Numerical and Implementation Issues in Food Quality Modeling for

Human Diseases Prevention

A. Galletti

, R. Montella

1,2

, L. Marcellino

, A. Riccio

, D. Di Luccio

, A. Brizius

and I. Foster

Department of Science and Technology, University of Naples “Parthenope”, Centro Direzionale,

Isola C4, 80143, Napoli, Italy

Computation Institute, University of Chicago, 5735 South Ellis Avenue, 60637, Chicago, IL, U.S.A.

Keywords:

Lagrangian Methods, Numerical Interpolation, Food Quality, Human Diseases, Cloud Computing, High Per-

formance Computing, Scientiﬁc Workﬂow, Smart Devices, Internet of Things, Marine Data Crowdsourcing.

Abstract:

Monitoring nearshore sea water pollution using connected smart devices could be nowadays impracticable due

to the aggressive saline environment, the network availability and the maintain and calibration costs. Accurate

forecast of marine pollution is most needed to evaluate the adverse effects on coastal inhabitants’ health when

ﬁshes and mussels farming economically characterizes the local social background. In an operational context,

numerical simulations are performed routinely on a dedicated computational infrastructure producing space

and temporal high-resolution predictions of weather and marine conditions of the Bay of Naples. In this paper

we present our results in developing a community open source Lagrangian pollutant transport and dispersion

model, leveraging on hierarchical parallelism implying distributed memory, shared memory and GPGPUs.

Some numerical details are also discussed. This system has been used to develop an alarm system to help local

authorities in making decisions regarding the collection of mussels. The model setup and the simulation results

will be improved using FairWind, an under development system dedicated to coastal marine crowdsourced data

gathering and sharing, based on smart devices and Internet of Things aﬂoat.

1 INTRODUCTION

Human health can be adversely affected by pollutants

emission into sea water, specially from seafood con-

tamination caused by inshore discharges or offshore

spills in areas close to aquaculture farms.

Fish and mussel farms are critically sensitive to

coastal water quality, and thus require continuous

monitoring to enforce food security and quality and to

prevent any possible disease affecting human health.

The potentially toxic substances emitted from

point sources can, in more or less short time, reach the

mussel farms and promote, in relation to the mussel-

pollutant contact time, the bioaccumulation in ﬁlter

feeders organisms. Several studies have shown that

pathogens, such as bacteria and enteric viruses, can

be transmitted by mussels and the widespread habit

of consuming raw or slightly cooked shellﬁsh con-

tributes to maintaining the incidence of hepatitis A

cases in the southern Italy at high level (Croci et al.,

2003).

On the other hand, nevertheless the availability of

technologies for remote water quality monitoring sys-

tem using wireless sensors (Haron et al., 2009), the

livestock sampling and the microbiological spottily

analysis fails if the goal is a consistent data time series

needed by any process aimed to make inference with

human health. The use of connected smart devices

is fully feasible in a context where the farms are in a

limited environment, while the challenges rise for ﬁsh

and mussel farms in marine nearshore, but open wa-

ters: the extreme weather events, the aggressive saline

environment, the network and energy availability and,

last but not least, the need for continuous maintenance

and sensors calibration could have a negative impact

on the use of a technical solution fully based on the

Internet of Things aﬂoat approach.

To face the above depicted scenario, we de-

signed and implemented WaComM (Water Commu-

nity Model), a three dimensional Lagrangian model

enforcing the decision support system and enabling

the simulation and prediction of pollutant spills, trans-

port and dispersion in both inshore and offshore envi-

ronments (Giunta et al., 2005).

Here, we provide some details about the way in

which the input data of our Lagrangian model are ob-

Galletti A., Montella R., Marcellino L., Riccio A., Di Luccio D., Brizius A. and Foster I.

Numerical and Implementation Issues in Food Quality Modeling for Human Diseases Prevention.

DOI: 10.5220/0006297905260534

tained. Moreover, we give some insights about the

related parallel implementation.

