-
Epic meme (sunglasses emoji)
Inspiration
We found an online guide to pesticide spraying (colloquially knows as the "Pesticide Bible" or "Pible" for short). It gaves us a plethora of information on the correct conditions for farmers to spray pesticides. However, we realized this information could seem very overwhelming for the average farmer, so we decided to build a program that would help farmer's make informed decisions on whether to spray pesticides.
What it does
Pesto-Mode interfaces with IoT devices in a farmer's field via The Things Network using LoraWAN. These devices would have sensors to collect data on temperature, wind speed, wind direction, pressure, humidity, etc. This is combined with the local whether forecast to assess the predicted weather conditions. We then feed this data to a model that gives each section of the field a score on how advisable it is to spray.
How we built it
For our back-end we used Python. We implemented a DataPoint class to represent each data point, a Sensor class to record the data of each sensor in the field, and a field class that contained all the information in one farmer's field. After collecting data we send it to our model which works as follows:
Pesto-Mode takes in a wide variety of factors when calculating the "score" for spraying pesticides on a certain day.
Wind speed: during the day, the ideal wind speed is between 5 and 15 km/h. If it's too windy the pesticides will spread too much resulting in less effective application and more environmental damage. If the wind speed is too small, the pesticides will not get distributed enough.
Wind direction: our web interface allows users to mark bodies of water that are nearby their farm. The model will then give a lower score if the wind is blowing in a similar direction as the water. This is to help prevent leaking of pesticides into the local ecosystem.
Temperature: based on our research, in most cases lower temperatures are ideal for pesticide application. As temperature rises, droplets evaporate faster and air becomes more turbulent resulting in less effective application. Therefore, our model prefers lower temperatures, with a temperature of roughly 60 degrees Fahrenheit being ideal.
Delta-T Value: the delta-t value represents the rate of evaporation of liquid. While we cannot directly measure this, it can be calculated using temperature, humidity, and pressure. A delta-t value between 2 and 12 is preferred with numbers near 2 being ideal.
Pressure differences throughout the field: a similar pressure across the farmer's entire field indicate ideal conditions for spraying. This is because pressure gradients throughout the field will results in unstable surface conditions and volatility of particle movement. Therefore, if a field has multiple sensors in it (which we envision most will), we check that the pressures at the various sensors are similar.
Rain: rain occurring soon after pesticide application is not ideal for many reasons. Firstly, this decreases the effectiveness of the application as many of the pesticides will be washed away. Secondly, as a consequence of this wash-away, the local ecosystem will be flooded with pesticides that are harmful to plants and animals. Therefore, our model looks to avoid rain in the days following pesticide application.
Challenges we ran into
We had little web development experience, and tried to build an advanced website. This resulted in an interesting front-end that can't quite keep up with the capabilities of the back-end.
Accomplishments that we're proud of
We successfully pulled weather data from an API. Analyzed it and created a model that works quite well to advise the use of pesticides. We also got a sample LoraWAN node up and running to demonstrate what a sensor in the field would look like.
What we learned
We need to have a clear plan for interfacing between different parts of our project before we start. We have all of the pieces built but had some trouble integrating it all together.
What's next for Pesto-Mode
We would like to improve the website to both look and function better. Also, we would need to test our sensors in the field. Ideally we could conduct research on pesticide application and crop yield to assess how well our model works.
Log in or sign up for Devpost to join the conversation.