Journal of Operational Oceanography

ISSN: 1755-876X (Print) 1755-8778 (Online) Journal homepage: https://www.tandfonline.com/loi/tjoo20

Modelling of marine ecosystem in regional

scale for short term prediction of satellite-aided
operational fishery advisories

Kunal Chakraborty, Sourav Maity, Aneesh A. Lotliker, Alakes Samanta,

Jayashree Ghosh, Nagaraja Kumar Masuluri, Naga Swetha & Rose P. Bright

To cite this article: Kunal Chakraborty, Sourav Maity, Aneesh A. Lotliker, Alakes Samanta,
Jayashree Ghosh, Nagaraja Kumar Masuluri, Naga Swetha & Rose P. Bright (2019): Modelling of
marine ecosystem in regional scale for short term prediction of satellite-aided operational fishery
advisories, Journal of Operational Oceanography, DOI: 10.1080/1755876X.2019.1574951

To link to this article: https://doi.org/10.1080/1755876X.2019.1574951

Published online: 11 Feb 2019.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tjoo20
JOURNAL OF OPERATIONAL OCEANOGRAPHY
https://doi.org/10.1080/1755876X.2019.1574951

Modelling of marine ecosystem in regional scale for short term prediction of

satellite-aided operational fishery advisories
Kunal Chakraborty , Sourav Maity , Aneesh A. Lotliker , Alakes Samanta , Jayashree Ghosh ,
Nagaraja Kumar Masuluri , Naga Swetha and Rose P. Bright
ESSO-Indian National Centre for Ocean Information Services, Hyderabad, India

ABSTRACT ARTICLE HISTORY

The operational Potential Fishing Zone (PFZ) advisory generated and disseminated by the ESSO- Received 25 March 2018
Indian National Centre for Ocean Information Services has a significant impact on the livelihood Accepted 22 January 2019
of coastal community of India. PFZs are identified as the relatively narrow zones in the ocean
KEYWORDS
where horizontal gradients of physical and/or biological properties are enhanced. The advisories PFZ; ROMS; satellite image;
are provided to fishermen on a daily basis using remotely sensed sea surface temperature (SST) SST; chlorophyll-a; SATCORE
and chlorophyll-a (Chl-a) data from NOAA-AVHRR and MODIS-AQUA and/or Oceansat-2 satellites,
respectively. Sometimes it becomes a major challenge to retrieve SST/Chl-a data from satellite
images, particularly during the extensive cloud coverage. To overcome this operational difficulty,
the satellite data is replaced by a coupled physical-biogeochemical model data capable of
simulating ocean features leading to PFZs. The use of model data provides an additional
advantage towards transforming the existing service from advisories to forecast. The average
length of PFZs identified from satellite (model) data (2010–2016) for off Gujarat is 27.80 ± 7.2 km
(33.07 ± 3.2 km) whereas for off Andhra Pradesh, it is 28.27 ± 10.9 km (52.48 ± 8.7 km).
Considering the capability of the model in identifying PFZs, the existing advisory service can be
transitioned into a short term PFZ forecast.

Introduction
1984; Arnonn, 1987; Narain et al. 1990; Nath et al.
More than 30 million people reside along the coastlines of 1991; Solanki et al. 2003, 2005a, 2005b; Wall et al.
India, and their life is related to the ocean in one way or the 2008; Miller, 2009; Solanki et al. 2015). In India, the con-
other. The coastal fisheries in India play an important role certed efforts of oceanographers along with remote sen-
in the national economy. Almost seven million people liv- sing specialists and fishery scientists have resulted in a
ing along the coastline spanning over 8100 km are directly unique service of PFZ advisory (http://www.incois.gov.
dependent on fishing for their livelihood. Locating fishery in/MarineFisheries/PfzAdvisory). Utilising the remotely
resources often ends up in spending considerable time and sensed SST and Chl-a data, retrieved respectively from
resources, thus increasing the cost leading to low profit- NOAA-AVHRR and Oceansat-2 and/or MODIS-Aqua
ability. A reliable and timely advisory/forecast on the satellites, the ESSO-Indian National Centre for Ocean
potential fishing zone (PFZ) has multifold benefits for Information Services (INCOIS), provides PFZ advisories
the fishing community such as reduction in search time, to the fishermen on a daily basis, in an operational mode,
savings valuable fuel and human effort, thus improving except during fishing ban periods. The advisories are
the profitability and hence, the socio-economic status of generated using the methodology described by Solanki
the fishermen. In addition, advisory/forecast is useful in et al. (2005a, 2005b).
the reduction of CO2 emissions due to burning of fossil The major limitation in generation of satellite-based
oils. There is also a need to find a solution for the sustain- PFZ advisories is non-availability of data due to cloud
able long-term economic utilisation of these resources by cover. In India, the monsoon period coincides with the
maintaining the exploitable fish stocks through proper peak fishing season thereby rationalising a need for
regulatory measures. PFZ advisories during cloudy days. Also, the present
The linkages between satellite-derived sea surface PFZ advisories are derived based on same day satellite
temperature (SST) and chlorophyll-a (Chl-a) with fish images which are valid only for the next 24 h. As the
aggregation are established by a number of researchers resources are depleting in nearshore regions, it is essen-
(Lasker et al. 1981; Fiedler et al. 1984; Laurs et al., tial to encourage the fishermen community to carry out

CONTACT Kunal Chakraborty kunal.c@incois.gov.in, kcincoishyb@gmail.com

© 2019 Institute of Marine Engineering, Science & Technology
2 K. CHAKRABORTY ET AL.

the pelagic fishing activities in deep seas which require and the same maps were used for the end use. After
multi-day fishing. Therefore, accurate forecasting of the the establishment of ESSO-INCOIS in 1999, PFZ was
PFZ advisories for next 3–5 days is necessary. To over- taken up as a mission to provide uninterrupted
come the operational difficulties in generating PFZ advi- advisories on an operational basis to the fishermen com-
sories due to non-availability of data and transforming munity of India. With the launch of the IRS-P4 (Ocean-
PFZ advisories into PFZ forecast, a coupled physical-bio- sat-1), in May 1999, Space Applications Centre (SAC)
geochemical model in regional scale is developed which developed the techniques to generate the PFZ advisories
is capable of adequately simulating the ocean biogeo- using SST and Chl-a data. Subsequently, the technology
chemical state of the marine ecosystem. This approach was transferred to ESSO-INCOIS in late 2001 (Solanki
not only ensures that there is no data gap in either SST et al. 2001a, 2001b, 2003, 2005a, 2005b, Dwivedi et al.
or Chl-a but also makes data available in forecasting 2005; Nayak et al. 2007).
mode. In late 2002, ESSO-INCOIS created a digital database
The modeling framework involves an online coupling for coastline, bathymetry, major landing centres, and
of the Regional Ocean Modeling System (ROMS) phy- lighthouses along the coastal states from the National
sics/dynamics integrated with an ecosystem model at Hydrographic Office (NHO) maps. New base maps
high resolution. ESSO-INCOIS executed a programme were created using these digital datasets and were used
for long-term in situ measurements of environmental to overlay the PFZ advisories. With the availability of
parameters in Indian coastal waters. The time-series digital data and the Geographical information System
stations, four stations each along the east and west (GIS) tools, the frequency of the PFZ advisories was
coast of India, established under SATellite Coastal and increased to alternate days. In the middle of 2007, SAC
Oceanographic REsearch (SATCORE) programme developed a technique to estimate the shifting of the
yield sustained in situ observations from 2008 to 2017. PFZ features due to winds prevailing over the ocean sur-
The coupled physical-biogeochemical model is validated face, which was subsequently operationalized at ESSO-
against a wide range of observational data including the INCOIS. During late 2011, ESSO-INCOIS started using
in-situ data collected under SATCORE project and has the SST and Chl-a data from multiple satellites and opti-
demonstrated considerable capability in reproducing mally interpolated SST data from GHRSST project
marine ecosystem dynamics at synoptic and seasonal enhancing the frequency of PFZ advisories from every
time-scales in the Indian Ocean (Chakraborty et al. two days to daily (Kumar et al. 2012).
2016, 2017, 2018). In mid-2012, ESSO-INCOIS enhanced the generation
The objective of the study is to demonstrate the effec- of the PFZ advisories from 267 fish landing centers to 587
tive use of the model to overcome the operational fish landing centers. In 2013, ESSO-INCOIS incorporated
difficulties in generating PFZ advisories and further the magnitude and direction of the surface currents in the
transforming into short term forecast. The existing oper- PFZ Maps. The entire process of identification of PFZ
ational frontal detection system and the associate areas from SST and Chl-a data and the generation of
changes incorporated into frontal detection system are PFZ advisories were completely automated in 2013 redu-
described in the material and methods section. The cing the processing time from a few hours to 15 min. The
results and discussion section includes the capability of utility of PFZ advisory was enhanced by dissemination in
the model in simulating temperature and Chl-a in the native languages spoken in the different coastal states.
Indian coastal waters, potential use of model in generat- The entire coastline of India was divided into 14 sectors
ing PFZ advisories. The competitive advantage of model namely – Gujarat, Maharashtra, Goa, Karnataka, Kerala,
over satellite in identifying persistent occurrence of PFZs Lakshadweep Islands, South Tamil Nadu, North Tamil
is also demonstrated. Nadu, South Andhra Pradesh, North Andhra Pradesh,
West Bengal, Odisha, Andaman Islands and Nicobar
Islands (Figure 1). The sector-wise PFZ maps integrated
2. Satellite aided marine fishery advisory
with coastline, landing centers, bathymetry contour,
services in the India and associated benefits
and surface current are disseminated operationally on a
The generation of PFZ advisories started at National daily basis.
Remote Sensing Agency (NRSA) during 1996–1997 A study conducted at Raigad district of Maharashtra,
using SST derived from NOAA-AVHRR. Having limit- India during 2013–2014 has shown that the use of PFZ
ation in the spatio-temporal coverage, three days of sat- advisories with 15% adoption level saved 9,00,000 litres
ellite data was required to cover the entire Indian coast, of diesel and a reduction of carbon emission by 2412
limiting the service twice a week. The features were tonnes (Singh and Singh, 2016) whereas a case study
identified on the base map using analog techniques conducted with 32 fishing boats at the same place has
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 3

