Possible Uses for Animal Tracking and Sensor Data

Ragnar Stølsmark and Erlend Tøssebro

Institute of Electrical Engineering and Computer Science, University of Stavanger, Stavanger, Norway

Keywords: Animal Tracking, Sensors.

Abstract: What other uses are there for animal tracking? We argue that the location data can be made more useful by

analyzing it to determine the future position of animals and optimize farming. Other sensors such as

temperature and RFID could be added to increase benefit for both farmers and scientists. Adding more

sensors is not trivial and major challenges include energy consumption, cost, size and establishing a market.

However these obstacles can be dealt with and the possibility of having mobile sensor platforms available to

the public is intriguing.

1 INTRODUCTION

The invention of the GPS technology has

revolutionized animal tracking. Animal tracking

used to be reserved for scientists tracking radio

tagged animals with big antennas. Now, with

inventions such as Telespor (Thorstensen et al.,

2004), farmers are tracking their animals and are

able to easily locate them over the internet. This is of

course a big improvement over the way it used to be,

when the farmer could spend weeks walking in

difficult terrain trying to retrieve his animals.

Large scale animal tracking generates vast

amounts of data. A sheep flock can consist of

several hundred animals that have their position

logged multiple times each day for approximately

three months. It is our opinion that using this data

only to establish the location of animals is a terrible

waste. Just by performing some simple analysis it

could be possible to improve farming and create

models of animal movement.

Adding different sensors to the animal is

certainly possible if a modular platform such as

Libelium’s Waspmote is used. This will open up a

range of new possibilities such as accurate local

weather forecasting and predator detection. However

it has far greater challenges and is therefore also

more difficult to achieve than simply adding an

analysis layer on top of the animal location data.

This article is focused on sheep and sheep

farming, since that is what we have previous

experience from. We feel however that many of the

observations are true also for other types of animals

such as cows, goats and horses.

This paper is structured as follows: Section 2

covers related work. In section 3 different uses for

animal location data is described, including farming

optimization and location extrapolation. Section 4

shows examples of what can be achieved by adding

different types of sensors. The fifth section discusses

difficulties concerned with adding more sensors.

Section 6 concludes the paper.

2 RELATED WORK

The electronic shepherd research project

(Thorstensen et al., 2004) culminated in the

commercial Telespor product, which is the

technology that Norwegian sheep farmers use to

track their sheep. It has become quite popular and is

also being used for cows. The technology is based

on obtaining the position of the sheep via GPS and

transmitting it via GSM to a central server. The

farmer can then view the position of his sheep on a

web site. This technology in itself already generates

location data and can therefore be used for those

things we suggest in the next section. It is not

modular and therefore will need to be modified in

order to support additional sensors.

In the survey on wireless sensor networks by

(Akyildiz et al., 2002), they list a lot of the possible

sensors that can be used in a wireless sensor

network. These include temperature, humidity,

movement, pressure and noise among others. This

244

Stølsmark R. and Tøssebro E..

Possible Uses for Animal Tracking and Sensor Data.

DOI: 10.5220/0004315402440249

In Proceedings of the 2nd International Conference on Sensor Networks (SENSORNETS-2013), pages 244-249

ISBN: 978-989-8565-45-7

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

survey, although a bit old, is also a good

introduction into many of the typical wireless sensor

network problems such as energy consumption, cost

and size issues.

Animal movement analysis is studied by (Jonsen

et al., 2003) in an article concerned with how the

data from GPS tracking can be used to create animal

movement models. It uses state-space modeling to

model animals’ interaction with the environment

while also dealing with measurement error. This

method could be used if one would try to make a

movement model for sheep based on the data from

Telespor. State-space modeling for creating animal

movement models has also been studied by

(Patterson et al., 2008).

Finkenzeller and Müller has written a good

introduction (Finkenzeller and Müller, 2010) into

near-field communication (NFC) and RFID which is

recommended for those interested in these emerging

technologies.

