Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

A Data Driven Approach to Dynamic

Geofencing for Sustainable and Profitable Fisheries
Bhargav Chaudhari

Abstract:- Geofencing is a technology that uses a virtual  Seer Fish (Scomberomorus guttatus)
perimeter around real world areas. This concept is  Bombay Duck (Harpadon nehereus)
discussed widely in multiple past publications, in the  Threadfin Bream (Nemipterus japonicus)
context of fishery. The most common application is for  Prawns and Shrimps
marking international coastal boundaries and preventing  Lobsters
their violation by fishermen. Here, geofencing is taken
into application for sustainable fishing practices along If the coast of Maharashtra State is taken into
with helping the individuals working in this industry consideration, the most found and demanding species are
guarantee their safety by maximizing fish yield. The Indian Mackerel, locally known as Bangda, King Mackerel
dynamic nature of the geofences is proposed in this (Surmai), Pomfret and the Bombay Duck (Bombil).
research, where the virtually marked areas are
susceptible to change based on the real time and past
generated data, thus safeguarding the breeding patterns
of multiple fish species. The approach’s implementation is
further explained with the example of the coastline of
Maharashtra state in India. This technique is possible to
be implemented at different levels of authority, namely
the district, state, and the country, for varied scales of
fishermen which is further discussed. This approach is
advantageous as it helps in balancing the fish population,
and guiding fishermen to find the most prone fishing
spots, also protecting them from accidental violation of
the country’s borders. The implementation accuracy of
this model depends on the data sources provided,
including the Orbital Thermal Imaging and other
weather-related data.

Keywords:- Fishery, Geospatial Data Analysis,

Sustainability, Predictive Modeling, Geofencing, Big Data,
Marine Protected Areas, Remote Sensing, Random Forest
Algorithm.

I. INTRODUCTION Fig 1 Coastline of Maharashtra. Marked in Gray are

Coastal Districts.
Fishery is a source of livelihood for a large amount of (Source. Wikipedia)
individuals on the coasts of India. Around 2.8 crore people are
directly or indirectly employed in this sector. India is blessed The species like surmai being migratory in nature, there
with a large coastline on both the east and west sides, making lies an entire sea to move, depending on multiple
India the Second largest producer of fish in the world. In the environmental factors such as Climate change, water
F.Y. 2021-22 India’s annual fish production was 162.48 lakh temperature, ocean currents, salinity level, food availability
tonnes. (source: PIB, Ministry of Fisheries, Animal and even more.
Husbandry & Dairying).
This uncertainty makes it quite difficult for the
 Indian Coast is the Home for a Diverse Range of Fish fishermen to find the right spot for catching.
Species, out of which the Species with High Commercial
value Include: Also, the fish species have specific conditions to breed
in, including season and other factors mentioned above. The
 Indian Mackerel (Rastrelliger kanagurta) fish breeding zones are always vulnerable to be affected from
 Sardines (Sardinella longiceps) fish catching activities, as the young fish being caught cause a
 Tuna (Thunnus spp.) population imbalance in species, also they lack the
 Pomfret (Pampus argenteus) commercial market value that a fully grown fish gets.

