Weather forecasting is an important application of Science to plan daily activities, being used in agriculture, air traffic, sea, forestry, severe weather alerts, military operations and enterprise systems. The weather forecast information is provided by several systems with different formats, parameters and with a large number of heterogeneous concepts used in different organizations. To minimize this problem, the number of these heterogeneous concepts must be reduced in order to provide uniform access for better decision making. This article describes the construction of a wellfounded ontology for the Numerical Weather Prediction data domain (NWPOntology), developed from philosophically based foundations of the Unified Foundational Ontology as a solution to promote the semantic enrichment of data from meteorological observations, aiming to reduce conceptual ambiguities.
Existing weather forecast centers focus on presenting data to the forecaster supporting some important decisions. Most of the data analysis and assembling key decisions in a forecasting policy takes place in the minds of analysts. This is usually transformed into forecast products finished in a simple text editor with large amounts of complex, semistructured and multivariate data, having to interpret them in a short time. Therefore, one of the biggest challenges in geospatial research is to generate accurate data.
In Meteorology, it is necessary ensure data consistency in the most efficient way. Inconsistencies and variations in the data forecasting series may occur due to several reasons, such as, human error in data acquisition, inadequate maintenance and calibration of the measuring instrument, erroneous data processing, random variations expected for any weather phenomenon and changes in the surrounding environment. In this context, the weather forecasting centers may use different syntaxes and terminologies for describing information as well as the same term may have different meanings and interpretations in different organizations.
For example, in the sentence "Maps of daily temperature and precipitation are
produced," an expert would recognize that the observation is "temperature" but could
not determine the details related to the temperature concept (atmospheric temperature,
sea surface temperature, etc.). He needs to ask the data provider to get more details
[
On the other hand, various terms may correspond to the same meaning. For
instance, the Observatory of Sahara and Sahel (OSS) may use the word “rainfall” for the
same real-world feature that usually refers to “precipitation” in other sources. The
variety of terms complicates the work of emergency responders who should be
familiar with the terms used in each discipline [
Semantic Sensor Web (SSW), SWEET and Next Generation Network Enabled
Weather (NNEW) are ontologies designed to represent the domain of weather data (to
be detailed in the Section2), but they did not emphasize numerical weather prediction,
which is essential for an area that is constantly evolving. To provide conceptual
methodological support for the exchange of knowledge on Numerical Weather
Prediction Data, NWPOntology is being developed, a domain ontology for numerical
weather prediction data based on Ontology of Unified Foundational Ontology (UFO)
[
This document introduces NWPOntology as follows: Section 2 reviews some related work from ontology in weather data domain; Section 3 details NWPOntology; and finally, Section 4 presents final observations and topics for further investigation.
There exists a huge number of sensor networks around the globe designed to monitor a
large set of different phenomena, creating a vast amount of data. Making them available
demands to structure such data as well as to allow the interoperability between different
sensor networks. Semantic Sensor Web (SSW) is an approach that builds upon Sensor
Web Enablement (SWE) [
The SSW ontology does not support forecast values. However, it would be
possible to extend the SSW ontology to implement them. In the context of
NWPOntology, the concept Observations, as well as its subconcepts, could be reused;
however, some additional subconcepts should to be added. The Basic (WGS84 lat/long)
for location [
Another approach towards a semantically-enriched sensor network, based on
SWE, is an OWL 2 ontology created by the W3C Semantic Sensor Network Incubator
group (SSN-XG) which is referred to as the SSN Ontology [
SWEET is a set of more than 200 ontologies comprising about 6000 concepts
[
In the context of weather forecasting, SWEET ontologies are a well-qualified approach for reusing them in a climate ontology. Time definitions, and geographical position specifications are supported. However, SWEET comes with some disadvantages, which ultimately led to the decision not to use them in the implementation of NWPOntology to support a future decision-making system: it does not currently cover future events, as well as does not cover all weather elements relevant to numerical weather prediction data, e.g., numerical weather prediction models.
The Next Generation Network Enabled Weather (NNEW) ontology is designed
to provide a comprehensive view on the weather across the country, built from
thousands of single weather observations. One of the tasks foreseen for the system was
the development of ontologies for the meteorological domain, mainly for the reduction
of delays related to the climate in the airspace of the USA to half of its current
magnitude. Recently, approximately 70 percent of all air traffic delays are attributable to
weather [
This ontology was built with respect to SWEET ontologies to map weather phenomena. Extensions to SWEET include additional weather phenomena and concepts and relations that lead to the 4-D Wx Data Cube (Four Dimensional Weather Data Cube) which uses time as the fourth dimension for the “location” of weather observations. Just like SWEET, NNEW ontology does not address the context of weather predictions. This was the reason for not using this ontology.
The Section 2 discusses no existing foundational ontology covers the domain of
numerical weather prediction data, in a way suitable for using it as a starting point for
NWPOntology. Thus, the creation of NWPOntology was based on UFO conceptual
requirements, reusing of existing meteorological ontology concepts (see Section3). The
ability to clearly express concepts of the real-world, reducing conceptual ambiguities,
leads the scientific community, as well as conceptual modeling professionals, to
consider UFO as an important resource to model the domain ontologies.