In order to support the numerical model solu-

tion, we leverage on crowdsourced data acquired by

the system called "FairWind" a smart, cloud-enabled,

multifunctional navigation system for leisure and pro-

fessional vessels. This system has been developed as

an open technologies developing platform based on

the Android operating system for smart devices.

In this context, the system is used to collect data

from on board marine instruments such as, but not

limited to, GPS, heading, pitch, roll, yaw and speed

sensors, water temperature, depth and weather sen-

sors. Collected data are sent on the land using grid

ﬁle transfer technologies, and then processed in order

to improve data quality and consistency.

Data, made available to the community as open

data, are used to improve the model setup (i.e. high

resolution coastal digital depth model) and will be

used for model evaluation and improvement (surface

temperature, surface current).

The rest of the paper is organized as follows: sec-

tion 2 is about related work; section 3 focuses on the

numerical issues driving the design of our Lagrangian

model; section 4 is on details about the WaComM im-

plementation with focus on parallel cores; section 5 is

about the novel contribution of the crowdsourcing ap-

proach in this application research ﬁeld; section 6 de-

scribes some preliminary computational evaluations;

ﬁnally section 7 presents conclusions and future de-

velopments.

2 RELATED WORK

The Regional Ocean Modeling System (ROMS)

is a free-surface, terrain-following, primitive equa-

tions ocean model widely used by the scientiﬁc

community to characterize and simulate the meso-

and sub-mesoscale ocean and coastal water dynam-

ics (Haidvogel et al., 2000; Wilkin et al., 2005;

Shchepetkin and McWilliams, 2003; Shchepetkin and

McWilliams, 2005). In this application ROMS is

forced by a high-resolution atmospheric forecast and

provides the ﬂow velocity on a curvilinear boundary-

ﬁtted grid.

ROMS was used to deploy a real-time forecast for

the transport and deposition of water pollutants us-

ing the particle-tracking WaComM model, which im-

plements a Lagrangian technique consistent with the

advection-diffusion equation (Rodean, 1996).

In this kind of model, the dispersion phenomenon

is reproduced by imaginary numerical particles; at

each of these particles different characteristics are as-

signed, e.g. pollutant concentration and settling ve-

locity.

At each time step, the particle position is calcu-

lated on the basis of the ﬂow velocity, computed by

the hydrodynamic model ROMS, and a random jump

representing the turbulence diffusion.

GVirtuS (Montella et al., 2011) is a general-

purpose virtualization service for high performance

computing applications on cloud environments, fo-

cusing on NVidia CUDA GPGPU virtualization and

MPI based virtual clusters.

The RAPID GVirtuS incarnation (Montella et al.,

2016b) was used as GPGPU remoting provider for hi-

erarchical parallelism, sending the instruction set ker-

nel to the accelerating hardware, processing data on

the device and then sending back results to the gen-

eral purpose CPU.

The FACE-IT (Framework to Advance Climate,

Economic, and Impact Investigations with IT) project

(Pham et al., 2012) has been developed, and contin-

ues to be developed, to provide a cloud-based sci-

ence gateway for the web-based access to a range of

data projects, simulation models and analysis tools

(Montella et al., 2015). FACE-IT builds on the

Globus Galaxies platform, which has been devel-

oped over the past several years at the University of

Chicago, initially in support of the Globus Genomics

project (Madduri et al., 2015). FACE-IT is used as

main computational playground for the implementa-

tion of the WaComM Lagrangian model running as an

on-demand and routinely workﬂow (Montella et al.,

2016a).

3 DESIGN

WaComM is the evolution of the LAMP3D model.

We optimized the algorithms in order to improve

its performance on high performance computing en-

vironment adding features as restarting and shared

memory parallelization. The description of the un-

derlying mathematical model is as follows.

Pollutants are considered as inert Lagrangian par-

ticles, tracing the marine circulation without feedback

interactions with sea current ﬁelds and other particles.