IJISRT24SEP1401 www.ijisrt.com 2026

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

The objective of this research is to help fishermen to find II. LITERATURE SURVEY
the right fishing spot and safeguarding the natural habitat of
the fish species. The concept of geofencing is widely discussed in
multiple past publications, one such example is “Fisherman
Dynamic geofencing, a technology that utilizes real-time Navigation and Safety System” by M.B. Mukesh Krishnan, D.
data and advanced analytics to create virtual boundaries, Saveetha, A. Arokiaraj Jovith, P. Rajasekar. This research
offers a promising solution to these problems. By establishing elaborates the implementation of Geofences for the safety of
dynamic geofences based on factors such as fish stock Fishermen, by using positioning systems to avoid their
distribution, environmental conditions, and legal boundaries, violation of international boundaries. An example of the Sea
it is possible to guide fishermen towards sustainable fishing Border between India and Sri Lanka, where fishermen from
grounds while preventing them from entering protected areas. India are frequently caught by the Naval Forces of Sri Lanka.
Although the paper suggests using the Mobile devices for
This research aims to develop a data-driven approach to tracking the fishermen, the devices usually do not get internet
dynamic geofencing for the Indian fisheries industry. By signal deep into the sea, so the actual implementation results
leveraging satellite imagery, historical data, and machine are dependant to the proximity of the devices from the Indian
learning techniques, we will create a system that can Coast.
accurately predict fish distribution, identify potential fishing
hotspots, and enforce marine regulations. This will not only The Fish production and its improvement is further
contribute to the conservation of marine ecosystems but also studied in “Estimating fishing effort from highly resolved
enhance the profitability of the fishing industry by ensuring geospatial data: Focusing on passive gears” by T. Mendo a, *,
sustainable and efficient resource management. G. Glemarec b, J. Mendo c, E. Hjorleifsson d, S. Smout e, S.
Northridge e, J. Rodriguez f, A. Mujal-Colilles g, M. James.
 Significance: The paper discusses the rate of output with respect to the
By leveraging technology, data analytics, and machine effort taken for it.
learning, this approach can create virtual boundaries that
guide fishermen towards sustainable fishing grounds while Fishing Effort Estimation Methods: Previous studies on
protecting marine ecosystems. Through real-time monitoring fishing effort, particularly in large-scale fisheries using active
and adaptive updates, dynamic geofencing can help to reduce gears, have predominantly focused on time spent at sea as a
overfishing, prevent illegal fishing, and promote the long-term metric of effort. Works like Jennings and Lee (2012) have
sustainability of marine resources. This innovative solution demonstrated the effectiveness of using vessel monitoring
has the potential to enhance the profitability of the fishing systems (VMS) to estimate fishing effort in large-scale
industry, protect jobs and livelihoods, and contribute to the trawling operations. However, such indicators do not perform
overall health of our oceans. well for passive gears, prompting researchers to seek better
alternatives. The paper focuses on improving the
 Problem Statement: representation of fishing activities in marine spatial planning,
The Indian fishing industry faces significant challenges especially for passive gears used in small-scale fisheries,
due to unsustainable practices, including overfishing, illegal which are often overlooked. By using highly detailed vessel
fishing, and habitat destruction. These factors have led to tracking data, the authors present a new method for mapping
declining fish stocks, environmental degradation, and adverse fishing effort, emphasizing that traditional methods based on
economic impacts on coastal communities. time spent at sea are less effective for passive gears. Instead,
they propose using the length of the vessel's track as a more
Traditional fisheries management approaches have accurate indicator of effort. The study introduces and
limitations in adapting to rapidly changing environmental validates a method to estimate "soak time" (the time gear is
conditions and the complex behaviors of various fish species. left in the water) from tracking data and shows that maps
Static geofencing, while helpful, may not be sufficient to based on this approach differ significantly from those based
address these challenges effectively. on fishing time alone. This new method enhances the ability
to manage fisheries sustainably by providing better spatial
 Therefore, there is a Pressing need for a Dynamic insights into the distribution of passive gear efforts, which
Geofencing System that can: account for a significant portion of global fish catches.

 Accurately predict fish distribution based on real-time data The research titled "Passive Georeferencing: A
and environmental factors. Promising Approach for Finding Probable Fishing Grounds"
 Create and update geofences in response to changing by Eldho Varghese, J. Jayasankar, Pratibha Rohit, Somy
conditions and species behaviors. Kuriakose, K. G. Mini, Grinson George, Vinay Kumar Vase,
 Guide fishermen towards sustainable fishing grounds Reshma Gills, Shelton Padua and A. Gopalakrishnan
while preventing them from entering protected areas. discusses the probable locations to find the fishing spots from
 Support effective enforcement of fishing regulations and the known data, Passive georeferencing refers to determining
reduce illegal activities. geographic information from indirect sources, such as
 Contribute to the long-term sustainability of the Indian satellite images, vessel movement patterns, or environmental
fisheries industry and protect marine ecosystems data, without active input from fishermen or direct tagging of
fishing vessels.