There is a strong tradition for farming

optimization. An example is the paper by (Mulder et

al., 2006) which tries to optimize cattle breeding by

testing out different breeding strategies and

determine which strategy leads to the highest milk

yield. In the article by (Zervas, 1998) he argues that

there is a need for a collective optimizing strategy to

best utilize the Greek pastures. A perfect mix of

sheep and goats should be used to maximize the

potential of the grazing areas.

3 USES FOR ANIMAL

LOCATION DATA

An animal tracking system delivers a set of time

stamped locations for each animal. This makes it

possible to see the last known location of the animal

and also create a trace of the animal’s movement

from the beginning of the tracking. But with some

further analysis of the data, it is possible to increase

the utility of such a system.

3.1 Location Extrapolation

Animals’ movements are seldom completely random

as they are conscious beings with certain needs.

Examples of such needs are the need to sleep, eat

and drink. This leads to patterns in their movement

that can be used to extrapolate their current or future

position based on the location data and a movement

model. In its simplest form this location

extrapolation can simply be to assume constant

velocity and calculate a future location using the last

known speed and direction.

Using knowledge of animal behavior it is

possible to obtain far better predictions. Many

animals sleep at night. Therefore their position at

dawn will be close to where they ended up the

previous evening. This can be used both in location

extrapolation, by implementing a no movement at

night rule, but also in duty cycling. Since the

animals do not move at night, the tracking device

can be switched off during this time and rather have

more updates during the day.

To predict the future location of the animal

weeks in advance, knowledge of animal behavior

and tracking data from previous years becomes

important. Their relative importance compared to

last measured location and direction should increase

the further into the future one tries to predict. Sheep

can be used as a good example of how animal

behavior can be used in far future prediction. Sheep

prefer to eat the more nutritious early spring grass.

To get to this grass, the sheep moves higher and

higher up into the mountains as the summer

progresses. This can be used to predict the altitude

the sheep will most likely be at in a few weeks.

Knowing the altitude it is possible to rule out areas

that are either too high or too low. For such a system

to work each animal type should have its own

behavior model. Including previous year’s location

data to adjust the behavior model on an individual

level would probably also be important. Animals are

individuals and some prefer to stay with the flock

while others are lone wolfs and do not come down

from the mountains until the winter sets in.

To create a behavioral model it will be important

to have one set of location data to use while

developing it and another set of data to verify the

validity of the model. It will also be important to test

the accuracy of the predictions on an animal species

basis. Some species have more regularity in their

movement than others and can therefore be predicted

further into the future. If it is even possible to give

any meaningful predictions more than a few days in

advance is an open question that requires further

investigation.

3.2 Farming Optimization

Farmers have a strong tradition for optimizing the

farms annual output. In case of sheep farms their

output is measured in the amount of meat they

produce. An example of such an optimization is the

strategic breeding that has led to sheep that produce

more meat and also meat of a more desirable quality

PossibleUsesforAnimalTrackingandSensorData

245

in terms of fat percentage. We argue that the

location data can be used to optimize farming in

several ways.

Location data combined with knowledge of an

animal’s weight before and after the grazing season

should be analyzed to establish the best grazing area.

If one area over time sets itself apart from the rest,

by providing a larger weight gain among the animals

that have grazed there, measures should be taken to

optimize the time the animals spend in that area.

This optimization can be achieved in at least three

ways. The animals can be released as close as

possible to this area. This will lead to more animals

discovering the good pastures and therefore graze

there. Selective breeding is also an efficient way of

controlling where the animals go to graze, since

small animals learn from their mothers where to go.

Therefore the grazing pattern of the animals should

be taken into consideration as one of the parameters

when deciding which animals should be allowed to

breed. The grazing area can also be affected by the

farmer by making it more desirable for the animals.