Figure 1. Schematic diagram of the study area showing PFZ boxes (the entire coastline of India is divided into 14 sectors denoted using
blue rectangle), domain of front maps presented in this paper (thick black boxes) and time-series stations established under SATCORE
( ) project. The red colour two big size boxes are used to show the domain of high resolution coastal physical-biogeochemical
models configured at the ESSO-INCOIS using ROMS.

shown that 70,000 litres of diesel can be saved and a an increasing acceptability of PFZ advisories among
reduction of 1,50,000 kgs of CO2 can be achieved in a them (Surendran and Velvizhi, 2011; Suvitha and Veda-
month through the use of PFZ Advisories (NAIP, valli, 2013; Vedavalli and Suvitha, 2013; Vedavalli and
2012). Another case study conducted at Kerala, India Velvizhi, 2015).
during 2008–2011 has shown that the use of PFZ advi-
sories resulted in enormous saving of diesel that varied
from 21.47 l to 1293.53 l for every tonne of fish catch 3. Material and methods
and a reduction of CO2 emission from 0.06 t to 3.45 t
3.1. Study area
(Kumar et al. 2018). The annual net economic benefits
due to the scientific identification of PFZs based on sat- The present study is focused on dynamic shelf-break
ellite information is estimated to lie in the range of Rs. regions in the northeastern Arabian Sea (off Gujarat;
34,000 to Rs. 50,000 crores (Venkatesan and Joshi, 67.14°E–73.54°E, 18.01°N–22.28°N) and northwestern
2010). A rapid change in the fishing practices is wit- Bay of Bengal (off Andhra Pradesh; 82.10°E–89.09°E,
nessed in Gilakaladindi village of Krishna district, 15.70°N–20.46°N). The northeastern Arabian Sea is
Andhra Pradesh, India where 50 fishing boats are con- influenced by variable physical processes associated
verted from bottom trawling to gillnetting. The avail- with the seasonal reversal of monsoonal winds. The
ability of PFZ advisories on a regular basis and, cool and dry continental air is carried over the region
subsequent adoption by the fishermen make it possible during winter monsoon (December to February) by the
to bring a change in fishing operations (Vedavalli et al. prevailing northeast trade winds thereby intensifying
2014). In addition to that, various independent studies evaporation and leading to surface cooling and convec-
are carried out by the M S Swaminathan Research Foun- tive mixing. This results in intrusion of subsurface nutri-
dation (MSSRF) on the impact of PFZ services on the ent-rich water to surface leading to higher primary
livelihood of the fishermen community and reported productivity (Kumar and Prasad, 1996; Madhupratap
4 K. CHAKRABORTY ET AL.

et al., 1996). The coastal upwelling driven by the south- The coastal models are configured with a goal to study
west winds stimulates high productivity along the north- the coastal processes along the Indian coastal waters. The
eastern Arabian Sea coast and shelf during the summer primary objective is to set up models, which can trans-
monsoon (June to September). On the other hand, the form the existing PFZ advisory into PFZ forecast. For
productivity in the northwestern Bay of Bengal is this purpose, it is necessary to make available short
influenced by the coastal upwelling and with the term forecast of the biogeochemical state along with
presence of a significant number of eddies. The strong physical state of the ocean. At ESSO-INCOIS, our oper-
upwelling occurs along the east coast of India (the wes- ational ocean general circulation forecasting model is
tern Bay of Bengal) during March to September and its forced by 6 hourly atmospheric fields available from
intensity decreases from south to north due to heavy National Centre for Medium Range Weather Forecast
river runoff in the northern Bay (Shetye et al., 1991). (NCMRWF). NCMRWF developed a data assimilated
The cold core eddies are considered an important Global Forecast Systems (GFS) at a horizontal resolution
source of nutrients that promote phytoplankton blooms of 0.25 (Rajagopal et al., 2012). This is made available to
in the Bay of Bengal (Kumar et al. 2002). The avail- ESSO-INCOIS to force the ocean model. Six-hourly ana-
ability of solar radiation, which is relatively lower lysed atmospheric forcing fields obtained from the
during summer due to large cloud cover than winter, NCMRWF are used to force ROMS-1/48° coastal models
also plays a major role in the primary production nested online with ROMS-1/12° model for a period of
in the western Bay of Bengal (Gomes et al., 2000; 01-January-2010–31-December-2016. Atmospheric for-
Madhupratap et al., 2003). cing component includes air pressure at 2 m, air temp-
erature at 2 m, net shortwave and longwave flux, rain
fall rate and wind velocity. Surface heat and momentum
3.2. Ecosystem model
fluxes are internally calculated by ROMS using the bulk
A series of high-resolution physical-biogeochemical parameterizations (Liu et al., 1979; Fairall et al., 1996a,
models have been configured at ESSO-INCOIS under a 1996b). Daily averaged output is stored for the analysis.
project titled, High-resolution Operational Ocean Fore-
cast and reanalysis System (HOOFS), which integrates
3.3. Sustained observational network
ocean simulation, observation and analysis to study the
role ocean physics plays in marine environmental health A sustained observation network of total eight-time
and ecosystem functioning in the coastal ocean and adja- series sampling stations (off Okha, Goa, Mangalore and
cent deep sea. The modeling framework involves an Kochi along the west coast, and off Parangipettai, Chen-
online coupling of the Regional Ocean Modeling System nai, Visakhapatnam, and Gopalpur along the east coast)
(ROMS; Haidvogel et al. 2008) physics/dynamics inte- were established by the ESSO-INCOIS in its SATCORE
grated with an ecosystem model (Fennel et al. 2006; programme (Figure 1). SATCORE programme was
2008). A series of high resolution models include two initiated with an objective of long-term continuous
very high resolution physical-biogeochemical models measurements of bio-optical and water quality par-
with horizontal resolution 1/48° (ROMS-1/48°; approxi- ameters in the Indian coastal waters. The principal objec-
mately 2.25 km spatially averaged) exclusively for the tive of these measurements (physical, optical and bio-
east and west coast of India, and a high resolution phys- geo-chemical) was validation, improvement and devel-
ical-biogeochemical model with horizontal resolution opment of bio-optical algorithms. Regular in situ
1/12° (ROMS-1/12°; approximately 9 km spatially aver- sampling and analysis at time series stations were carried
aged) for the entire Indian Ocean basin. ROMS-1/48° out during 2008 to 2017. To ensure the accuracy of the
models viz. the east coast model (77°E–99°E, 04°N–23° parameters measured at various laboratories at different
N; red colour box in Figure 1) and the west coast time-series locations, SATCORE Inter-comparison Exer-
model (65°E–77.5°E, 08°N–26°N, red colour box in cises (SICOME) were also conducted during 2014 and
Figure 1) are nested with ROMS-1/12° model (30°S– 2015 (Lotliker et al., 2016).
30°N; 30°E–120°E). The boundary conditions for
ROMS-1/48° models are prescribed from ROMS-1/12°
3.4. Operational satellite data processing chain
model. All the regional scale models use 40 vertical levels
in a terrain-following s-coordinate system. The detail As a part of the Chlorophyll Global Integrated Network
description of ROMS-1/12° model is available in (ChloroGIN) Project, endorsed by the Global Ocean
Chakraborty et al. (2018). ROMS-1/48° coastal models Observing System in the Indian Ocean (IOGOOS),
are configured in a similar way as described in ESSO-INCOIS is providing ocean colour data products
Chakraborty et al. (2018). to users at near real-time (NRT) for research and
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 5