Each particle is assumed to have:

• position r(t) = (x(t),y(t),z(t)) at time t;

• initial position r

= r(0) = (x

) at the initial

time t = t

= 0;

• velocity v(r(t),t) = U(r(t),t) + η(r(t),t) at time

t, where U(r(t),t) denotes the deterministic ve-

locity, and η(r(t),t) is the stochastic ﬂuctuation

arising from the Langevin equation model in or-

der to describe the Brownian motion of particles

(Rodean, 1996).

Given U(r(t),t), or an estimate of it, for each posi-

tion r(t) and at each time t, the ﬁnal particle posi-

tion r(t

k+1

) = r(t

+∆t), at time t

k+1

= t

+∆t, can be

computed by means of the equation:

r(t

+∆t) = r(t

+∆t

v(r(t),t) dt, (k = 0,1, ...),

(1)

where r(t

) is a starting position, at the starting time

, and ∆t > 0 denotes a time interval length (in Wa-

ComM we set ∆t = 1 h).

Numerical integration of (1) could be made in sev-

eral ways; in our approach we use the Eulero method

that considers a discretization of the time interval

+ ∆t] in the grid

j,k

= t

+ j · dτ, j = 0,. ..,N

where dτ = ∆t/N denotes the discretization step.

To do this, at each time τ

the evaluations of

U(r(τ

j,k

),τ

j,k

) and η(r(τ

j,k

),τ

j,k

) are required. How-

ever, these values are provided by ROMS only at

some discrete time instants, and on a discrete irregu-

lar three-dimensional grid. Such a grid can be thought

of as the set of vertexes of a ﬁnite number of poly-

hedrons V

(cells). These cells are all topologically

homeomorphic to a cube and their union is the space

domain of U.

Figure 1: Example of cell. The cells form the irregular grid

where the values of U are known.

An example of the form of such cells is in Fig-

ure 1. Notice that the irregular polyhedron is deﬁned

by assigning its eight vertexes.

Possible choices of interpolants, which do not

need any information about the mesh (i.e. the so-

called mesh-free methods) are the radial basis func-

tions methods (Cuomo et al., 2013; Fasshauer, 2007).

However, when some structure of the grid is assigned,

i.e. when a smart subdivision of the domain in geo-

metrical polyhedral structures is known, one can take

advantage of this fact and so several kinds of inter-

polants, exploiting the geometry of the cells that form

the mesh, can be deﬁned (Galletti and Maratea, 2016;

Cuomo et al., 2014a; Cuomo et al., 2014b; Cuomo

et al., 2015). In this case we choose a simple trilinear

interpolation approach using barycentric coordinates.

In particular, by referring to the notations introduced

in Figure 1, we compute velocities at any spatial lo-

cation, e.g. at particle position r and desired time, by

the linear interpolation of velocities made available

by ROMS at the vertexes of any grid cell and regular

time intervals.

In order to assign the stochastic ﬂuctuations, Wa-

ComM relies on the standard ‘K-theory’, based on

a diffusion coefﬁcient which is estimated by pre-

processed ROMS data. An exponential decay which

uses the T90 parameter (the time required to de-

grade 90% of the biodegradable matter in a given

environment) is applied to takes into account de-

caying processes. A sedimentation velocity, w

sed

(0,0, −w

sed

), is added to the deterministic component

of velocity to simulate settling particles. At the end

of each suitably chosen time interval, a scaled con-

centration ﬁeld C

i, j,k

is computed by simply counting

the number of particles found within each grid cell.

4 IMPLEMENTATION

The modeling system can be used in an ex-ante fash-

ion, as a decision support tool to aid in the selection

of the best suitable areas for farming activity deploy-

ment, or in an ex-post fashion, in order to achieve a

better management of offshore activities. We tested

the system on several case studies where pollutants

are spilled out from well known punctual sources lo-

cated along the coasts of Campania region.