IJISRT24SEP1401 www.ijisrt.com 2027

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

This approach being static in nature, it does not take the  Variable Data:
change in conditions according to the weather and seasons
into account. Although the algorithm proposed here for  Satellite Imagery:
approximating the location for Fishing spots is well studied,
and the difference between the Predicted and Actual Fishing  Sea Surface Temperature: Maps of ocean surface
locations is also presented. temperatures to identify areas with suitable conditions for
fish species.
III. METHODOLOGY  Chlorophyll-a Concentration: Indicates the presence of
phytoplankton, which is a primary food source for many
The model incorporates data from multiple sources and fish species.
with different parameters. The collected data can be either  Ocean Currents: Maps of ocean currents to understand the
non-changing (fixed) or real-time which do continuously movement of fish and plankton.
change.  Precipitation related Cloud Imagery: Satellite Images of
the Cloud Patterns over a region. For example, the INSAT
 Constant Data: 3DR is a satellite by ISRO with a capability for Weather
related imagery.
 Seabed Topography: Study and mapping of the  Salinity Level: The water salinity can change to a certain
underwater landscape, including the depth, shape, and levels depending on climatic conditions. For example, the
features of the ocean floor. It is the underwater equivalent SMOS orbiter by European Space Agency have
of topographic mapping on land. Bathymetry helps capabilities to monitor Ocean Salinity.
scientists understand the physical structure of the ocean,
including its valleys, mountains, and plains.  Weather Related Real Time Forecasts:
 Species specific preferences: Every fish species have its The Meteorological Department Issued Weather
own preference for the water depth, temperature, salinity Forecasts and Alerts, and the real-time weather that affects
level, and type of food. the fish movement and can be a potential threat for
 Historical Data: This includes past records of fishing fishermen.
activities, environmental conditions, and fish distribution.
While this data is collected over time, it is generally  Temperature: Air temperature and water temperature can
considered static for the purposes of the geofencing influence fish behavior and distribution.
system. Continuous addition is made to the dataset of Past  Wind Speed and Direction: Wind can affect ocean currents
trends and Behaviors of Fish species and seasonal and wave patterns, which can impact fish migration and
changes. habitat suitability.
 Geographic Information: Data such as coastline  Precipitation: Rainfall and snowfall can influence water
boundaries, protected areas are generally constant over temperature, salinity, and nutrient levels.
time.  Atmospheric Pressure: Changes in atmospheric pressure
can affect ocean currents and weather patterns.
 Humidity: Humidity levels can influence the evaporation
rate of water, affecting salinity and temperature.
 Cloud Cover: Cloud cover can affect water temperature
and solar radiation, which can influence fish behavior.

Fig 2 Seabed Topography (Credits: Abdullah Hisam Omar)

IJISRT24SEP1401 www.ijisrt.com 2028

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

Fig 3 Sea Surface Temperature

(Source: NIWA) (Representational Image)

Fig 4 Chlorophyll-a Concentration in Coastal India Observed over Intervals.

(Source: incois.gov.in)

IJISRT24SEP1401 www.ijisrt.com 2029

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

Fig 5 Cloud Pattern by Weather Satellite INSAT 3DR

 Data Collection and Preprocessing: The proposed model is automated, hence for the real
The Data from the above mentioned sources is gathered time data the geofences should be created with minimal
and pre-processing is carried out, which involves handling human intervention.
missing values, outliers, and inconsistencies. This makes the
data uniform and suitable for analysis.  Model Training:
The Random forest algorithm is used for developing the
The data is normalized and is further utilized for feature model for dynamic geofencing.
extraction.
Random Forest is a powerful ensemble learning
 The Relevant Features here Include: algorithm that combines multiple decision trees to make
predictions. This approach helps to reduce overfitting, a
 Distance from coastlines common problem in machine learning where a model
 Depth becomes too complex and fits the training data too closely,
 Sea surface temperature leading to poor performance on new data. Each decision tree
 Salinity in a random forest is trained on a random subset of the
 Ocean currents features and a random subset of the training data. This helps
 Historical fishing activity to prevent the trees from becoming too dependent on any
particular feature or set of data points. When making a
 Protected areas
prediction, the forest combines the predictions from all of its
 Weather conditions
trees using a voting or averaging method.

IJISRT24SEP1401 www.ijisrt.com 2030

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

Fig 6 Random Forest Algorithm

The Dataset is partitioned here based on its data source,  This Approach has the Following Advantages:
and the decision trees are formed for each data subset. This
allows each tree to specialize in capturing patterns from a  Leverages Diverse Data: By using separate decision trees
specific data source. for different data sources, you can capture the unique
insights from each type of data.
Each Decision tree will be trained on the respective  Reduces Overfitting: The ensemble nature of random
subset. The methodology for training an individual tree can be forest can help to mitigate overfitting, which is a common
unique, based on the nature of the data. This diverse ensemble problem when training models on complex datasets.
of models that can improve overall accuracy and reduce  Improved Accuracy: Combining predictions from multiple
overfitting. models can often lead to better overall accuracy than using
a single model.
In the next step, the trained trees would be used for
making predictions for new data points. The data points would  Below are the Summarized Steps for the model Training:
be the sites with the favorable outcome, which can be the
probable spot with high fish concentration, or the breeding  Data Partitioning
spot.  Decision Tree Creation
 Feature Extraction
The random forest algorithm being based on Ensemble  Training and Prediction
Learning approach, the predictions from individual trees are
 Ensemble Prediction
combined using the voting or averaging method. This helps to
reduce the impact of individual tree errors and improve the
overall accuracy of the model. The resultant output will be the
refined predictions of the data points.