operational activities. An Automatic Data Processing The Oceansat-2 OCM is mainly designed to keep the
Chain (ADPC) has been setup for acquisition, processing continuity to the Oceansat-1 OCM instrument and to
and dissemination of ocean colour data products from obtain quantitative information of ocean-colour variables,
MODIS-Aqua using Ocean Colour processor (OCSSW) e.g. Chl-a concentration, vertical diffuse attenuation of the
and Graph Processing Toolkit (GPT) of SeaDAS v7.4. light, (Kd) and total suspended matter (TSM) concen-
In addition to standard products (https://oceancolor. tration in coastal waters (Figure 2).
gsfc.nasa.gov/products/), value-added products such as
Total Suspended Matter (TSM), Bloom Indices (BI)
3.5. Detection of oceanic fronts
(Ahn and Shanmugam, 2006), Quasi-true colour compo-
site (BRS) and phytoplankton types (Dwivedi et al., ESSO-INCOIS is providing advisories on Potential Fish-
2015) are also generated. ing Zones (PFZ) to the fishermen community on a daily
In addition to the Chl-a images available from ocean basis utilising remotely sensed SST and Chl-a. In the pre-
colour data products as described above, ESSO-INCOIS sent setup, SST is acquired from NOAA-AVHRR (spatial
has also setup ground station for real-time acquisition resolution: 1.1 km) and Chl-a from either Oceansat-2
and processing of data from Oceansat-2 Ocean Colour OCM (spatial resolution: 360 m) or MODIS-Aqua
Monitor (OCM) satellite and makes use of Chl-a images (spatial resolution: 1 km). It is possible that either or
with 360 m spatial resolution available from Oceansat-2 both the ocean colour satellite data may not have
satellite for operational generation of PFZs. Ocean-colour sufficient spatial coverage due to cloud cover or algor-
data retrieved from the first Indian ocean observation sat- ithm failure. In that case, Chl-a data is obtained by
ellite Oceansat-1 was extensively used for various societal blending these two products.
and scientific applications like Potential Fishing Zone The biological productivity is considered to be higher
(PFZ) identification, estimation of primary productivity, in regions where strong SST fronts are observed. These
algal bloom detection and studying the coastal processes. are the regions where SST gradients are large and rapid
Oceansat-2 spacecraft of Indian Space Research Organiz- change is taking place in SST. The high resolution infra-
ation (ISRO) is the second satellite in ocean series, red daily SST data are used to identify such regions.
which was successfully launched on September 23, 2009. Apart from this, the chlorophyll data which is sensed

Figure 2. Schematic diagram of the existing PFZ Mission, describes the operational satellite data processing chain, PFZ advisory gen-
eration, validation and dissemination including feedback collection and user interaction/awareness.
6 K. CHAKRABORTY ET AL.

by the satellite as ocean colour is used as a direct marker Table 2. The Canny edge detection parameters used in this study
of biological productivity. Regions in which SST gradi- are listed below.
ents occur along with a higher chlorophyll concentration Arabian sea Bay of Bengal

are considered to be strong potential for fishing. Again, Canny

threshold
Canny
Sigma
Canny
threshold
Canny
Sigma
the advection of the fronts by the large scale flow leads Model [0.75 0.78] 3 [0.65 0.70] 3
to increased productivity in and around the frontal Satellite
regime. ESSO-INCOIS relies on the above principle for
generation of PFZ advisories. on the logarithm (base 10) of the input Chl-a data to
The existing operational mechanism to demarcate reduce the dynamic range. An overlapping window of
PFZ, the thermal fronts is identified, as a first step, 64×64 pixels is translated over the entire region to
using the algorithm prescribed by Cayula and Cornillon detect fronts. The algorithms are directly applied to
(1992). The algorithm employs overlapping window of SST and Chl-a data available in NetCDF format from
32×32 pixels over the entire region to detect fronts. satellite and ROMS.
The ratio of variance between the two populations to The ROMS simulated SST and Chl-a data, having
the variance within the populations is assigned as 0.76. approximately 2.25 km resolution in horizontal direc-
The mean temperature difference between the two popu- tion, are projected on to the observation grid to make a
lations is kept at 0.45°C for identifying the strong ther- comparison between the fronts detected from satellite
mal fronts. The potential fronts from satellite Chl-a are and ROMS. These algorithms are implemented using
detected manually using Aeronautical Reconnaissance Matlab™ software (Mathworks, Inc). In order to evalu-
Coverage Geographic Information System (ArcGIS) ate overlapping front length of satellite and model fronts,
software. Subsequently, the frontal vectors delineated the front pixels of satellite are considered as reference. If
from satellite SST and Chl-a are superimposed to ident- a model front pixel is detected within 5 km from the sat-
ify the common fronts. A tolerance limit of 15 km has ellite front pixel then overlapping pixel counter is
been fixed to identify common fronts derived from increased by one. This process has been carried out for
both SST and Chl-a (Nammalwar et al., 2013). In the all the satellite front pixels each day separately for SST,
absence of Chl-a data, the strong persistent thermal Chl-a fronts, and PFZs throughout the study period
fronts from SST data are identified as PFZ. (2010–2016). Following this process, the value stored
The present study uses Cayula and Cornillon (1992) in the overlapping pixel counter is multiplied by the
algorithm with a modified mean temperature difference corresponding resolution of data to get overlapping
between the two populations as 0.25°C to enhance the front length. The fronts are clustered depending upon
performance to detect moderately strong or weaker SST their length and a frequency distribution (histogram) is
fronts in the Indian coastal waters (Table 1). To track generated to assess the capability of the model.
an edge and form a contour, Cayula and Cornillon
(1992) edge tracking technique is used. Wall et al.
3.6. Detection of persistent occurrence of PFZs
(2008) showed that the regionally tuned Single Image
Edge Detection (Cayula and Cornillon, 1992) is most The PFZ advisories generated during the entire study
suitable for identifying thermal fronts whereas the period (2010–2016) are considered for calculating
algorithm prescribed by Canny (1986) works better in monthly frequency of occurrence and persistence in
delineating Chl-a fronts. Therefore, Canny (1986) interannual time scale. As the spatial coverage of the sat-
algorithm is used to detect Chl-a fronts (Table 2). ellite data is limited, therefore in the first step the inter-
The Canny (1986) edge detection algorithm is applied secting coverage of SST and Chl-a data are identified
each day during the study period. Then, a counter has
been set to keep a record of the availability of both SST
Table 1. The Cayula and Cornillon (1992) edge detection
and Chl-a data in each grid. Now to calculate the number
parameters used in this study are listed below.
Parameter Value
of PFZs available in each grid of the data coverage area
Window size 32 × 32
during the entire study period, a tolerance limit of
Window clarity (% of data have to present in the window) 75% 5 km has been set. The number of PFZs available within
Overlap 50% 5 km from the centre of each grid is recoded in another
The ratio of variance between the two populations to the 0.76
variance within the populations θ(τopt) counter. The ratio of the number of PFZs available in
Gradient ratio 0.9 each grid and the number of data available (both SST
Cohesion coefficients [C1, C2, C] [0.9, 0.9,
0.92] and Chl-a) days gives the occurrence of persistent
Mean temperature difference between the two populations 0.25 PFZs. Finally, the persistent occurrence of PFZs is con-
(surface waters)
verted into percentage.
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 7

Table 3. Performance indices for the relative errors between in situ measured SST and Chl-a with ROMS simulated SST and Chl-a.
Statistical indices are correlation coefficient (R2), slope, intercept (I), root mean squared error (RMSE), absolute (APD), relative (RPD)
and unbiased percentage difference (UPD).
Parameter Region N Slope I R2 RMSE APD RPD UPD
SST West coast 234 1.02 −0.26 0.82 0.80 2.31 −0.78 −0.74
East coast 122 1.06 −1.60 0.88 0.82 2.69 0.24 0.29
All 356 1.06 −1.54 0.89 0.81 2.44 −0.43 −0.39
Chl-a East coast 240 0.86 −0.11 0.83 0.57 55.59 20.73 29.60
West coast 342 0.80 0.04 0.78 0.58 46.58 11.35 18.87
All 582 0.83 −0.02 0.80 0.58 50.30 15.22 23.29

4. Results and discussion In the case of Chl-a, the APD is observed to be higher
for both west (55.59) and east coast (46.58) shelf waters.
4.1. Validation of model-simulated SST and Chl-a
However, the model is able to simulate the data trend as
The performance of model-simulated SST and Chl-a is evident from high correlation coefficient for west (0.83)
evaluated by comparing with the in situ data generated at an east coast (0.78) shelf waters. The low RMSE
SATCORE stations. The evaluation is based on the statisti- (<0.58), bias, RPD and UPD (<29.6%) is indicative of
cal indices such as correlation coefficient (R 2), slope, inter- satisfactory performance of model in simulating Chl-a.
cept, root mean squared error (RMSE), absolute (APD), In an earlier study, an assessment of the accuracy of
relative (RPD) and unbiased percentage difference (UPD) Chl-a simulated by a coupled physical-biogeochemical
(Table 3). The shelf waters of both west and east coast of model configured using ROMS against in situ data
India experiences variable physical processes (Shetye and measured in surface waters of the eastern Arabian Sea
Shenoi, 1988; Shetye et al., 1991; McCreary et al. 1993; was performed. In comparison, the ROMS model under-
Shankar et al., 2002), at seasonal scale, controlling the estimated Chl-a, though the correlation was significant
change in biogeochemistry. Therefore, validation is per- indicating that the model is capable of reproducing the
formed independently on the data from west coast and trend in in situ Chl-a (Chakraborty et al., 2016). The
east coast as well as combining the data for both the coasts superiority of a coupled physical-biogeochemical model
(Figure 3). configured at high resolution (1/12°; approximately
The statistical indices indicated model is performing very ∼9 km spatially averaged) was clearly demonstrated in
well in simulating SST and Chl-a in the Indian shelf waters. a recent study by Chakraborty et al. (2018).
The validation statistics for SST is similar for west and east
coast shelf waters except for bias which was relatively larger
4.2. Validation of model-generated PFZs
at east coast (1.60) compared to the west coast (0.26). The
comparison of measured and model-simulated SST show The capability of coupled physical-biogeochemical
high correlation coefficient (>0.82), slope closer to unity, model in detection of fronts is evaluated by comparing
low RMSE (<0.82) and percentage difference error it with that derived from satellite data (Figures 4–7).
(<2.69%) in both west and east coast shelf waters. The comparison is performed separately for SST and

Figure 3. Scatter plot showing the relationship between in situ measured and model simulated (a) Chlorophyll-a (Chl-a) and (b) Sea
Surface Temperature (SST) in the western and eastern shelf waters of India. The scale represents the actual values. The dotted line
represents 1:1.
8 K. CHAKRABORTY ET AL.