As arguing from the numerical approach, the

model is computing intensive and parallelization is

needed for its usage in real-world applications. The

problem size increases with the number of emission

sources and the number of emitted particles. More-

over, the computing time is inﬂuenced by the integra-

tion time step which should be short enough to cor-

rectly represent the turbulent diffusion processes.

Although consistent results can be guaranteed us-

ing the sequential implementation of the WaComM

model, the wall-clock performance actually makes the

production unfeasible.

Hence, the growing need of on-demand results,

which involves a large computational effort for Wa-

ComM, suggests to use general purpose GPUs in or-

der to efﬁciently perform computing intensive tasks.

In particular, a GPU implementation is considered

for the WaComM main cycles involved for the inter-

polation and evaluation steps of the 3D momentum

and dispersion parameters. This concerns with a par-

allel design schema hierarchical and heterogeneous.

The implementation of the GPGPU enabled code

is realized with the NVIDIA CUDA programming

model, and using heterogeneously both the CPU and

GPU (Ortega et al., 2016) supported by an MPI based

distributed memory approach. Such an implementa-

tion considers a dynamical load balancing on the par-

ticle number.

More in detail, the distributed memory paralleliza-

tion has been introduced in the hourly inner cycle of

WaComM in order to enhance the performance. Such

a cycle computes the path of each particle and, since

no interaction between particles is assumed, each par-

ticle path can be virtually tracked independently of

the others. As explained in the design section, the

interpolation stage is time consuming. The problem

size scales with the input data (the resolution of the

momentum, the sea current 3D vector components

- U, V, W - and the vertical T-diffusion - AKT -

grid). Then, previous discussion justiﬁes even more

the shared memory parallel approach for the evalua-

tion of the interpolation model.

5 DATA CROWDSOURCING

In the ﬁeld of ocean modeling, the need for compu-

tational and storage resources is currently satisﬁed by

the availability of cloud based services that reduce the

total cost of ownership and accelerate the democra-

tization of science (Foster, 2011). Nevertheless, to

have more robust and accurate models, there is a need

for detailed, high resolution, spatio-temporal data for

initialization and validation. While data can be hard

to obtain from traditional sources, due the lack of

available public data in some coastal areas, the chal-

lenges of surveying large areas and other technical

constraints, this data can be easily obtained using in-

ternet of things based crowdsourcing tools like Fair-

Wind

(Montella et al., 2016c).

FairWind is an integrated, multifunctional, navi-

gation software based on open technologies designed

and developed by a very interdisciplinary team in or-

der to maximize the beneﬁts and the advantages. It

is a marine single board computer device leveraging

a custom-built infrastructure on top of stable and well

documented mobile and cloud technologies.

From the marine electronics point of view, the

most remarkable innovation introduced by FairWind

are the Boat Apps that extend the FairWind basic fea-

tures, integrating with already present onboard instru-

ments and straightforwardly interacting with indus-

trial or self - made internet of things based instru-

ments. The board dataset, collected by FairWind, is

a scientiﬁcally intriguing source of huge amounts of

geolocated data (big-data) about marine coastal envi-

ronment (weather and sea conditions, surface sea cur-

rents, water temperature, water depth, etc.), boat en-

gine status, boat performances (speed, heading, pitch,

roll), presence of board water and waste management,

fuel consumptions and, above all, safety at sea and

search and rescue systems (Figure 2).

Data is collected on board and, when possible,

sent to cloud storage and computing facilities using

reliable, affordable, and safe technologies such as the

Globus data transport services. Users can choose

what data to share in a named or anonymous way.

Operationally, once data is collected, it will be ana-

lyzed and processed in order to extract sensor calibra-

tion using big-data algorithms, evaluated with a qual-

ity model comparing it with data acquired by trustful

equipment and, ﬁnally, made available as open data

for ocean model initialization and/or validation.

6 EVALUATION

This section describes the WaCoMM model use case

for the Campania Region (Italy), applied to prevent

the consumption of contaminated food harvest in

mussel farms. This farms have, generally, a long lines

organization in which the mussels are attached to sub-

merged ropes hung from a back-bone supported by

large plastic ﬂoats.