This can be done by easing access via improved

trails or by putting the salt stone in that area. The

salt stone is a block of salt that the farmer place in

the grazing area so that the animals have access to

salt. The animals will go to lick the salt stone quite

often. Therefore it is quite a popular spot for the

animals and it makes sense to put it in the best

grazing area.

When the grazing season ends it is time to collect

the animals. Already this is greatly simplified by

using an animal tracking system, but we think that

there is still room for improvement. It should be

possible for an animal tracking system to

automatically find the shortest possible path to

retrieve the animals. This is not as trivial as it might

sound. To get a good solution it would have to take

into account many factors: The start and end

location, the presence and quality of trails in the

area, the future position of the animals at the time of

collection and the benefit of leading a smaller flock.

There could probably be other factors as well. It

would certainly be beneficial to the farmer to get a

suggested path as the areas can be large and difficult

to walk across and therefore choosing the wrong

path can lead to many hours of extra work.

3.3 Other Applications

The location data collected from the sheep can also

be used in non-farming related purposes. An

example is automated collection of trails and paths.

The sheep follows paths through the landscape and

these can be mapped by a simple algorithm. The

hypothesis is that the more the trail is used by the

sheep, the better the trail. There are two reasons for

this. First, the sheep is an animal that prefers

comfort and will therefore choose the best trails

from A to B. Secondly; the sheep creates their own

trails through use alone. Therefore the more use the

larger and better the trail. These trails should

therefore be ranked based on their use and could

then be presented to the public who wants to plan a

hiking trip.

4 OTHER SENSORS

It is possible to attach sensors to an animal that can

be used in combination with the location data to

provide interesting information for the farmer as

well as others. This section covers different sensor

types and the benefit of adding them.

4.1 Temperature Sensor

The use of a temperature sensor can be divided into

two categories depending on the sensors placement.

The sensor could be placed as an external sensor that

measures the ambient temperature. It is also possible

to attach an internal sensor that monitors the

animal’s body temperature.

The external sensor is interesting as a tool in

weather forecasting. The data collection is

synchronized and performed multiple times in many

locations. This gives a unique possibility to create

accurate local weather models about how the

temperature changes with the terrain. This kind of

data collection would be almost impossible to

achieve any other way. It is easy to get synchronized

temperature readings from stationary weather

stations and a scientist walking with a thermometer

can add mobility. However, having 500 scientists

walking around for 3 months, doing regular

synchronized temperature readings, is impossible.

The animals could easily have other weather related

sensors attached as well. Examples include a

barometer to measure atmospheric pressure and a

moisture meter to detect rain.

An internal body temperature sensor could be a

useful feature for farmers. A lot of animals die each

year from diseases. By monitoring the animal’s body

temperature it could be possible to detect fever. The

farmer could then take the appropriate action. This

could mean seeing the animal to administer

treatment or retrieving it back to the farm for

observation. It could also be used to improve animal

SENSORNETS2013-2ndInternationalConferenceonSensorNetworks

246

death detection. This can be achieved today by

movement detectors, but they have a longer

detection time. They require the animal to remain

stationary for 24 hours. The long detection time is

unavoidable since it needs to distinguish between

sleeping and dead animals. A temperature sensor can

alert the farmer the moment the internal temperature

of the animal drop below normal values. This could

save animals lives in a situation where they have

been attacked by predators. If some dies, the farmer

would be alerted much quicker. He can then tend to

those that are wounded.