Figure 4. (a) Comparison of SST fronts detected from satellite (Pink) and that of ROMS (Black) data for 10 January 2011; background
image (1.1 km spatial resolution) is from AVHRR sensor onboard MetOp 2 satellite, colour bar is in °C. (b) Comparison of Chl-a fronts
detected from satellite and that of ROMS (Black) data for 10 January 2011; background image (1 km spatial resolution) is from MODIS-
Aqua, colour bar is in logarithmic scale in mg/m3. (c) Comparison of PFZs identified using both SST and Chl-a fronts detected from
satellite (Pink) and ROMS (Black) data for 10 January 2011. (d) Histogram consisting of the binned frequency of the identified PFZs
derived from satellite (Pink) and ROMS (Black) with respect to length of PFZs as class interval of 5 km for 10 January 2011.

Chl-a fronts for the entire study period (2010–2016), and 2.8 km (31.99 ± 4.4 km) whereas for off Andhra Pradesh
for the identified PFZs combining both SST and Chl-a during the entire study period, the average length of sat-
fronts. Considering the availability of maximum spatial ellite-detected thermal fronts (Chl-a fronts) is 26.85 ±
coverage of AVHRR SST and MODIS Chl-a data, a 5.2 km (26.06 ± 5.8) and that using model data is 46.85
total of four cases are presented in shelf waters, three ± 7.1 km (34.78 ± 5.4 km). Figures 4(c)–7(c) are depict-
along the west coast and one case along the east coast. ing the comparison of PFZs, identified from both satellite
The thermal fronts detected, from both satellite retrieved and model, following the methodology described in Sec-
and ROMS simulated SST data, overlaid on the 1.1 km tion 3.5. In comparison, the average length of PFZs
surface temperature image of AVHRR sensor onboard identified from satellite (model) data for off Gujarat
MetOp 2 satellite is shown in Figures 4(a)–7(a) for two during the entire study period is 27.80 ± 7.2 km (33.07
different regions (Figures 4(a)–6(a) for off Gujarat and ± 3.2 km) whereas for off Andhra Pradesh during the
Figure 7(a) for off Andhra Pradesh). In a similar way, entire study period, the average length of PFZs identified
the Chl-a fronts, detected from both satellite retrieved using satellite (model) data is 28.27 ± 10.9 km (52.48 ±
and ROMS simulated surface chlorophyll data, overlaid 8.7 km). A comparison of PFZ lines identified from sat-
on the 1 km surface chlorophyll image of MODIS- ellite and model data is shown by matching their binned
Aqua is presented in Figures 4(b)–7(b). In comparison, frequencies with respect to length of PFZs represented in
it is noted that for off Gujarat during the entire study a class interval (Figure 4(d)–7(d)).
period (2010–2016), the average length of satellite- The satellite-based observations are found to be extre-
detected thermal fronts (Chl-a fronts) is 26.19 ± 4.4 km mely useful in identifying fronts that extend to the sea sur-
(22.72 ± 4.1 km) and that using model data is 31.31 ± face and proved very effective in environmental
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 9

Figure 5. (a) The caption of the figure is same as described in Figure 5 except the date of comparison which is 17 February 2013.

Figure 6. (a) The caption of the figure is same as described in Figure 5 except the date of comparison which is 15 February 2015.
10 K. CHAKRABORTY ET AL.

Figure 7. (a) The caption of the figure is same as described in Figure 5 except the date of comparison which is 18 February 2014.

monitoring of the open and coastal ocean for mapping and A major portion of the profit comes from the savings on
monitoring (Blondeau-Patissier et al., 2014). Rao et al. the cost of fuel due to the avoidance of search for potential
(2002) indicated that PFZ based on SST fronts is useful fishing grounds. On average, the utilization of PFZ advi-
in detecting the pelagic fish landings. The water tempera- sories for fishing reduced the time spent on fishing by
ture is the mostly used environmental parameter in studies 30–70% (Nayak et al. 2007; Kumar et al., 2008).
concerning relationships between the environment and The PFZ advisories are generated and disseminated by
fish behaviour and abundance. However, Rao et al. ESSO-INCOIS have enormous societal impact. However,
(2002) and Solanki et al. (2003, 2005a, 2005b) hypoth- the major problem being faced at the operational level is
esized that an improved methodology of the identification non-availability of satellite data, especially during mon-
of PFZ areas consists of taking into account both SST and soon, mainly due to cloud cover, which also coincides
Chl-a fronts. Subsequently, Choudhury et al. (2007) gener- with the peak-fishing season. The NOAA-AVHRR and
ated PFZ map using NOAA-AVHRR SST and IRS-P4 MODIS-Aqua sensors operate in the infrared and visible
OCM derived chlorophyll maps. regions respectively. These sensors are not capable of pro-
In the case of PFZ advisories provided by INCOIS, sat- viding data (SST and Chl-a) over areas under cloud cover.
ellite-derived SST and Chl-a are considered as key par- Ocean colour sensors take many days, even months, to
ameters. The methodology has been extensively sample the ocean surface enough to provide complete
validated with controlled experiments that have consist- global coverage due to persistent clouds and aerosols
ently indicated that the net profit increased by 2–5 (Gregg and Casey, 2007). In the microwave part of the
times due to the marked increase in Catch Per Unit spectrum, the SST can be observed in the presence of
Effort (CPUE) by 2–4 times (Solanki et al., 2003; Choudh- cloud cover, but the limitation lies in its low spatial resol-
ary et al. 2007; Pillai and Nair, 2010; Deshpande et al. ution. Therefore, there is a need for an alternate source of
2011; Nair and Pillai, 2012; Nammalwar et al., 2013; SST/Chl-a information to generate PFZ advisory which
Maity et al., 2015). The PFZ advisories also contribute can supplement the data from IR and Visible sensors.
to reduction in search time and in locating fish school. Kumar et al. (2012) made an attempt to make use of an
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 11

optimally interpolated merged product of SST (https:// SST fronts obtained from in situ data. They concluded
www.ghrsst.org/) in the operational services. However, that single image frontal detection algorithm can be use-
the coarse resolution and near real-time availability of ful in providing accurate statistics of front occurrence at
this product is still a challenge. scales >10 km. Oram et al. (2008) developed an algor-
Many methods have been developed to predict PFZ ithm utilizing a gradient-based edge detector which is
distributions such as the habitat suitability index (HSI) less sensitive to noise in the input image compared to
model (Chen et al., 2009), generalized linear model previously used detectors. They established that the
(GLM) and generalized additive model (GAM) (Tian algorithm has capability to detect edges at different
et al., 2009). Wang et al. (2015) evaluate a suitable length scales. Zhang et al. (2013) made use of habitat
range of three key environmental variables such as suitability index (HSI) model combining bathymetry
SST, Chl-a and SSH to explore potential fishing zones and model-derived environmental factors such as temp-
(PFZs) using an artificial neural network. Their results erature, salinity, currents etc. to predict Japanese com-
showed that the time, geographical location and SST mon squid in the coastal waters.
had significant influences on the PFZ distribution of The frequency distribution of SST, Chl-a and PFZ
O. bartramii commercial fishery in the Northwest fronts, based on satellite and model data are compared
Pacific Ocean. Ullman and Cornillon (2000) evaluated off Gujarat and Andhra Pradesh (Figures 8–10). The fronts
different gradient and histogram-based edge detection are generated using the daily data for a period from 01-Jan-
algorithms using Advanced Very High Resolution uary-2010 to 31-December-2016. In general, the fronts off
Radiometer SST data and compared their results with Andhra Pradesh (east coast of India) are found to have

Figure 8. (a) The frequency distribution of SST fronts detected from satellite data (red box) and model data (blue box) during the entire
study period (2010–2016) off Gujarat. (b) The frequency distribution of SST fronts detected from satellite data (red box) and model data
(blue box) during the entire study period (2010–2016) off Andhra Pradesh. (c) The frequency distribution of SST fronts detected from
satellite data (red box) and model data (blue box) during the entire study period (2010–2016) combining both off Gujarat and Andhra
Pradesh.
12 K. CHAKRABORTY ET AL.