In 2006, the Campania Region published the

“Guidelines for mussel farms: classiﬁcation and

monitoring system of mussels production and relay-

ing areas” to delineate the guideline to mussel farms

monitoring.

This document identiﬁed the skills to perform

the analysis provided by the Experimental Zoo-

prophylactic Institute of Southern Italy (IZSM) for

mussels samples and by the Campania Regional En-

vironmental Protection Agency (ARPAC) for water

samples. The in situ analysis also included the com-

pulsory microbiological parameters: Escherichia Coli

and Salmonella spp. In accordance with the current

European legislation (2073/2005EC), the concentra-

tion of E. Coli and Salmonella spp in mussels must be

less than 230 MPN/100g (Most Probable Number per

100 grams) and zero respectively.

The considered mussels rows (MF

in Figure 3)

are located about 500 m distance to the coast in Punta

Terone-Capo Miseno and cover an area of about 257

, with a depth of about 20 m. In this case, the

reared mussels are mostly Mytilus Galloprovincialis.

The MF

mussels are allowed to be placed on the

market for human consumption directly, without fur-

Figure 2: Smart devices in a context of internet of things aﬂoat make data crowdsourcing affordable. FairWind equipped

professional and leisure boats can contribute to model setup improvement and data assimilation.

ther treatment in a puriﬁcation centre or after relaying

(class type A). This makes the mussels quality very

important to human diseases prevention.

In some cases ﬂooding events can bring too much

pollutants to shellﬁsh farms, banning them from har-

vesting.

In this context, a forecasting system of the meteo-

rological and hydrodynamic circulation, coupled with

the WaComM model, can be a valid support to the

mussel farm management.

WaComM is included in a scientiﬁc workﬂow to

ingest the forecast input data needed to initialize the

simulation and track particles trajectories.

The Weather Research and Forecasting Model

(WRF) (Skamarock et al., 2001) simulates the

weather conditions for driving the ROMS model.

The WRF model has already been used to simulate

weather conditions on the Campania region (Barone

et al., 2000; Ascione et al., 2006). WaComM model

domain is 715×712×11 grid points (Lat

min

=40.48N,

Lat

max

=40.95N; Lon

min

=13.91E, Lon

max

=14.52E).

The pollutant sources in the Gulf of Naples are con-

sidered as points all along the coast, spilling out 100

particles for each simulation hour. Actually we used

50 particle sources. Our system has been tested com-

paring the numerical forecast and mussel microbio-

logical analysis. The simulation spanned the time

interval 07/12/2015 Z00 – 21/12/2015 Z23 and the

output was stored at a hourly interval. On days

09/12/2015 and 21/12/2015 the local authorities car-

ried out the microbiological analysis on Mytilus Gal-

loprovincialis in Punta Terone mussel farm. Results

showed a concentration of E. Coli much greater than

the legal limits in the ﬁrst day (5400 MPN/100g) and

lower in the second day (45 MPN/100g); Salmonella

spp was absent in both samples.

The mean wind direction started blowing from

NW from 7 to 9 December 2015; in this period parti-

Figure 3: Mussel farms locations in the Bay of Pozzuoli. The studied mussel farm is in Punta Terone (MF

) area.

cles spilled out by sources in the eastern part moved

towards the center, while particles emitted by sources

in the western part remained close to the coast (see

Figure 4). After, the mean wind shifted to NE and all

particles moved towards the center of the gulf with a

progressive increase in the concentration of the tracer

in the area surrounding the two mussel farms with a

maximum value on day 15/12/2015. Subsequently,

the rotating ocean currents contributed to the disper-

sion of the tracer away from the mussel farms. This

picture is also conﬁrmed by microbiological analy-

sis carried out on Mytilus Galloprovincialis mussels

(Figure 5). The comparison between numerical fore-

cast and microbiological analysis showed a remark-

able similarity in trends, although this kind of analysis

have to be performed in a more extensively fashion.

