Content uploaded by Alessandro Samuel-Rosa
Author content
All content in this area was uploaded by Alessandro Samuel-Rosa on Feb 20, 2018
Content may be subject to copyright.
Bringing Together Brazilian Soil Scientists to Share Soil Data
Alessandro Samuel-Rosa(1); Ricardo S. D. Dalmolin(2); Paulo Ivonir Gubiani(2); Wenceslau
Teixeira(3); Stanley R. de M. Oliveira(4); João Herbert M. Viana(5); Carlos G. Tornquist(6);
Lúcia Anjos(7); José João L. L. de Souza(8); Eloi Ribeiro(9); Marta Ottoni(10); Paula S. C. de
Medeiros(11); José Miguel Reichert(2); Diego S. Siqueira(12); José Marques Júnior(12); José A. M.
Demattê(13); André C. Dotto(13); Leonardo Collier(14); Gustavo M. Vasques(3); Gustavo
Valladares(15); Fabrício A. Pedron(2); João C. Pedroso Neto(16); José M. F. Alba(17); Ronaldo P. de
Oliveira(3); João Henrique Caviglione(18); Pablo Miguel(19); Humberto G. dos Santos(3); Carlos
A. Flores(17); Igo Lepsch(13); Diego José Gris(2); Nícolas Augusto Rosin(2); Jean M. Moura-
Bueno(2)
(1) Universidade Federal de Santa Maria, Av. Roraima 1000, Cidade Universitária, Bairro Camobi, Santa Maria - RS, CEP 97105-900,
Brasil, e-mail: alessandrosamuelrosa@gmail.com; (2) Universidade Federal de Santa Maria; (3) Embrapa Solos; (4) Embrapa
Informática Agropecuária; (5) Embrapa Milho e Sorgo; (6) Universidade Federal do Rio Grande do Sul; (7) Universidade Federal Rural
do Rio de Janeiro; (8) Universidade Federal do Rio Grande do Norte; (9) ISRIC - World Soil Information; (10) Serviço Geológico do
Brasil; (11) Instituto Brasileiro de Geografia e Estatística; (12) Universidade Estadual Paulista; (13) Universidade de São Paulo; (14)
Universidade Federal de Goiás; (15) Universidade Federal do Piauí; (16) Empresa de Pesquisa Agropecuária de Minas Gerais; (17)
Embrapa Clima Temperado; (18) Instituto Agronômico do Paraná; (19) Universidade Federal de Pelotas.
INTRODUCTION
A great deal of soil data has already been
produced as part of soil survey and research
projects. Most of the published datasets are
published as a single paper, and the primary data
is unavailable to other researchers. If not
published soon, then the data probably will remain
unpublished forever. As data underutilization is a
waste of resources and refrains the advancement
of knowledge, many isolated soil data rescue and
sharing efforts have emerged (Arrouays et al.,
2017). But a consistent solution to the problem of
permanently safeguarding and promoting the
reusability of all kinds of soil data has yet to be
developed.
Lately, soil scientists have increased their
concerns with data discoverability and reusability,
the first two of the Three Laws of Open Data
(Eaves, 2009). Discoverability is the ability of a
dataset to be found by someone else. Reusability
is the ability that a dataset has to be used again by
its producer and/or someone else. Both
discoverability and reusability are critical to
ensuring the reproducibility of the research, a
basic principle of the scientific method.
Brazilian soil scientists have recently created a
soil data repository using community-built
standards and following open data policies in an
attempt to address the issues mentioned above.
The Free Brazilian Repository for Open Soil Data
– febr –, accessible through www.ufsm.br/febr, is
a centralized repository targeted at storing open
soil data and serving it in a standardized and
harmonized format, for various applications. This
paper describes the features of febr and the
opportunities that it creates for soil science.
FEATURES
Data model
Unlike existing soil databases, febr was designed
to allow individualized management of datasets.
First, because such a design highlights datasets
authors, helping them to be properly
acknowledged and cited by others. Second,
because it gives the flexibility to accommodate
many types of data of any soil variable. This is
accomplished by storing datasets using a directory
structure, each dataset being in its own directory.