Figure 9. (a) The frequency distribution of Chl-a fronts detected from satellite data (red box) and model data (blue box) during the
entire study period (2010–2016) off Gujarat. (b) The frequency distribution of Chl-a fronts detected from satellite data (red box)
and model data (blue box) during the entire study period (2010–2016) off Andhra Pradesh. (c) The frequency distribution of Chl-a fronts
detected from satellite data (red box) and model data (blue box) during the entire study period (2010–2016) combining both off Gujarat
and Andhra Pradesh.

larger length scales as compared to that off Gujarat (west et al. 1993; Shankar and Shetye, 1997; Gupta et al.,
coast of India). The distribution of SST, Chl-a and PFZ 2016), and winter convective mixing, during winter
fronts off Gujarat ranges from 15 to 288, 14 to 377 and monsoon (Kumar and Prasad, 1996; Madhupratap
15 to 284 km respectively. Whereas the distribution of et al., 1996) are the major physical processes along the
SST, Chl-a and PFZ fronts off Andhra Pradesh ranges west coast of India. The persistent eddy (Figure 7)
from 15 to 402, 14 to 311 and 15 to 389 km, respectively. along the northern part of east coast (Choudhury et al.
Using daily-averaged SST retrievals, Castelao et al. 2007; Chen et al., 2012) is likely to have an influence
(2006) identified the seasonal evolution of coastal upwel- on water properties. The PFZ fronts with the larger
ling fronts in the California Current System. The spatial length scales off Andhra Pradesh (east coast of India)
and temporal patterns of Chl-a and SST fronts in the are the results of the eddy. In a similar study by Kahru
coastal zone of the northwest Atlantic over a 2-year et al. (2012) observed large-scale distributions of frontal
period revealed that the coastal current is a major driving frequency of both Chl-a and SST of the order 500–
force in establishing congruent chlorophyll and thermal 700 km along the coast coinciding with the area of
fronts (Stegmann and Ullman, 2004). The east and west increased mesoscale eddy activity. They also observed
coast of India, although on the same geographical lati- that more than one year longer variability in SST fronts
tude, experiences diverse physical processes which ulti- is influenced by the large-scale SST gradient, while at
mately controls the biogeochemistry of the shelf waters shorter timescales the influence of Coastal Upwelling
(Wiggert et al., 2009). The coastal upwelling along Index is evident. Mesoscale activity associated with the
southern coast, during summer monsoon (McCreary coastal currents leading to the formation of mesoscale
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 13

Figure 10. (a) The frequency distribution of PFZs identified from satellite data (red box) and model data (blue box) during the entire
study period (2010–2016) off Gujarat. (b) The frequency distribution of PFZs identified from satellite data (red box) and model data
(blue box) during the entire study period (2010–2016) off Andhra Pradesh. (c) The frequency distribution of PFZs identified from sat-
ellite data (red box) and model data (blue box) during the entire study period (2010–2016) combining both off Gujarat and Andhra
Pradesh.

eddies which can eventually move across the coast are In the present study, we have successfully demon-
responsible for the high Chl-a values in the region (Sar- strated the application of a physical-biogeochemical
aceno et al., 2005). model configured using ROMS in providing gap-free
Prominent thermal fronts are observed in the north- continuous data of SST and Chl-a. The PFZ, identified
east Arabian Sea between the 50 m bathymetry contour using the modeled data, matched well with the existing
and continental shelf break (∼200 m contour) and a operational output. In addition, the model is also able
total of 45–75% of thermal fronts occurred at the annual to provide additional features that are missing in oper-
timescales whereas 20–30% occurred at the semi-annual ational advisories due to non availability of satellite data.
time-scales (Mohanty et al., 2017). For the northwestern
Bay of Bengal, the strong inter-annual variability in the
4.3. Persistent occurrence of fronts
spatial location and intensity of eddies is a major force
for the change in the physical-biogeochemical distri- The oceanic fronts are indicative of many oceanographic
bution leading to the rapid change in thermal gradients processes including increased biological activity affecting
(Gopalan et al., 2000; Jyothibabu et al., 2004; Kumar all oceanic life forms from microbes to seabirds and mar-
et al. 2004, 2007; Chen et al., 2012; Nuncio and Kumar ine mammals (Kahru et al., 1984; Quartly and Srokosz,
2012). The results of Jing et al. (2015) indicated that pro- 2003; Bost et al., 2009). In the present study, we use a
nounced surface cooling and upwelling-related fronts 7-year long time series of satellite and model data to
occur around Hainan Island with a width of 20–50 km. identify zones of persistent occurrence of PFZs. The per-
These fronts persist through the summer upwelling season. sistent occurrence PFZs identified using model simulated
14 K. CHAKRABORTY ET AL.

data (in shade) overlaid with satellite contours for the of PFZs identified using model data whereas, in the sum-
months of January and February during winter season, mer season, persistent occurrence of PFZs are not dis-
and for the months of August and September during cernible from the available satellite data.
summer season are shown off Gujarat (Figure 11). It is Figure 12 depicts a comparison of persistent occur-
observed that the persistent occurrence of PFZs ident- rence of PFZs identified from both satellite and model
ified using model data has more spatial coverage in com- data off Andhra Pradesh for the months of December
parison to persistent areas identified from satellite data. and January during winter monsoon, and for the
The persistent occurrence of PFZs in the case of model months of April and May during pre-monsoon. During
(satellite) data has maximum 70% (20%) occurrence pre-monsoon, the region is highly influenced by
during the winter season. Whereas during the summer eddies/gyres resulting in strong gradient in temperature,
season though 30–40% persistent occurrence of PFZs and by coastal upwelling that pumps nutrient to the sur-
can be identified from the model simulated data, but face which in turn triggering high productivity. There-
no such persistent occurrence of PFZs is identifiable fore, April and May are the most relevant months for
from satellite data. In this context, it is to be noted that making such comparison. The availability of satellite
for off Gujarat, the availability of satellite retrieved SST retrieved SST and Chl-a data in this region for the
and Chl-a data for the months of January (February) months of December (January) during the entire study
during the entire study period is, respectively 79.8% period is, respectively 59.6% (73.4%) and 13.4%
(78.5%) and 43.5% (46.3%) whereas for the months of (19.1%). Whereas for the months of April (May), the
August (September), the availability of satellite retrieved availability of satellite retrieved SST and Chl-a data is
SST and Chl-a data is respectively, 15.9% (44.9%) and respectively, 67.8% (45.9%) and 7.7% (3.7%). In pre-
2.3% (8.3%). Also, it is to be noted that a PFZ can only monsoon, the model simulated data indicates, 40–50%
be identified if both SST and Chl-a data are available at persistent occurrence of PFZs in this region, but in the
the same time. As, the spatial coverage of Chl-a data is case of satellite, the availability of data is not sufficient
very poor in both seasons, specifically in the summer sea- to identify persistent occurrence of PFZs. In winter
son, therefore during the winter season, the percentage of monsoon, the maximum 20% persistent occurrence of
persistent occurrence of PFZs identified using satellite PFZs is observed from the available satellite data
data are less in comparison to the persistent occurrence whereas, in case of model, the maximum persistent

Figure 11. Satellite derived persisting PFZs (in contour) overlaid on ROMS derived persisting PFZs (in shade) for the entire study period
(2010–2016) off Gujarat.
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 15

Figure 12. Satellite derived persisting PFZs (in contour) overlaid on ROMS derived persisting PFZs (in shade) for the entire study period
(2010–2016) off Andhra Pradesh.

occurrence of PFZs is 20–30% except at a few small incorporated as a proxy for pelagic diversity in the des-
patches with 50% occurrence. ignation of Marine Protected Areas in a UK Continental
There have been many studies demonstrating the vital Shelf. The distribution and variability of SST fronts, over
applications of the areas experiencing persistent fronts the shelf and slope along the east coast of North America
(Ullman and Cornillon, 2001; Priede and Miller, 2009). using a 12-year time series of advanced very high-resol-
There are also reports of relationships between fronts ution radiometer images revealed the presence of persist-
and fish abundance such as swordfish (Podesta et al., ent fronts off the northeast US coast (Ullman and
1993), tuna and billfish (Worm et al., 2005). In addition, Cornillon, 1999).
Worm et al. (2005) determined a global correlation of The use of a model simulated gap-free data is found to
predator diversity with fronts. The studies also revealed be an additional advantage over satellite data, since the
the strong relationship between the track of a tagged study region tends to be cloudy in all seasons except
basking shark and a thermal front (Priede and Miller, few days in the winter season, and only relatively few
2009). The shark followed the N–S front for a whole clear MODIS-Aqua images is generally obtained
day, keeping just to the warmer side. Analyzing 12 (Gregg and Casey, 2007; Miller, 2009). The use of
years (1985–1996) of SST imagery, Ullman and Cornil- model-simulated gap-free data allowed to identify
lon (2001) has revealed the presence of persistent fronts, more persistent areas of PFZs with greater percentage
having cool water inshore and warmer outer shelf water in comparison to the persistent areas identified through
during winter over the middle shelf off northeast coast of satellite data in the study region.
US. The winter surface cooling was found to play an
important role in the formation of these persistent ther-
5. Conclusions
mal fronts. The highest frequency of these fronts was
observed in the vicinity of the 50 m isobath using satellite To address operational and scientific needs of the ESSO-
data. Miller and Christodoulou (2014) processed 10-year INCOIS, a series of high resolution coupled physical-bio-
time-series satellite images and aggregated to generate a geochemical models have been configured at ESSO-
front climatology of the UK continental shelf, indicating INCOIS. The ability of these models to reproduce the sat-
the regions where strong fronts are most frequently ellite-derived PFZs in the Indian Exclusive Economic
observed during each season. These maps were also Zone (EEZ) is investigated. Our evaluation shows that
16 K. CHAKRABORTY ET AL.