4.2 Gas Sensors

A gas sensor can measure the concentration of

different gases in the air. Since some animals

wander around in forests, it could be possible to

attach a gas sensor to detect forest fires. Early

detection is essential to minimize the damage caused

by such fires. Therefore it would be very useful to

protect the most fire prone forests by having animals

walk around as mobile fire detectors. During a forest

fire a nearby gas sensor would detect a sharp

increase in carbon monoxide (CO) and carbon

dioxide (CO

) levels. This alone could be sufficient

as a detection device. Combining it with an increase

in temperature measured by a temperature sensor

would further enhance the possible forest fire

detection rate. Such a combination should be of the

type where a fire is detected if the sum of gas

concentration and temperature increase is higher

than some defined threshold. This would make it

possible to detect situations where there is a lot of

smoke, and therefore a high CO concentration, but

not yet a great increase in temperature. The

temperature sensor can help in detecting fires when

the wind is blowing the smoke away from the

animal. An animal-based forest fire detection system

would have to go through rigorous testing before

being deployed. These tests would have to include

the particular animal that the system is intended for.

Without animal-specific tests it is impossible to tell

how close the animal would go to a nearby fire. If

the animals generally run away before the sensors

can detect fires the system is of very little use as it

could only detect fires that the animal could not

escape from.

4.3 Microphone

A microphone can be used to listen for, and

recognize, special noises in the environment. The

question is what such a system should listen for.

Since animals wander around in the forest or

mountains an obvious choice is other animals or

humans. The microphone could for instance try to

detect the presence of predators such as wolves or

foxes. This could be useful for public statistics. If an

area has too many predators, maybe a choice has to

be made between killing some of the predators or

ban farmers from letting their animals graze in those

areas. Birds are another thing that could be detected

by a microphone. Since they often make very

characteristic noises that are loud enough to be

carried some distance, they are the perfect match for

microphone-based detection. The information about

the presence of birds in an area can be interesting for

researchers as well as bird watchers. Also if it is a

predator bird such as an eagle, it could be interesting

for the farmer as well.

4.4 NFC or RFID

NFC and RFID are similar technologies since they

both work over distances under a few meters and

they only require one of the communicating parties

to have power. Either of them can be used in an

animal tracking network as proximity sensors. In the

sheep tracking example it could be used to ensure

that the lambs are still with their mother. It could

also be used to check how frequently a particular

spot is visited without continuous GPS tracking. In

sheep farming this could be used to evaluate a salt

stone placement. By placing a NFC or RFID reader

next to the salt stone, the chip-equipped sheep could

be counted. If few sheep use the salt stone, it can be

moved to a new location. Similar applications are

sure to exist in other types of animal monitoring as

well. Animal proximity can also be used to trigger

some event. In cattle farming it is already used to

control that each cow gets the right amount food and

do not steal food reserved for others. It could also

trigger cameras so that it would be possible for the

farmer to see images of their animals while they are

grazing. This could be used to determine their

current size which is an important economic factor

as the income of the farmer is dependent on the total

weight of the animals that are sold at the end of the

season.

5 DISCUSSION

As have been documented in the previous chapter

there are a lot of sensing tasks that can be performed

by animals equipped with modular sensor boards.

However there are four main obstacles: energy

PossibleUsesforAnimalTrackingandSensorData

247

consumption, size, cost and establishing a

marketplace.

Energy consumption is a major obstacle,

especially if multiple sensors are added. Some

sensing tasks also require the sensor device to be

switched on all the time. This is not compatible with

scenarios where animals are without supervision for

several months. Having to change the battery on all

the animals mid-season will in most cases be

unacceptable. There are two solutions to this issue.

The easiest is to fit a bigger battery. This solution is

limited in two ways. The first is size. Big batteries

are bulky and heavy. Forcing the animals to carry

these for months on end could be considered as

animal cruelty. The second problem is price as high

capacity batteries are quite costly. The other solution

is energy harvesting. As this study (Nadimi et al.,

2011) shows, an animal, such as a sheep, generate a

lot of energy while grazing. This could be used to

power the sensors. This technology is relatively new

and therefore it should be expected to become better,

cheaper and smaller over the next years. This will

most likely make energy harvesting a better solution

than using big batteries.

A similar problem to energy consumption is the

issue of size. If too many sensors are put on an

animal it is not possible for it to comfortably carry

the equipment. The size limit is of course animal-

specific. A big cow can carry more than a small

goat.