IJISRT24SEP1401 www.ijisrt.com 2031

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

Fig 7 Workflow of The System

Upon training the model based on the Random forest  Refine model: Make adjustments to the machine learning
algorithm, the next step would be the generation of the model as needed based on evaluation results.
geofences. Thresholds can be set according to the condition,
about the concentration of fish. The geofences will be in The detailed usage and scale of operation is further
accordance to the availability of fish and protected zones. discussed Into the implementation section.

The fishing vessels can be monitored live-time by IV. IMPLEMENTATION

authorities if needed, in case any unauthorized activity or
violation of protected zones is noticed, the necessary action The model can be re-trained at some given intervals,
can be taken. which can be at a week or a month. This has to be done to
retain the prediction accuracy of the model. The predictions
The frequency to update the geofences can be preset, as would be more accurate as the dataset expands, and a strong
per the season or depending on the fishing activity observed at set of historical records has to be maintained. The process of
a given location. model training being computationally intensive, since
multiple decision trees are formed and the size of dataset
 Evaluation and Refinement: would continue to expand.

 Monitor performance: Continuously evaluate the The scale of Implementation is dependent upon the
performance of the geofencing system to identify areas for computational capabilities, the larger the area for geofencing,
improvement. bigger is the dataset. If the coastline of the state Maharashtra
is taken into account, it spans for 720 kilometers, from the

IJISRT24SEP1401 www.ijisrt.com 2032

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

Gujarat state at the north to the Goa state at the south. The fish activity varies differently for every species. Some species
coastal districts in Maharashtra are Palghar, Thane, Mumbai possess high rate of migration, so their probable location of
Suburban, Mumbai City, Raigad, Ratnagiri, Sindhudurg. migration should also be studied by the model, as the fish
specific information is also used as a data source.
The satellite sourced data, such as the Sea Surface
temperature, Salinity Level, Weather Related Data, Another aspect of Implementation would be about the
Chlorophyll-a concentration, etc. are measured at different species of fish. For the example of Maharashtra, the
scales, as per the specifications of the satellite. Then it can demanding fish species are King Mackerel, Pomfret, Bombay
range from a few kilometers to a few hundred meters. Ideally Duck, Indian Mackerel, and many others.
for the oceanic tides, a medium resolution data would be
suitable with a larger scale, whereas for the data such as sea The fish species like the Pomfret do not migrate for long
surface temperature, and chlorophyll-a concentration, a high distances and is generally found in shallow waters, whereas
resolution data would be more appropriate. As the fish the King Mackerel are highly migratory species and may
concentration can be measured at a smaller area. travel for long distances and have good tolerances of water
temperatures. The Bombay duck has a lesser market value and
If the system for geofencing is set up on district level, it is easily available in abundant quantities in shallow waters
would be the most ideal scenario. In a state like Maharashtra, with a sedentary pattern.
with a long coastline, almost each district has a coastline of
110-130 kilometers, which would a suitable range for As the fish species have different behaviors, the system
observation. Also it would be easier for the personnel can provide suggestions and probable locations as per the
managing the system than a centralized system for the entire species of fishes. The predictions must not lead to over
state’s coast. exploitation of a certain species of fish, so the scale of
predictions must not be much accurate.
In the historical data, it can be limited for the past 5
years at max. Maintaining legacy data older than that would If the fishermen try to enter a zone marked with
be computationally expensive while model training, also due conditions for fish breeding, immediate alerts must be given
to global changes in the environment, including the global to the fishermen. Since the positioning system is Satellite
average temperature or the precipitation levels the older data based, it would continue to work even in locations farther
becomes lesser relevant for inference in this scenario. It may from the coast. The GPS tracker can be fit to the boat and
affect the accuracy of the model. paired with the smartphone device or would use the available
sensor on the device.
The system can be in the form of a website or a
smartphone application, where the map with the locations will The implementation of this system is complex on certain
be made available for the fishermen. For the system to be aspects and requires multiple factors to be taken into
accessible for a wider section of fishermen, ranging from consideration.
larger sized trawlers entering in deep sea, to small scale
fishermen with smaller boats. V. CONCLUSION