the high-resolution models are able to reproduce the Acknowledgements

thermal fronts leading to PFZs in reasonably good confi- We are grateful to the anonymous reviewers for their careful
dence. The statistics of the above-mentioned compari- reading, constructive comments and helpful suggestions,
son indicates that though the models are capable of which have helped us to significantly improve the presentation
reproducing PFZs but in most of the cases the models of this work. The first author earnestly thanks MDG team
overestimate the total length of all the PFZs. However, members for the help in setting up the ROMS simulations.
The physical setup of the ROMS-1/48° coupled model is confi-
it is worth noting that the model simulates spatio-
gured by Dr Francis P.A., but the C-preprocessing options of
temporarily gap-free data but the satellite images do the physical setup have been changed after integrating ecosys-
have data gap. Therefore, all the potential fonts may tem module. We acknowledge the DMG group and CWG
not be detectable from the satellite images in the study Group of INCOIS, respectively for making available satellite
area which leads to this overestimation. Additionally, data used in this study and necessary support to run ROMS
though the model simulated SST/Chl-a data are capable in INCOIS-HPC. We gratefully acknowledge the Directors of
INCOIS for his encouragement to this work. The ecosystem
of producing ocean features leading to PFZs reasonably model configured as a part of the Ocean – Modeling, Data
well but the model is producing large scale features better Assimilation and Process SpeCific ObservaTions (O-MAS-
than the smaller scale features. COT) programme coordinated by the Indian National Centre
Nevertheless, a physical-biogeochemical model simu- for Ocean Information Services (INCOIS) and funded by Min-
lated SST and Chl-a data can be used in complementing istry of Earth Sciences (MoES) (Ref: MoES/36/OOIS/O-MAS-
COT/2017). In situ data used in this paper is collected as a part
spatio-temporarily gap-free generation of operational
of Satellite Coastal and Oceanographic Research (SATCORE)
PFZ advisories. Moreover, one of the future goals programme coordinated by the Indian National Centre for
towards MFAS programme is to provide the PFZ advi- Ocean Information Services (INCOIS) and funded by Ministry
sories in the forecast mode to fisherman community, of Earth Sciences (MoES) (Ref: MoES/36/OOIS/SATCORE/
and seasonal outlooks to policy makers – for effective 2012). We are thankful to Dr Arya Paul for helping us in
and sustainable fishery in the region. As the ocean fea- improving the English of the manuscript. We thank Dr
Jason J. Roberts, Marine Geospatial Ecology Lab, Duke Uni-
tures observed from model simulated gap-free data is versity and Dr Carrie C. Wall Bell, Cooperative Institute for
well in agreement with the ocean features observed Research in Environmental Sciences, University of Colorado
through satellite data, therefore the model simulated at Boulder for their help in developing automated front detec-
SST and Chl-a data can be used not only in complement- tion system. This is INCOIS contribution no. 332.
ing satellite aided PFZ advisories but also in transform-
ing PFZ advisories into PFZ forecast.
Disclosure statement
At present, surface chlorophyll is the only biological
component of the marine ecosystem that can be opera- No potential conflict of interest was reported by the authors.
tionally derived with the help of remote sensing. How-
ever, high spatio-temporal coverage of vertical profile
Funding
of the chlorophyll concentration or other key parameter
such as dissolved oxygen is not available by any means. This work was supported by Ministry of Earth Sciences [grant
Although, Argo-float based biogeochemical observations number MoES/36/OOIS/O-MASCOT/2017].
have been initiated in the Indian Ocean region recently,
but the float density is very sparse. Therefore, the ORCID
existing operational system does not consider the
subsurface properties as potential indicators while iden- Kunal Chakraborty http://orcid.org/0000-0001-6940-9355
Sourav Maity http://orcid.org/0000-0003-2337-5457
tifying PFZs. Again, it is evident that the scientific
Aneesh A. Lotliker http://orcid.org/0000-0001-5467-173X
acceptability of the PFZs will be definitely enhanced if Alakes Samanta http://orcid.org/0000-0002-6582-5850
the subsurface properties can also be included as poten- Jayashree Ghosh http://orcid.org/0000-0002-5043-6617
tial indicators towards demarcation of PFZs. In this Nagaraja Kumar Masuluri http://orcid.org/0000-0002-
regard, numerical ocean modeling is the only solution 9692-7892
to meet such operational requirements. A biogeochem- Naga Swetha http://orcid.org/0000-0001-5782-2003
Rose P. Bright http://orcid.org/0000-0001-7030-0955
ical model coupled with physical model allows to capture
the spatio-temporal complexities inherent to the ecosys-
tem. Such model is extremely useful to understand the References
complex biogeochemical interactions that constitute
Ahn YH, Shanmugam P. 2006. Detecting the red tide algal
the food web and, subsequently make use of the subsur- blooms from satellite ocean color observations in optically
face properties as potential indicators for demarcation complex northeast-Asia coastal waters. Remote Sens
of PFZs. Environ. 103(4):419–437. doi:10.1016/j.rse.2006.04.007.
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 17