The data model used in febr to organize each
dataset takes into consideration that soil
observations generally have four operational
dimensions. The first two are the x and y
horizontal spatial coordinates, referring to some
predetermined standard coordinate reference
system (CRS). The third is the temporal
coordinate, t, the moment – according to some
predetermined standard calendar and time zoning
system – when the soil was observed. The fourth
and last operational dimension is the soil
observation depth, z, as measured using some
predetermined standard scale, e.g. metres. At a
point in (geographic) space and time, [x, y, t], or
in space, time and depth, [x, y, t, z], a soil
observation is accompanied by an attribute space.
The latter is a multi-dimensional space defined by
a set of attributes of the environment (land use,
slope, parent material) or soil layer (pH, cec,
carbon content), respectively.
Two tables per dataset are used to cope with
the multi-dimensional character of soil data:
'observacao', for space-time data, and 'camada',
for space-time-depth data. The relation between
them is established using a identification key
included in both tables – the column
'observacao_id'. A third table, 'metadado', is
used to store the data about the methods used to
produce the soil data, i.e. the metadata. A fourth
table, 'dataset', stores data about the dataset as a
whole (general description, author name and
contact, version). Last, all febr maintainers tasks
(changes, improvements, todo list) are recorded in
table 'tarefa'.
The directory structure of febr is implemented
in Google Drive, while spreadsheets (Google
Sheets) are used to store data tables. Spreadsheets
are familiar to any soil scientist, eliminating the
hurdle of having to learn a new software and/or
data structures. For this reason, it is very easy to
enter, manipulate, and visualize soil data in febr.
It facilitates the participation of soil students and
experts, as well as non-specialists in soil research,
in soil data recovery and quality assessment
exercises.
Visualization
There are two ways to search for data in febr. The
first is via the global visualization page,
www.ufsm.br/febr/view, which shows the
distribution of soil observations across the
Brazilian territory – provided they have spatial
coordinates (Figure 1). Various additional
geographical data layers can be accessed in the
global visualization page to assist navigation and
geolocation (terrain, street and road network,
vegetation). These data comes from different
providers such as Esri, OpenStreetMap,
OpenWeatherMap, and Stamen, through the
JavaScript library Leaflet accessed through the R-
packages leaflet and mapview. By clicking on a
point, a popup window appears, showing a link to
the page of the dataset in the febr catalog.
Figure 1. Snapshot of the visualization page of
febr. The popup window on a point in Xanxerê
(SC) has a link to its page in the febr catalog.
Search
The second search tool available in febr is the
dedicated search page, www.ufsm.br/febr/search
(Figure 2). It was implemented using the
JavaScript library DataTables – a plugin for the
JavaScript library jQuery – through the R-package
DT. In addition to finding a dataset, the search
page helps discovering other data sets and
learning how they are related to each other. For
this purpose, seven search criteria are used:
dataset title, dataset first author name, dataset first
author organization, federative unit (state) with
the largest number of observations, total number
of observations, indexing terms and knowledge
area of Soil Science according to CAPES and
CNPq. Clicking on the identification code of a
dataset opens its page in the febr catalog.
Figure 2. Snapshot of the search page of febr.
The search term “Xanxerê” returned a single
result (dataset_id = 'ctb0629').
Catalog
Each dataset has an individual page in the febr
catalog, www.ufsm.br/febr/catalog/. Implemented
using the R-package bookdown, the catalog
details the key features of each dataset. This
includes a general description, especially its origin
and how it was produced, and the people and
institutions responsible for its production.
Secondary information and the spatial distribution
of observations – provided they have spatial
coordinates – are also available. Unlike the global
visualization page, the local catalog spatial
visualization tool can be used to evaluate the
quality of geospatial data in more detail. The
catalog also includes a search tool via the jQuery
JavaScript library. Finally, the catalog gives
access to dataset directories and respective
spreadsheets in Google Drive, where they can be
edited (upon request) or downloaded from.