That conﬁrmed the possibility to use the system as a

decision maker tool for applications correlated with

sea quality and as a support system for experimental

observations and controls.

7 CONCLUSION

The quality of coastal marine waters depends strictly

on the impact of human activities. Urban settlements,

industries, agriculture, livestock farming and weather

conditions produce effects which, individually or to-

gether, can heavily compromise or even disrupt the

equilibrium of aquatic ecosystems.

In this paper we presented our research efforts in

designing and developing WaComM, a community

water quality model, with the main aim, but not lim-

ited to, to develop a forecast system and perform op-

erational numerical predictions in the context of mus-

sel farms management, in order to prevent E. Coli

and Salmonella human diseases with a strong effort

in data dissemination for local management decision

support (Montella et al., 2007).

WaComM is under continuous active develop-

ment.

From the implementation point of view a deep

refactoring is needed in order to better exploit the hi-

erarchical parallelism. The current implementation

leverages on a naive application designed load bal-

ancing (Laccetti et al., 2013). A Portable, Extensible

Toolkit for Scientiﬁc Computation (PETSc) approach

could enhance the overall perforance and, above all,

the problem scalability (Carracciuolo et al., 2015).

A short-term goal of our project is to extend the

studied area to the whole coasts of the Campania

Region in order to promote its use as an effective

tool dedicated to improve the management of coastal

farms. In order to achieve this target, we need to im-

prove the robustness of the WaComM model and the

scalability of the ofﬂine coupling system.

From the scientiﬁc point of view, we will en-

hance the simulation quality with data collected us-

ing the FairWind technology as depicted above im-

proving the data acquisition from boat sensor as inter-

connected smart devices using the Internet of Things

aﬂoat technologies. This issue is a source of novelty

even because the whole FairWind ecosystem is based

on the SignalK marine data interchange open proto-

col (http://signalk.org). The proposed system is ex-

tensible in order to collect data from other sensors

as, but not limited to, surface ph and salinity sen-

sors that could improve the simulation quality and the

model validation process and, ﬁnally, if supported by

Figure 4: Sea surface currents (vectors) and pollutants concentration (red=high; yellow=medium; green=low; blue=very low;

white=absent) in in Gulf of Pozzuoli (Campania Region, Italy) in days 08/12/2015 Z12 (ﬁgure A), 09/12/2015 Z12 (ﬁgure

B), 15/12/2015 Z12 (ﬁgure C) and 20/12/2015 Z23 (ﬁgure D). The red dotted line is the area of Study with mussel farms in

Punta Terone (number 1 in ﬁgure A) area.

a ground true based on better microbiological analy-

sis and consistent epidemiological studies on mussels

originated enterogastric diseases (Suffredini et al.,

2014), enhance the overall system trustability.

Figure 5: Forecasted averaged particles concentration timeseries in the study area.

ACKNOWLEDGEMENTS

This work is supported in part by the Euro-

pean Union Grant Agreement number 644312-

RAPID–H2020-ICT-2014/H2020-ICT-2014-1 “Het-

erogeneous Secure Multi-level Remote Acceleration

Service for Low-Power Integrated Systems and De-

vices”, using GVirtuS open source software compo-

nents; project “Development of a software for model-

ing pollution events based on high-performance and

distributed computing technologies with advanced

web interface”; project “MytiluSE - Modelling

mytilus farming System with Enhanced web tech-

nologies”’; NSF cyberSEES program award ACI-

1331782; NSF Decision Making Under Uncertainty

program award 0951576; DOE contract DE-AC02-

06CH11357. IBM-P8 resources were generously pro-

vided through an agreement between IBM and the

University of Naples “Parthenope” - Department of

Science and Technology - High Performance Scien-

tiﬁc Computing Smart-Lab.

REFERENCES

Ascione, I., Giunta, G., Mariani, P., Montella, R., and Ric-

cio, A. (2006). A grid computing based virtual lab-

oratory for environmental simulations. In European

Conference on Parallel Processing, pages 1085–1094.