Arnone RA. 1987. Satellite-derived color-temperature Fairall CW, Bradley EF, Rogers DP, Edson JB, Young GS. 1996a.
relationship in the Alboran Sea. Remote Sens Environ. 23 Bulk parameterization of air-sea fluxes for Tropical ocean-
(3):417–437. Global Atmosphere coupled-ocean Atmosphere Response
Blondeau-Patissier D, Gower JF, Dekker AG, Phinn SR, Experiment. J Geophys Res Oceans. 101(C2):3747–3764.
Brando VE. 2014. A review of ocean color remote sensing Fennel K, Wilkin J, Levin J, Moisan J, O’Reilly J, Haidvogel D.
methods and statistical techniques for the detection, map- 2006. Nitrogen cycling in the middle Atlantic Bight: results
ping and analysis of phytoplankton blooms in coastal and from a three-dimensional model and implications for the
open oceans. Prog Oceanogr. 123:123–144. North Atlantic nitrogen budget. Global Biogeochem
Bost CA, Cotté C, Bailleul F, Cherel Y, Charrassin JB, Guinet Cycles. 20(3):1–14.
C, Weimerskirch H. 2009. The importance of oceano- Fennel K, Wilkin J, Previdi M, Najjar R. 2008. Denitrification
graphic fronts to marine birds and mammals of the effects on air-sea CO2 flux in the coastal ocean: simulations
southern oceans. J Mar Sys. 78(3):363–376. for the northwest North Atlantic. Geophys Res Lett. 35
Canny J. 1986. A computational approach to edge detection. (24):1–5.
IEEE Trans Pattern Anal Mach Intell. 8(6):679–698. Fiedler PC, Smith GB, Laurs RM. 1984. Fisheries applications
Castelao RM, Mavor TP, Barth JA, Breaker LC. 2006. Sea sur- of satellite data in the eastern North Pacific. Fisheries. 46
face temperature fronts in the California current System (3):1–13.
from geostationary satellite observations. J Geophys Res. Gomes HR, Goes JI, Saino T. 2000. Influence of physical pro-
111(C9):1–13. cesses and freshwater discharge on the seasonality of phyto-
Cayula JF, Cornillon P. 1992. Edge detection algorithm for SST plankton regime in the Bay of Bengal. Cont Shelf Res. 20
images. J Atmos Ocean Technol. 9(1):67–80. (3):313–330.
Chakraborty K, Gupta A, Lotliker AA, Tilstone G. 2016. Gopalan AKS, Krishna VV, Ali MM, Sharma R. 2000.
Evaluation of model simulated and MODIS-aqua retrieved Detection of Bay of Bengal eddies from TOPEX and in
sea surface chlorophyll in the eastern Arabian Sea. situ observations. J Mar Res. 58(5):721–734.
Estuarine Coastal Shelf Sci. 181:61–69. Gregg WW, Casey NW. 2007. Sampling biases in MODIS and
Chakraborty K, Kumar N, Gupta GVM. 2017. Getting the right SeaWiFS ocean chlorophyll data. Remote Sens Environ. 111
wind-forcing for an ecosystem model: A case study from the (1):25–35.
eastern Arabian Sea. J Operat Oceanogr. 10(2):176–190. Gupta GVM, Sudheesh V, Sudharma KV, Saravanane N,
Chakraborty K, Valsala V, Gupta, GVM, Sarma VVSS. 2018. Dhanya V, Dhanya KR, Lakshmi G, Sudhakar M, Naqvi
Dominant biological control over upwelling on pCO2 in SWA. 2016. Evolution to decay of upwelling and associated
sea east of Sri Lanka. J Geophys Res Biogeosci. 123. biogeochemistry over the southeastern Arabian Sea shelf. J
doi:10.1029/2018JG004446. Geophys Res Biogeosci. 121(1):159–175.
Chen XJ, Liu BL, Tian SQ, Qian WG, Li G. 2009. Forecasting Haidvogel DB, Arango H, Budgell WP, Cornuelle BD,
the fishing ground of Ommastrephes bartramii with SST- Curchitser E, Di Lorenzo E, Fennel K, Geyer WR,
based habitat suitability modeling in northwestern pacific. Hermann AJ, Lanerolle L, et al. 2008. Ocean forecasting
Oceanol Limnol. 40(6):707–713. in terrain-following coordinates: Formulation and skill
Chen G, Wang D, Hou Y. 2012. The features and interannual assessment of the regional ocean modeling System. J
variability mechanism of mesoscale eddies in the Bay of Comput Phys. 227(7):3595–3624.
Bengal. Cont Shelf Res. 47:178–185. doi:10.1016/j.csr.2012. Jing Z, Qi Y, Du Y, Zhang S, Xie L. 2015. Summer upwelling
07.011. and thermal fronts in the northwestern South China Sea:
Choudhury SB, Jena B, Rao MV, Rao KH, Somvanshi VS, observational analysis of two mesoscale mapping surveys.
Gulati DK, Sahu SK. 2007. Validation of integrated potential J Geophys Res Oceans. 120(3):1993–2006.
fishing zone (IPFZ) forecast using satellite based chlorophyll Jyothibabu R, Maheswaran PA, Madhu NV, Ashraf TM,
and sea surface temperature along the east coast of Gerson VJ, Haridas PC, Gopalakrishnan TC. 2004.
India. Int J Remote Sens. 28(12):2683–2693. doi:10.1080/ Differential response of winter cooling on biological pro-
01431160600987878. duction in the northeastern Arabian Sea and northwestern
Deshpande SP, Radhakrishnan KV, Bhat UG. 2011. Direct and Bay of Bengal. Curr Sci. 87:783–791.
indirect validation of potential fishing zone advisory off the Kahru M, Di Lorenzo E, Manzano-Sarabia M, Mitchell BG.
coast of Uttara Kannada, Karnataka. J Indian Soc Remote 2012. Spatial and temporal statistics of sea surface tempera-
Sens. 39(4):547–554. ture and chlorophyll fronts in the California current. J
Dwivedi R, Rafeeq M, Smitha BR, Padmakumar KB, Thomas Plankton Res. 34(9):749–760.
LC, Sanjeevan VN, Raman M. 2015. Species identification Kahru M, Elken J, Kotta I, Simm M, Vilbaste K. 1984. Plankton
of mixed algal bloom in the northern Arabian Sea using distributions and processes across a front in the open Baltic
remote sensing techniques. Environ Monit Assess. 187 Sea. Mar Ecol Prog Ser. 20:101–111.
(2):51. doi:10.1007/s10661-015-4291-2. Kumar TS, Kumar MN, Nayak S. 2008. Benefits derived by the
Dwivedi RM, Solanki HU, Nayak SR, Gulati D, Somvanshi VS. fishermen using potential fishing zone (PFZ) advisories.
2005. Exploration of fishery resources through integration Proceedings of SPIE – The International Society for
of ocean colour with sea surface temperature: Indian experi- Optical Engineering, 7150, “p. 71500N”.
ence. Indian J Mar Sci. 34(4):430–440. Kumar MN, Kumar TS, Swetha NN, Maity S, Jemima U. 2012.
Fairall CW, Bradley EF, Godfrey JS, Wick GA, Edson JB, A Comparative study on feasibility of utilization of
Young GS. 1996b. Cool-skin and warm-layer effects on GHRSST product for operational generation of Potential
sea surface temperature. J Geophys Res Oceans. 101 Fishing Zone (PFZ) advisories in India. Int J Earth Sci
(C1):1295–1308. Eng. 5:240–246.
18 K. CHAKRABORTY ET AL.

Kumar SP, Muraleedharan PM, Prasad TG, Gauns M, NAIP. 2012. Annual Report 2011-12: National Agricultural
Ramaiah N, De Souza SN, Madhupratap M. 2002. Why is Innovation Project. Indian Council of Agricultural
the Bay of Bengal less productive during summer monsoon Research (ICAR), New Delhi.
compared to the Arabian Sea? Geophys Res Lett. 29(24):1–4. Nair PG, Pillai VN. 2012. Satellite based potential fishing zone
Kumar MN, Nair P, Pillai VN, Kumar TS. 2018. (PFZ) advisories-acceptance levels and benefits derived by
Environmental benefits due to adoption of satellite-based the user community along the Kerala coast. Indian J Fish.
fishery advisories. Fish Technol. 55(2):100–103. 59(2):69–74.
Kumar SP, Nuncio M, Narvekar J, Kumar A, Sardesai S, De Nammalwar P, Satheesh S, Ramesh R. 2013. Applications of
Souza SN, Madhupratap M. 2004. Are eddies nature’s trig- remote sensing in the validations of potential fishing
ger to enhance biological productivity in the Bay of Bengal? zones (PFZ) along the coast of North Tamil Nadu, India.
Geophys Res Lett. 31(7):1–4. Indian J Geo-Mar Sci. 42(3):283–292.
Kumar SP, Nuncio M, Ramaiah N, Sardesai S, Narvekar J, Narain A, Dwivedi RM, Solanki HU, Kumari B, Chaturvedi N,
Fernandes V, Paul JT. 2007. Eddy-mediated biological pro- James PSBR, Somvanshi VS. 1990. The use of NOAA-
ductivity in the Bay of Bengal during fall and spring inter- AVHRR data in fisheries exploration in the Indian EEZ.
monsoons. Deep Sea Res Part I. 54(9):1619–1640. Proc. Of the Seminar on Remote Sensing for marine
Kumar SP, Prasad TG. 1996. Winter cooling in the northern fisheries studies, Beijing, China, ESCAP/UNDP, 226-232.
Arabian Sea. Curr Sci. 71:834–841. Nath AN, Rao MV, Reddy SR, Das NK, Baral NC. 1991.
Lasker R, Peláez J, Laurs RM. 1981. The use of satellite infrared Application of satellite derived sea-surface temperature for
imagery for describing ocean processes in relation to spawn- estimation of fish catch: A pilot study. Ind J Mar Sci.
ing of the northern anchovy (Engraulis mordax). Remote 20:152–154.
Sens Environ. 11:439–453. Nayak S, Kumar TS, Masuluri NK. 2007. Satellite based fishery
Laurs RM, Fiedler PC, Montgomery DR. 1984. Albacore tuna service in India THE FULL PICTURE. Geneva, Switzerland:
catch distributions relative to environmental features Tudor Rose on behalf of the GEO; p. 256–257.
observed from satellites. Deep Sea Res Oceanogr Res Nuncio M, Kumar SP. 2012. Life cycle of eddies along the wes-
Papers. 31(9):1085–1099. tern boundary of the Bay of Bengal and their implications. J
Liu WT, Katsaros KB, Businger JA. 1979. Bulk parameteriza- Mar Sys. 94:9–17.
tion of air-sea exchanges of heat and water vapor including Oram JJ, McWilliams JC, Stolzenbach KD. 2008. Gradient-
the molecular constraints at the interface. J Atmos Sci. 36 based edge detection and feature classification of sea-surface
(9):1722–1735. images of the southern California Bight. Remote Sens
Lotliker AA, Baliarsingh SK, Kumar ST. 2016. Satellite Coastal Environ. 112(5):2397–2415.
and Oceanographic Research Inter-comparison Exercise Pillai VN, Nair PG. 2010. Potential fishing zone (PFZ) advi-
(SICOME). Technical Report No. ESSO/INCOIS/ASG/TR/ sories-Are they beneficial to the coastal fisherfolk? A case
06. INCOIS, Hyderabad. study along Kerala coast, South India. Biol Forum Int J. 2
Madhupratap M, Gauns M, Ramaiah N, Kumar SP, (2):46–55.
Muraleedharan PM, De Sousa SN, Muraleedharan U. Podesta GP, Browder JA, Hoey JJ. 1993. Exploring the associ-
2003. Biogeochemistry of the Bay of Bengal: physical, ation between swordfish catch rates and thermal fronts on
chemical and primary productivity characteristics of the U.S. longline grounds in the western North Atlantic. Cont
central and western Bay of Bengal during summer monsoon Shelf Res. 13(2-3):253–277.
2001. Deep Sea Res Part II. 50(5):881–896. Priede IG, Miller PI. 2009. A basking shark (Cetorhinus max-
Madhupratap M, Kumar SP, Bhattathiri PMA, Kumar MD, imus) tracked by satellite together with simultaneous
Raghukumar S, Nair KKC, Ramaiah N. 1996. Mechanism remote sensing II: new analysis reveals orientation to a ther-
of the biological response to winter cooling in the northeast- mal front. Fish Res. 95(2-3):370–372.
ern Arabian Sea. Nature. 384(6609):549–552. Quartly GD, Srokosz MA. 2003. A plankton guide to ocean
Maity S, Kumar TS, Dutta S, Akhand A, Hazra S. 2015. Satellite physics: Colouring in the currents round South Africa and
based integrated potential fishing zone advisories: a feasi- Madagascar. Ocean Chall. 12(3):19–23.
bility analysis in the coastal water of West Bengal. Proc Rajagopal EN, Iyengar GR, George JP, Gupta MD, Mohandas
Zool Soc. 68:14–19. S, Siddharth R, Ashish A. 2012. Implementation of unified
McCrearyJr JP, Kundu PK, Molinari RL. 1993. A numerical model based analysis-forecast system at NCMRWF.
investigation of dynamics, thermodynamics and mixed- NMRF/TR/2/2012, 45.
layer processes in the Indian ocean. Prog Oceanogr. 31 Rao KH, Choudhury SB, Rao MV. 2002. Potential fishing zone
(3):181–244. mapping-an overview. Indian Soc Geomat Newslett.
Miller P. 2009. Composite front maps for improved visibility 9:48–51.
of dynamic sea-surface features on cloudy SeaWiFS and Saraceno M, Provost C, Piola AR. 2005. On the relationship
AVHRR data. J Mar Sys. 78(3):327–336. between satellite-retrieved surface temperature fronts and
Miller PI, Christodoulou S. 2014. Frequent locations of chlorophyll a in the western South Atlantic. J Geophys
oceanic fronts as an indicator of pelagic diversity: application Res. 110(C11):1–16.
to marine protected areas and renewables. Mar Policy. Shankar D, Shetye SR. 1997. On the dynamics of the
45:318–329. Lakshadweep high and low in the southeastern Arabian
Mohanty PC, Mahendra RS, Nayak RK, Kumar N, Kumar TS, Sea. J Geophys Res Oceans. 102(C6):12551–12562.
Dwivedi RM. 2017. Persistence of productive surface ther- Shankar D, Vinayachandran PN, Unnikrishnan AS. 2002. The
mal fronts in the northeast Arabian Sea. Reg Stud Marine monsoon currents in the north Indian ocean. Prog
Sci. 16:216–224. Oceanogr. 52(1):63–120.
 JOURNAL OF OPERATIONAL OCEANOGRAPHY 19