Download
Datasets in febr can be downloaded in two
different ways. The first is to access and download
the spreadsheets directly from Google Drive,
where the output file format can be selected in the
File menu, with the options XLSX, ODS, PDF,
HTML, CSV, and TSV. The second way is to use
the R-package febr (Figure 3). It has four key
functions, dataset, observation, layer, and
metadata, each of them designed to download the
tables 'dataset', 'observacao', 'camada', and
'metadado', respectively. Various arguments can
be passed to these functions to select the datasets
and variables that should be downloaded. The
connection with Google Sheets is established
using the R-package googlesheets.
# Install packages
if (!require(devtools)) {
install.packages(pkgs = "devtools")
}
devtools::install_github(
repo = "febr-team/febr-package")
# Download observations with all variables
obs <- febr::observation(
dataset = "ctb0629", variable = "all")
# Download layers with all variables
lrs <- febr::layer(
dataset = "ctb0629", variable = "all")
# Merge data frames
ctb0629 <- merge(
x = obs, y = lrs,
by = c("dataset_id", "observacao_id"))
Figure 3. Installation and usage of the R-package
febr. The code chunk shows how to download and
merge data from dataset 'ctb0629'.
Routines for data standardization and
harmonization have been implemented in the R-
package febr. The former includes dealing with
measurement units and number of decimal places,
irregular transitions between sampling layers,
symbols used to indicate detection limits (+, <, >)
or missing data (N/A, NA, blank), coordinate
reference systems, and so on. The later consists of
rules to translate soil attributes determined using
disparate analytical methods into a common
attribute space. For example, to harmonize the
iron contents in the sulfuric acid extract
determined by atomic absorption spectrometry
and by inductively coupled plasma spectrometry.
The benefit of these routines is that upon
download the data is virtually ready for analyses
and modelling.
Discussion group
The febr is a project with challenging goals. This
requires much discussion and the collaboration of
countless people. A public discussion group was
created to facilitate these. One can send a message
to febr-forum@googlegroups.com to join the
discussion group and then influence the definition
of standards and data management choices,
collaborate in research and development activities,
propose new features and improvements, help
solving questions and provide technical support,
and point out data inconsistencies. Also, by
posting to febr-forum@googlegroups.com one can
learn how to publish data in febr.
Documentation
A comprehensive documentation, prepared using
the R-package bookdown, is available at
www.ufsm.br/febr/book/. Constantly updated, the
documentation is fundamental to facilitate the
adoption of agreed standards by those interested
in publishing data in febr as well as by the febr
maintainers. This guarantees, for example, that
febr maintainers can be replaced without harming
the progress of the activities.
OPPORTUNITIES
The febr currently has 14,477 soil observations
from 232 datasets covering the Brazilian territory
(Figures 1 and 4). Figure 4 in special shows that
many states have a poor coverage in febr, mainly
from the North and Centre-West regions of Brazil,
where soil data sharing should be further
encouraged. The potential benefits of having this
enormous volume of soil data freely available,
standardized and harmonized, for short and long
term scientific reuse, are countless. For example,
improving the Brazilian Soil Classification
System and international ones (Universal, WRB,
Soil Taxonomy), creating intelligent fertilizer
recommendation engines, developing specialized
databases, calibrating pedotransfer functions,
supporting the upcoming Brazilian National Soil
Survey Program (PronaSolos) and international
soil mapping initiatives (GlobalSoilMap, Global
Soil Partnership), and many others. In the long
term, the project results should promote a cultural
change towards a more open and collaborative
soil science in Brazil. By sharing data through a
centralized, collaborative and community-based
soil data storing and sharing facility, soil scientists
from different fields have the opportunity to
increase collaboration and the much needed soil
knowledge.
Figure 4. Relative distribution and areal density
(per 1000 km2) of observations among federative
units (states) – 42% of observations are from the
South and Southeast regions.
ACKNOWLEDGEMENTS: The authors thank
all soil scientists who have collaborated in the
development of febr by sharing data and ideas.
REFERENCES
Arrouays D et al. Soil legacy data rescue via
GlobalSoilMap and other international and national
initiatives. GeoResJ. 2017;14:1-19.
doi:10.1016/j.grj.2017.06.001
Eaves D. Three Laws of Open Data. Nov/2009.
Accessed in: 29 Nov. 2009. Available at:
https://eaves.ca/2009/11/29/three-laws-of-open-data-
international-edition/.
