The world is facing the worst locust attack in decades. As of May 3rd 2020, over 5.5 million hectares of land have already been affected by locusts. The swarms are only growing in number with each swarm having up to 80 million desert locusts. A very small part of an average swarm eats the same amount of food in one day as about 10 elephants or 25 camels or 2,500 people.
COVID-19 has exacerbated the situation having impaired various activities like the movement of goods and services. The lock-down imposed across the world has affected surveillance that helps in identifying areas that are potentially vulnerable to invasion by locusts. In addition to this, the supply chain of bio-pesticides involved in control operations has also been disrupted due to restricted movement. This has resulted in places like Somalia seeing delays as much as 3 weeks in receiving shipments of bio-pesticides.
N-Visage has come up with Locust Attack Detector which takes into consideration the various environmental factors like vegetation, rainfall and temperature to identify areas that are likely to be attacked by locusts. This will help farmers and the various disaster mitigation organisations to be well prepared to tackle and mitigate a locust invasion.
India is on the verge of its biggest food scarcity in 25 years, when the last major locust attack happened. If we don’t plan and respond to the impending disaster to the crops caused by these locusts, it might lead to millions of people across the country going without food.
In order to effectively foresee and tackle such a locust invasion, it is paramount to study the various environmental factors that influence the breeding and movement of these insects.
We aim to identify the areas that are likely to be attacked by locusts through factors measuring:
We procured data for cities in 3 countries, namely, Kenya, Saudi Arabia and Ethiopia that have recorded frequent locust attacks in the past to learn about the patterns involved in locust presence.
We relied on the data provided by
We developed 3 classification models for this use-case:
Perceptron
Logistic Regression
Decision Tree
*precision = precision refers to the percentage of your results which are relevant
*recall = recall refers to the percentage of total relevant results correctly
*F-score = measure of test accuracy
Result:
Taking into account, both the Precision and Recall, we have chosen the Decision Tree model for predicting the potential areas that are likely to be attacked. This model was incorporated into the “Locust Attack Detector” website to predict if an area is likely to be affected or not based on the inputs given by the user.
Try the Locust Attack Detector: n-visage.co
Tools Used:
For statistical modelling, we used Pandas, NumPy, SciKit-learn, JobLib libraries with Python as the coding language on Jupyter Notebook.
For building the website, we used Vanilla HTML, CSS and Javascript.
For visualizing the map, we used the Mapbox API.
Future Scope:
We would like to further our understanding of the influence of wind speed and wind direction in the movement of locust swarms and incorporate the same in the "Locust Attack Detector" to predict the potential path of the locust swarms. The vision is to build a full-fledged system that can be utilized globally to increase awareness and preparedness about locusts and its movements