions using
as an input, the layer and a number of maximum vertex
13https://postgis.net/docs/ST_Subdivide.html
parameters to each fraction. It was recently optimized in
PostGIS to tune intersect operations but has been neglected
by the community [6]. This technique also explores spatial
indexes, inevitably used for spatial data. An overview of how
“subdivided” features and indexes are compared with regular
ones is presented in Figure 7. The idea is to take advantage of
small indexes to improve performance of database operations
since they work faster than large indexes.
Figure 7: ”Subdividing” PRODES Cerrado project
deforestation features from 2000 for query performance
optimization.
4. CONCLUSIONS
Remote sensing and geoprocessing techniques have been
applied to identify critical areas of historical clear-cut
deforestation rates and increments, as well as alerts in
near-real time of potentially dangerous forest suppression
circumstances. TerraBrasilis leverages emergent technologies
to enable DETER and PRODES project easily and flexible
manipulation. The results evidenced that TerraBrasilis is fast
and lightweight to use with spatio-temporal deforestation data
of large mapping areas. It provides APIs that interact with
GIS and analytics environments, and is able to depict the
dynamics of thematic mapping data such as high complex
data such as deforestation.
5. REFERENCES
[1] ESCADA, M. I. S. et al. Avaliação de dados dos sistemas de alerta da amazônia:
Deter e sad. Simpósio Brasileiro de Sensoriamento Remoto, v. 15, p. 2934–2943,
2011.
[2] DINIZ, C. G. et al. Deter-b: The new amazon near real-time deforestation detection
system. IEEE Journal of Selected Topics in Applied Earth Observations and Remote
Sensing, IEEE, v. 8, n. 7, p. 3619–3628, 2015.
[3] PRODES, P. Monitoramento da floresta amazônica brasileira por satélite. Instituto
Nacional de Pesquisas Espaciais Projeto Prodes. Available at http://www. obt. inpe.
br/prodes/index. php Accessed, v. 25, p. 2013, 2013.
[4] KLINK, C. A.; MACHADO, R. B. Conservation of the brazilian cerrado.
Conservation biology, Wiley Online Library, v. 19, n. 3, p. 707–713, 2005.
[5] INPE. Terrabrasilis. 2018. <terrabrasilis.dpi.inpe.br/>. [Online; accessed 25-
September-2018].
[6] RAMSEY, P. Scaling PostgreSQL and PostGIS. 2018. <https://vimeo.com/
250498574>. [Online; accessed 05-October-2018].
... Our hypothesis is to build on the research findings of a previous study [28] and design a spatial data analytics infrastructure called TerraBrasilis (http://terrabrasilis.dpi.inpe.br), that splits the main components of SDI into micro-services. ...
Article
Full-text available
The physical phenomena derived from an analysis of remotely sensed imagery provide a clearer understanding of the spectral variations of a large number of land use and cover (LUC) classes. The creation of LUC maps have corroborated this view by enabling the scientific community to estimate the parameter heterogeneity of the Earth’s surface. Along with descriptions of features and statistics for aggregating spatio-temporal information, the government programs have disseminated thematic maps to further the implementation of effective public policies and foster sustainable development. In Brazil, PRODES and DETER have shown that they are committed to monitoring the mapping areas of large-scale deforestation systematically and by means of data quality assurance. However, these programs are so complex that they require the designing, implementation and deployment of a spatial data infrastructure based on extensive data analytics features so that users who lack a necessary understanding of standard spatial interfaces can still carry out research on them. With this in mind, the Brazilian National Institute for Space Research (INPE) has designed TerraBrasilis, a spatial data analytics infrastructure that provides interfaces that are not only found within traditional geographic information systems but also in data analytics environments with complex algorithms. To ensure it achieved its best performance, we leveraged a micro-service architecture with virtualized computer resources to enable high availability, lower size, simplicity to produce an increment, reliable to change and fault tolerance in unstable computer network scenarios. In addition, we tuned and optimized our databases both to adjust to the input format of complex algorithms and speed up the loading of the web application so that it was faster than other systems
Article
Full-text available
The Brazilian Legal Amazon (BLA), the largest global rainforest on earth, contains nearly 30% of the rainforest on earth. Given the regional complexity and dynamics, there are large government investments focused on controlling and preventing deforestation. The National Institute for Space Research (INPE) is currently developing five complementary BLA monitoring systems, among which the near real-time deforestation detection system (DETER) excels. DETER employs MODIS 250 m imagery and almost daily revisit, enabling an early warning system to support surveillance and control of deforestation. The aim of this paper is to present the methodology and results of the DETER based on AWIFS data, called DETER-B. Supported by 56 m images, the new system is effective in detecting deforestation smaller than 25 ha, concentrating 80% of its total detections and 45% of the total mapped area in this range. It also presents higher detection capability in identifying areas between 25 and 100 ha. The area estimation per municipality is statistically equal to those of the official deforestation data (PRODES) and allows the identification of degradation and logging patterns not observed with the traditional DETER system.
Article
Full-text available
This paper presents an analysis comparing two deforestation monitoring system data, DETER and SAD, for the period of August 2008 to July 2009 for Amazônia Legal region. This study uses as reference clear cut deforestation data from PRODES, and forest degradation data from DEGRAD (INPE). DETER is a deforestation Alert System developed by INPE, and SAD is a non governmental Alert System developed by IMAZON. Both systems use low spatial resolution (250m) data from MODIS image, but different techniques to detect and map deforestation areas. In the period of analysis, DETER detected larger clear cut deforestation area than SAD, 2.705 km 2 against 1.510 km 2 , showing better capacity to support law-enforcement. However, the proportion of clear cut areas non-detected by these systems was high, 78% for DETER and 81% for SAD. The worse result obtained was for deforestation polygons size equal or smaller than 0,25 km 2 (25 ha), with ~4% of confirmation for both systems, despite the fact that SAD minimum mapping area is 6,25 ha and DETER minimum area is 25 ha. This result reinforce that 25 ha is the limit size area of MODIS detection capacity. The analysis of PRODES data showed a tendency of increasing the proportion of clear cut areas smaller than 25 ha, from 22% of the total area in 2002 to 60% in 2009. To improve the Alert Systems detection capacity, considering this scenario, is necessary to investigate the possibility of incorporating other sensors data with higher spatial resolution, maintaining high temporal resolution. Palavras-chave: forest monitoring system, clear cut deforestation, forest degradation, MODIS, sistema de monitoramento da floresta, desmatamento por corte raso, degradação florestal.
Conservation of the brazilian cerrado
  • C A Klink
  • R B Machado
KLINK, C. A.; MACHADO, R. B. Conservation of the brazilian cerrado. Conservation biology, Wiley Online Library, v. 19, n. 3, p. 707-713, 2005.
2018. <terrabrasilis.dpi.inpe.br/>. [Online; accessed 25
  • Inpe
  • Terrabrasilis
INPE. Terrabrasilis. 2018. <terrabrasilis.dpi.inpe.br/>. [Online; accessed 25-September-2018].
Monitoramento da floresta amazônica brasileira por satélite. Instituto Nacional de Pesquisas Espaciais Projeto Prodes
  • P Prodes
PRODES, P. Monitoramento da floresta amazônica brasileira por satélite. Instituto Nacional de Pesquisas Espaciais Projeto Prodes. Available at http://www. obt. inpe. br/prodes/index. php Accessed, v. 25, p. 2013, 2013.
Scaling PostgreSQL and PostGIS
  • P Ramsey
RAMSEY, P. Scaling PostgreSQL and PostGIS. 2018. <https://vimeo.com/ 250498574>. [Online; accessed 05-October-2018].