Cost is also a big factor. The cost of adding

another sensor cannot exceed the perceived benefit

of adding that sensor to the party responsible for

paying for it. Adding a temperature sensor to

measure the weather is not important to most

farmers. Therefore the cost of adding a new sensor

must be paid by those interested in using the data

produced by that sensor. This would involve the

weather service paying the farmer to be allowed to

put temperature sensors on his animals. This leads to

the last problem, the establishment of a common

marketplace where farmers with free sensor capacity

can offer that capacity to customers willing to pay

the extra cost of adding more sensors. A web-based

marketplace where customers can browse based on

area and sensor type would seem like a good option

for this kind of service. It should also be possible to

discover already existing sensor deployments to

allow new customers access to those data, for a

price.

6 CONCLUSIONS

Animal tracking is already being used by both

farmers and researchers to locate animals. By

applying some logic, it is possible to add value to the

location data. Location extrapolation to predict

future locations and optimized farming is just a

couple of examples. Since this can be done with the

equipment being used today, this is merely a

question of implementing the logic into the systems.

Adding more sensors could also be beneficial.

The best candidates are those that can work well

with discrete measurements, since they require less

energy. Temperature sensors and RFID stands out as

sensors that could be quite easily added and are very

useful both to farmers and scientists. Generally

adding more sensors are more complicated than

doing more reasoning on the location data.

Challenges such as energy consumption, cost and

size are all serious concerns. The biggest challenge

however is to establish a market for such a solution.

To be successful it needs to be easy for farmers to

offer sensor capacity to customers and for customers

to find available sensors in an area. On top of all this

it needs to be affordable. It is hard to tell if such a

service is economically viable, as that depends on

the application of the sensor data, but it would

definitely be interesting to see it developed.

Something so radically different could easily spawn

new inventions only made possible by effortless

access to mobile sensor data.

REFERENCES

Akyildiz, I. F., Su, W., Sankarasubramaniam, Y.,

Cayirci,E., 2002. Wireless sensor networks: a

survey.Computer Networks, 38(4), 393-422, Elsevier.

Finkenzeller K., Müller, D., 2010, RFID handbook:

Fundamentals and applications in contactless smart

cards, radio frequency identification and near-field

communication, Wiley.

Jonsen, I. D., Ransom, A. M., Flemming, J. M., 2003.

Meta-analysis of animal movement using state-space

models. Ecology, 84(11), 3055-3063, Ecological

society of America.

Mulder, H. A., Veerkamp, R. F., Ducro, B. J., van

Arendonk, J. A. M., Bijma, P., 2006. Optimization of

dairy cattle breeding programs for different

environments with genotype by environment

interaction, Journal of Dairy Science, 89(5), 1740-

1752, Elsevier.

Nadimi, E. S., Blanes-Vidal, V., Jørgensen, R. N.,

Christensen, S., 2011 Energy generation for an ad hoc

wireless sensor network-based monitoring system

SENSORNETS2013-2ndInternationalConferenceonSensorNetworks

248

using animal head movement, Computers and

Electronics in Agriculture, 75(2), 238-242, Elsevier.

Patterson, T. A., Len, T., Wilcox, C., Ovaskainen, O.,

Matthiopoulos, J., 2008. State-space models of

individual animal movement, Trends in Ecology and

Evolution, 23(2), 87-94, Elsevier.

Thorstensen, B., Syversen, T., Bjørnvold, T. A.,

Walseth,T., 2004. Electronic Shepherd – a low cost,

lowbandwith, wireless network system. Mobisys ’04

Proceedings of the 2nd international conference

onMobile systems, applications, and services. 245-

255.ACM.

Zervas, G., 1998, Quantifying and optimizing grazing

regimes in Greek mountain systems. Journal of

Applied Ecology, 35, 983-986, Wiley.

PossibleUsesforAnimalTrackingandSensorData

249