Detailed surveys on ontology development methodologies can be found in the works of
[
In the Purpose Identification and Requirements Specification phase, it was defined that NWPOntology aims to serve as a shared conceptualization to resolve the semantic heterogeneity caused by divergent interpretations of data according to the different contexts in which they are used, and the green part of the model will support an analytical environment of historical events. Its main objective is the modeling and implementation of a domain ontology for numerical weather prediction data (NWPOntology), as a solution to reduce the conceptual ambiguities of the weather forecast data existing in the various forecast centers, promoting semantic enrichment meteorological observation data. In this way, the information can be observed, analyzed and disseminated by weather reports, assisting in the decision making process.
The focus of the Capture and Formalization phase is to capture the conceptualization of the discourse universe. Relevant concepts and relationships must be identified and organized. This phase begins with conceptual modeling using highly expressive languages to create strongly axiomatized ontologies that are as close as possible to the domain's ideal ontology. The focus of these languages should be on the adequacy of the representation of the main concepts and relationships in the domain using taxonomies.
The NWPOntology fragment presented in this article is based mainly on the
taxonomy of meteorological data terms proposed in [
NWPOntology model illustrated in Figure 1, making the distinction between
different types of entities explicit in the ontological sense. The UFO-B stereotypes
present in our modeling (the green part) were adapted from [
The meteorological parameters, modeled as the <kind> stereotype, is used to represent rigid concepts, that provide an identity principle for their instances and do not require a relational dependency. They are basically divided into variables, e.g., values of relative humidity, precipitation, surface wind, air temperature and sea surface temperature, atmospheric pressure, and cloudiness. Such concepts are modeled by the <subkind> stereotype, so they have their own identities and inherit the properties of the meteorological parameters class. In this case, the data subtypes were mapped as a <subkind> sortal, precisely because they are unique, but at the same time inherit characteristics in common.
Furthermore, according to our model, they are observed by the Global Observation Systems sensors, that was modeled by the <kind> stereotype, which is mapped by the formal relationship created between the Global Observation System and the observed meteorological parameters. These data are received at the National Meteorological Services (modeled by the <kind> stereotype) and the data is transformed into a graphical visualization, through the Local Data Managers (LDM). The LDM was modeled by the <role> stereotype, because is a construct used to represent anti-rigid specializations of identity providers, in this case the Software Agent, that was modeled by the <kind> stereotype.
Meteorological data transformed into a graphical visualization is observed by the Weather Forecasters and compared with the data generated by the outputs of the Regional Numerical Models. Weather forecasters, modeled by the <role> stereotype, which has its identity principle inherit by the National Meteorological Services class. Regional Numerical Models, modeled by the <subkind> stereotype, for they have their own identities and inherit the properties of the Numerical Weather Prediction Models class, modeled by the <kind> stereotype.
Global Numerical Models, modeled by the <subkind> stereotype, for they have
their own identities and inherit the properties of the Numerical Weather Prediction
models class, input the meteorological parameters data for the Regional Numerical
models. If there are no discrepancies between the outputs of the numerical models and
the meteorological data of the LDM, the Weather Forecaster produces the Weather
Forecast, modeled by the <kind> stereotype to be disseminated as the Weather Reports,
modeled by the <datatype> stereotype. This stereotype represents data whose members
are numerical values belonging to a given set of possible values. A concept represented
as a <datatype> is a theoretical representation of a conceptual space and the restrictions
imposed by its respective geometric structures [
Severe Weather Events refers to any dangerous meteorological phenomena with
the potential to cause damage, serious social disruption, or loss of human life. Types of
Severe Weather Events vary depending on the latitude, altitude, topography and
atmospheric conditions. High winds, hail, excessive precipitation and wildfires are
forms and effects of severe weather, as are thunderstorms, downbursts, tornadoes,
waterspouts, tropical cyclones, and extratropical cyclones, and should be released
through weather reports as soon as possible to users [
Severe Weather Events consist of a complex event that brings about the victim or property to a damage situation, represented at the time (Severe Weather Event Moment), when the victim or property suffers that damage. A Severe Weather Events only occurs if it is triggered by a situation of significant change in atmospheric parameters (Severe Weather Indicators). An example of a Severe Weather Event would be a Tropical Cyclone occurring in a given location, causing material damage and death to people.
The need to reduce the conceptual ambiguities of meteorological data, which make up the Global Observation System, motivated the construction of NWPOntology, aiming at an open and generic approach to interoperability between the different systems for the correct decision making. To complete the development of the ontology it is still necessary to have a dynamic verification and validation of the behavior of the operational ontology, in a finite set of test cases against the expected behavior in relation to Competence Questions (CQs). Investigations and experiments in NWPOntology are still needed, mainly with regard to studies on UFO-B event stereotypes, as the incorporation of new entities is necessary, such as material damage caused by serious climatic events in regions of the world potentially affected. Therefore, future research can benefit from the ontological concepts identified in this research.