Page 438 - Economia Azul - versão: inglês

years of use increased beyond the target available, such as ice thickness and concen- data assimilation methods extract the best the initial condition. The smaller the initial
value of 0.01 psu, reaching more than 0.05 tration, carbon dioxide concentration and information from these sources and com- condition error, the greater the probabili-
psu (K. Sato, 8th OceanPredict OS-Eval-TT pH. In ecosystem models, in addition to all bine them to produce the valuable objec- ty of producing accurate predictions (e.g.,
Meeting, April 28, 2021). This fact moved these quantities, nutrients, chlorophyll, and tive analyses, which are employed both for DERBER and ROSATI, 1989; KALNAY, 2003;
the scientific community to remove data a myriad of parameters of the trophic chain scientific studies and for the practice of op- EVENSEN, 2006; DAVIDSON , 2019).
bias and identify the impacts of these data must be considered. The models therefore erational ocean forecasting. Observing the right scales for
on the scientific results derived from them. offer a spatio-temporal representation of Data assimilation assimilation
In addition to the instrumentation error, oceanic processes, the biogeochemical cy-
there is also the sampling error, associated cle and ecosystems that serve or can serve In order to extract the maximum infor- The ability of assimilation to constrain
with the fact that it is often not possible to diverse applications, including the conser- mation from the observed data and reduce or correct the numerical model solution in
collect data with the frequency necessary vation of the marine environment, and a the errors of numerical models, data assim- the direction of observations is limited by
to characterize the variability of the natural wide range of users of oceanographic in- ilation methods are used. They improve the the quality and quantity of observations.
phenomenon. Some frequencies of variabili- formation, specialists or not. However, the representation of circulation and the physi- In relatively low-resolution global models,
ty obtained from the observed data are erro- models also have limitations. cal state of the system of interest produced with grid spacing of 20 km or greater, the
neous projections of the natural frequency. In order to obtain a solution, nonlinear by numerical models by correcting the mod- global SST and SSH fields available today
Therefore, great care must always be taken numerical models undergo a series of ap- el fields in the direction of observations and with resolutions, respectively, of approxi-
when using observed data to study process- proximations and have systematic or ran- extrapolating observational information into mately 5 km and 25 km, together with the
es, as both instrumentation and representa- dom errors in all their quantities. The mod- model space (DALEY, 1991; KALNAY et al., T/S profiles obtained by the Argo profilers
tiveness errors have to be considered. el equations, after being approximated 1996; EVENSEN, 1996; EVENSEN, 2003). and other equipment, offer good quali-
with numerical methods and mathemat- The assimilation methods optimally or sub ty and quantity to effectively correct the
Models
ical approximations, still face difficulties optimally combine in a mathematical sense, models. However, when considering re-
Numerical models of the oceans, as well with the spatial and temporal resolutions model fields with observed data and produce gional or even global models with high res-
as those of the atmosphere and the Earth used to discretize the space-time, consid- new fields, objective analyses, with smaller olution (2 km or smaller), the observational
system, are fundamental tools for society ering that the increase in resolution impos- errors than those of the models. They con- systems available today are no longer ef-
today. They reflect much of the scientific es more computational costs, not only in sider observed data errors and model errors, fective. This challenge requires the design
knowledge gained to date in oceanogra- processing but also in in the storage of the observed data and model fields that will be and implementation of new high-resolu-
phy and meteorology considering nonlin- outputs. This prevents models from solving corrected. When model errors are large rela- tion observational systems to meet a new
ear equations that translate momentum, processes with small spatial scale and/or tive to observations errors, the analysis relies stage in the evolution of predictive systems
mass, and energy balances (FOX-KEMPER high frequency. Another source of model more on observations. When the opposite for features with high frequency variability
et al., 2019). Unlike observations, model errors is the so-called physical parameter- happens, the analysis relies more on the of at spatial scales of less than 2 km, in the
fields are produced across the three-di- izations, which are not universal, that is, the model fields. so-called submesoscale.
mensional domain of interest, which can they often do not work in the same way Despite the diverse and relevant ap- The international scientific community
be global or regional. These fields include in different regions and are built to provide plications, the use of analysis as an initial involved in operational oceanography
fluxes of heat, salt, and momentum, in relationships between quantities in specific condition of the forecasting system was
addition to the primitive quantities of sea ranges of values. Finally, there are sources and still is the main motivation for the de- Considering the need to develop op-
surface height, velocity, temperature, sa- of model errors associated with boundary velopment of data assimilation methods in erational oceanography, in 1998 a major
linity, and density or pressure, in the case conditions, whether in bathymetry, in the meteorology and oceanography. Like the international effort was initiated through
of a purely ocean circulation model. If the kinematic conditions of the boundaries, atmosphere, oceans, and climate and Earth the Global Ocean Data Assimilation Ex-
model includes sea ice and biogeochemical or in atmospheric forcing. Considering the systems are chaotic, and part of the system periment (GODAE) project (CHASSIGNET
cycle quantities, other quantities are also limitations of models and observations, predictability depends on the quality of and VERRON, 2006; BELL et al., 2009).

436 BLUE ECONOMIY From Observation to Data Use 437