Shetye SR, Shenoi SSC. 1988. Seasonal cycle of surface circula- northwest Pacific Ocean. J Shanghai Ocean Univ. 18
tion in the coastal north Indian ocean. Proc Indian Acad Sci (05):586–592.
Earth Planet Sci. 97(1):53–62. Ullman DS, Cornillon PC. 1999. Satellite-derived sea surface
Shetye SR, Shenoi SSC, Gouveia AD, Michael GS, Sundar D, temperature fronts on the continental shelf off the northeast
Nampoothiri G. 1991. Wind-driven coastal upwelling U.S. coast. J Geophys Res: Oceans. 104(C10):23459–23478.
along the western boundary of the Bay of Bengal during doi:10.1029/1999JC900133.
the southwest monsoon. Cont Shelf Res. 11(11):1397–1408. Ullman DS, Cornillon PC. 2000. Evaluation of front detection
Singh VV, Singh DP. 2016. mKRISHI® Fisheries – A Blue methods for satellite-derived SST data using in situ obser-
Ocean Innovation. Marine Fisheries Information Service; vations. J Atmos Ocean Technol. 17(12):1667–1675.
Technical and Extension Series (230) ISSN 0254-380 X. 3-6. Ullman DS, Cornillon PC. 2001. Continental shelf surface
Solanki HU, Bhatpuria D, Chauhan P. 2015. Signature analysis thermal fronts in winter off the northeast US coast. Cont
of satellite derived SSHa, SST and chlorophyll concentration Shelf Res. 21(11–12):1139–1156.
and their linkage with marine fishery resources. J Mar Sys. Vedavalli L, Suvitha D. 2013. Role of helpline services in the
150:12–21. lives of fishermen: case studies from Tamil Nadu and
Solanki HU, Dwivedi RM, Nayak SR. 2001b. Synergistic analy- Andhra Pradesh. Chennai: M S Swaminathan Research
sis of SeaWiFS chlorophyll concentration and NOAA- Foundation. (MSSRF). PB:20.
AVHRR SST features for exploring marine living resources. Vedavalli L, Velvizhi S. 2015. Utility and impact of ESSO-
Int J Remote Sens. 22(18):3877–3882. INCOIS Services: reflections of fishers from Andhra
Solanki HU, Dwivedi RM, Nayak SR, Jadeja JV, Thakar DB, Pradesh, Tamil Nadu, Kerala & Puducherry. Chennai: M
Dave HB, Patel MI. 2001a. Application of ocean colour S Swaminathan Research Foundation (MSSRF). MSSRF/
Monitor chlorophyll and AVHRR SST for fishery forecast: RR/15/38, PB57.
preliminary validation results off Gujarat coast, northwest Vedavalli L, Velvizhi S, Suvitha D, Rao NV, Rao DS, Dhanraj
coast of India. Indian J Mar Sci. 30(3):132–138. K, Maity S. 2014. Potential fishing zone advisories and con-
Solanki HU, Dwivedi RM, Nayak SR, Naik SK, John ME, version from bottom trawling to gillnetting: role of MSSRF-
Somvanshi VS. 2005a. Cover: application of remotely INCOIS Partnership in Gilakaladindi village in Krishna dis-
sensed closely coupled biological and physical processes trict, Andhra Pradesh – a preliminary study. Chennai: M S
for marine fishery resources exploration. Int J Remote Swaminathan Research Foundation. (MSSRF). MSSRF/RR/
Sens. 26(10):2029–2034. 14/37,PB:48.
Solanki HU, Dwivedi RM, Nayak SR, Somvanshi VS, Gulati Venkatesan R, Joshi L. 2010. Impact assessment and economic
DK, Pattnayak SK. 2003. Fishery forecast using OCM chlor- benefits of weather and marine Services. New Delhi, India:
ophyll concentration and AVHRR SST: validation results off National Council of Applied Economic Research (NCAER).
Gujarat coast, India. Int J Remote Sens. 24(18):3691–3699. Wall CC, Muller-Karger FE, Roffer MA, Hu C, Yao W, Luther
Solanki HU, Mankodi PC, Nayak SR, Somvanshi VS. 2005b. ME. 2008. Satellite remote sensing of surface oceanic fronts
Evaluation of remote-sensing-based potential fishing in coastal waters off west–central Florida. Remote Sens
zones (PFZs) forecast methodology. Cont Shelf Res. 25 Environ. 112(6):2963–2976.
(18):2163–2173. Wang J, Yu W, Chen X, Lei L, Chen Y. 2015. Detection of
Stegmann PM, Ullman DS. 2004. Variability in chlorophyll potential fishing zones for neon flying squid based on
and sea surface temperature fronts in the long Island remote-sensing data in the northwest Pacific Ocean using
Sound outflow region from satellite observations. J an artificial neural network. Int J Remote Sens. 36
Geophys Res. 109(C7):1–12. (13):3317–3330.
Surendran S. and Velvizhi S. 2011. The Fisher Friend Mobile Wiggert JD, Hood RR, Naqvi SW, Brink KH, Smith SL. 2009.
Application: Case Studies on Effective Use of Ocean State Indian ocean biogeochemical processes and ecological
forecast and Potential Fishing Zone for a Better variability. Washington, DC: American Geophysical
Livelihood. Chennai: M S Swaminathan Research Union Geophysical Monograph Series. 185; p. 429.
Foundation (MSSRF). MSSRF/MA/11/53,PB:44. Worm B, Sandow M, Oschlies A, Lotze HK, Myers RA. 2005.
Suvitha D, Vedavalli L. 2013. Improving Livelihoods of fisher- Global patterns of predator diversity in the open oceans.
men- case studies from Andhra Pradesh and Union Science. 309(5739):1365–1369.
Terriotory of Puducherry. Chennai: M S Swaminathan Zhang X, Saitoh SI, Hirawake T, Nakada S, Koyamada K,
Research Foundation. (MSSRF).PB:40. Awaji T, Ishikawa Y, Igarashi H. 2013. An attempt of disse-
Tian SQ, Chen XJ, Feng B, Qian WG. 2009. Spatio-Temporal mination of potential fishing zones prediction map of
distribution of abundance index for Ommastrephes bartra- Japanese common squid in the coastal water, southwestern
mii and its relationship with habitat environment in the Hokkaido, Japan. Proc Asia-Pacific Adv Netw. 36:132–141.