Springer.

Barone, G., D’Ambra, P., di Seraﬁno, D., Giunta, G., Murli,

A., and Riccio, A. (2000). Application of a paral-

lel photochemical air quality model to the campania

region (southern italy). Environmental Modelling &

Software, 15(6):503–511.

Carracciuolo, L., D’Amore, L., and Mele, V. (2015). To-

ward a fully parallel multigrid in time algorithm in

petsc environment: A case study in ocean models. In

High Performance Computing & Simulation (HPCS),

2015 International Conference on, pages 595–598.

IEEE.

Croci, L., De Medici, D., Ciccozzi, M., Di Pasquale, S.,

Suffredini, E., and Toti, L. (2003). Contamination of

mussels by hepatitis a virus: a public-health problem

in southern italy. Food control, 14(8):559–563.

Cuomo, S., Galletti, A., Giunta, G., and Marcellino, L.

(2014a). A class of piecewise interpolating functions

based on barycentric coordinates. Ricerche di Matem-

atica, 63(11):87–102.

Cuomo, S., Galletti, A., Giunta, G., and Marcellino, L.

(2014b). A novel triangle-based method for scattered

data interpolation. Applied Mathematical Sciences,

8(134):6717–6724.

Cuomo, S., Galletti, A., Giunta, G., and Marcellino, L.

(2015). Piecewise hermite interpolation via barycen-

tric coordinates. Ricerche di Matematica, 64(2):303–

319.

Cuomo, S., Galletti, A., Giunta, G., and Starace, A. (2013).

Surface reconstruction from scattered point via rbf in-

terpolation on gpu. In Computer Science and Informa-

tion Systems (FedCSIS), 2013 Federated Conference

on, pages 433–440. IEEE.

Fasshauer, G. E. (2007). Meshfree approximation methods

with MATLAB, volume 6. World Scientiﬁc.

Foster, I. (2011). Globus Online: Accelerating and democ-

ratizing science through cloud-based services. IEEE

Internet Computing, 15(3):70.

Galletti, A. and Maratea, A. (2016). A bound for the accu-

racy of sensors acquiring compositional data. Proce-

dia Computer Science, 98:485 – 490. The 7th Interna-

tional Conference on Emerging Ubiquitous Systems

and Pervasive Networks (EUSPN 2016)/The 6th In-

ternational Conference on Current and Future Trends

of Information and Communication Technologies in

Healthcare (ICTH-2016)/Afﬁliated Workshops.

Giunta, G., Montella, R., Mariani, P., and Riccio, A. (2005).

Modeling and computational issues for air/water qual-

ity problems: A grid computing approach. Nuovo Ci-

mento C Geophysics Space Physics C, 28:215.

Haidvogel, D. B., Arango, H. G., Hedstrom, K., Beck-

mann, A., Malanotte-Rizzoli, P., and Shchepetkin,

A. F. (2000). Model evaluation experiments in the

north atlantic basin: simulations in nonlinear terrain-

following coordinates. Dynamics of Atmospheres and

Oceans, 32(3):239–281.

Haron, N. S., Mahamad, M. K., Aziz, I. A., and Mehat, M.

(2009). Remote water quality monitoring system us-

ing wireless sensors. In Proceedings of the 8th WSEAS

International Conference on Electronics, Hardware,

Wireless and Optical Communications (EHAC’09),

Cambridge, UK, pages 148–154.

Laccetti, G., Lapegna, M., Mele, V., and Romano, D.