The Large boats are sometimes equipped with onboard The approach towards sustainable and profitable fishery
systems, for communication and other navigation presented in this research potents to bring a positive change in
applications, the geofencing system can be integrated with the fishing industry. The actual taking into application for this
these systems. methodology will take a large amount of resources, including
the presence of Orbiting Satellites with the necessary
The most useful implementation of this systems would capabilities to produce data useful for Training the Machine
be in the mobile devices; it would be accessible for a wider Learning model. Currently the static Geofencing is being
section of people involved in fishery. The mobile application taken into effect into fishery. The adaption of dynamic
for this system should have a capability to load an offline geofencing would take a significant time ranging into years.
version of the geofenced map, before the fishermen begin the This would need a complete overhaul of technologies, on both
journey into the sea. Most internet services continue to work the authority and fisherman point of views. The actual
up to a distance of 10 kilometers from the coast. So, till the effectiveness of this approach can be studied in further
point the device gets network, the tracking data of the device researches, provided the necessary sources of data and
will be recorded into the system, and after the device loses availability of satellites positioned in orbit with necessary
internet connectivity, the application will locally store the path instruments for observation of oceanic bodies.
of the boat, which will be uploaded into the system upon
entering into coverage area. This recorded information can be REFERENCES
useful for the fishermen to track and maintain the records of
the locations they visited. [1]. Eldho Varghese, J. Jayasankar*, Pratibha Rohit, Somy
Kuriakose, K. G. Mini, Grinson George, Vinay Kumar
The geofences being dynamic, would be updated at Vase, Reshma Gills, Shelton Padua and A.
intervals of days, or hours. So the offline version of map Gopalakrishnan. Passive georeferencing: A promising
stored on devices would held valid for the given interval. The approach for finding probable fishing grounds.

IJISRT24SEP1401 www.ijisrt.com 2033

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1401

[2]. T. Mendo, G. Glemarec, J. Mendo, E. Hjorleifsson, S.

Smout, S. Northridge, J. Rodriguez, A. Mujal-Colilles,
M. James. Estimating fishing effort from highly
resolved geospatial data: Focusing on passive gears
[3]. C. Hentry* and S. L. Rayar, S. Saravanan, N.
Chandrasekar and A. Ponniah Raju, K. Kulathuran.
Application of Gps in Fisheries and Marine Studies.
[4]. Technologies for Improving Fisheries Monitoring by
Rod Fujita, Christopher Cusack, Rachel Karasik,
Helen Takade-Heumacher and Colleen Baker.
[5]. Mahima Gokhe1, N.M. Wagdarikar2, Ishita Sharma3,
Atul Kumar4. Embedded systems for Geo fencing and
Rescue System for Fisherman (Fisherman Protection
System)
[6]. M.B. Mukesh Krishnan, D. Saveetha, A. Arokiaraj
Jovith, P. Rajasekar. Fisherman Navigation and Safety
System
[7]. Ming Kun Tan, Muzzneena Ahmad Mustapha.
Application of the random forest algorithm for
mapping potential fishing zones of Rastrelliger
kanagurta off the east coast of peninsular Malaysia.
[8]. B. Padmaja, K. Mounika, P. Sowjanya, P.V.S.
Keerthana, K. Tejaswini. Catching Illegal Fishing
Using Random Forest and Linear Regression Models.
[9]. Alessandro Colombelli, Jacopo Pulcinella, Sara
Bonanomi, Emilio Notti, Fabrizio Moro, Antonello
Sala. Fishing the waves: comparing GAMs and
random forest to evaluate the effect of changing
marine conditions on the energy performance of
vessels.
[10]. Simone Vincenzi a, Matteo Zucchetta B, Piero
Franzoi B, Michele Pellizzato C, Fabio Pranovi B,
Giulio a. De Leo D, Patrizia Torricelli B. Application
of a Random Forest Algorithm to Predict Spatial
Distribution of the Potential Yield of Ruditapes
Philippinarum in the Venice Lagoon, Italy.

IJISRT24SEP1401 www.ijisrt.com 2034