(2013). A study on adaptive algorithms for numer-

ical quadrature on heterogeneous gpu and multicore

based systems. In International Conference on Paral-

lel Processing and Applied Mathematics, pages 704–

713. Springer.

Madduri, R., Chard, K., Chard, R., Lacinski, L., Rodriguez,

A., Sulakhe, D., Kelly, D., Dave, U., and Foster,

I. (2015). The Globus Galaxies platform: Deliver-

ing science gateways as a service. Concurrency and

Computation: Practice and Experience, 27(16):4344–

4360. CPE-15-0040.

Montella, R., Coviello, G., Giunta, G., Laccetti, G., Isaila,

F., and Blas, J. G. (2011). A general-purpose vir-

tualization service for HPC on cloud computing: an

application to GPUs. In International Conference on

Parallel Processing and Applied Mathematics, pages

740–749. Springer.

Montella, R., Di Luccio, D., Troiano, P., Brizius, A.,

and Foster, I. (2016a). WaComM: A parallel Water

quality Community Model for pollutant transport and

dispersion operational predictions. In International

Workshop on Numerical Algorithms and Methods for

Data Analysis and Classiﬁcation, 2016. NAMDAC’16.

IEEE. IEEE.

Montella, R., Ferraro, C., Kosta, S., Pelliccia, V., and

Giunta, G. (2016b). Enabling android-based devices

to high-end gpgpus. In Algorithms and Architectures

for Parallel Processing, pages 118–125. Springer.

Montella, R., Giunta, G., and Riccio, A. (2007). Using grid

computing based components in on demand environ-

mental data delivery. In Proceedings of the second

workshop on Use of P2P, GRID and agents for the de-

velopment of content networks, pages 81–86. ACM.

Montella, R., Kelly, D., Xiong, W., Brizius, A., Elliott, J.,

Madduri, R., Maheshwari, K., Porter, C., Vilter, P.,

Wilde, M., et al. (2015). FACE-IT: A science gateway

for food security research. Concurrency and Compu-

tation: Practice and Experience, 27(16):4423–4436.

Montella, R., Luccio, D. D., Ferraro, C., Izzo, F., Troiano,

P., and Giunta, G. (2016c). Fairwind: a marine

data crowdsourcing platform based on internet of

things and mobile/cloud computing technologies. In

8th International Workshop on Modeling the Ocean

(IWMO), Bologna, Italy.

Ortega, G., Puertas, A., de Las Nieves, F. J., and Martin-

Garzón, E. (2016). Gpu computing to speed-up the

resolution of microrheology models. In Algorithms

and Architectures for Parallel Processing, pages 457–

466. Springer.

Pham, Q., Malik, T., Foster, I. T., Di Lauro, R., and Mon-

tella, R. (2012). SOLE: Linking Research Papers with

Science Objects. In IPAW, pages 203–208. Springer.

Rodean, H. C. (1996). Stochastic Lagrangian models of

turbulent diffusion, volume 45. Springer.

Shchepetkin, A. F. and McWilliams, J. C. (2003). A method

for computing horizontal pressure-gradient force in an

oceanic model with a nonaligned vertical coordinate.

Journal of Geophysical Research: Oceans, 108(C3).

Shchepetkin, A. F. and McWilliams, J. C. (2005).

"the regional oceanic modeling system (roms):

a split-explicit, free-surface, topography-following-

coordinate oceanic model". Ocean Modelling,

9(4):347 – 404.

Skamarock, W. C., Klemp, J. B., and Dudhia, J. (2001).

Prototypes for the wrf (weather research and fore-

casting) model. In Preprints, Ninth Conf. Mesoscale

Processes, J11–J15, Amer. Meteorol. Soc., Fort Laud-

erdale, FL.

Suffredini, E., Lanni, L., Arcangeli, G., Pepe, T., Mazzette,

R., Ciccaglioni, G., and Croci, L. (2014). Qualitative

and quantitative assessment of viral contamination in

bivalve molluscs harvested in italy. International jour-

nal of food microbiology, 184:21–26.

Wilkin, J. L., Arango, H. G., Haidvogel, D. B., Lichten-

walner, C., Glenn, S. M., and Hedström, K. S. (2005).

A regional ocean modeling system for the long-term

ecosystem observatory. Journal of Geophysical Re-

search: Oceans, 110(C6).