Locust Locator

https://youtu.be/tNhQprsEAy4

https://github.com/Ezard/nestjs-microservice-protoc-plugin

In 2020, African and Middle Eastern nations experienced some of the worst locust infestations for nearly a century. Initial reports on the 2020 invasion by the Nairobi based GAD Climate Prediction and Applications Centre, which monitors 11 countries, suggested that 2,500 square kilometers of crop was damaged in Kenya, largely among marginalized communities. The Famine Early Warning Systems Network (FEWS Net) predicted locusts could damage enough crops to feed 280,000 people for six months in Somalia. In the face of a changing climate, increasingly unpredictable weather is expected to significantly impact marginalized communities, many of which already suffer from existing issues such food insecurity, safety issues, poor education, weak infrastructure, and problems associated with the ongoing pandemic. Locust Locator aims to minimize the impact of locust swarms on communities in Asia, Africa, and the MENA regions by using satellite data, machine learning, and community reporting to predict locust swarm development and movement, and better manage response. 

Adult locusts can eat their body weight every day and fly up to 150km a day in search of new food. Daily, a small swarm, covering just 1km area, can devour the same amount of food as thirty-five thousand people. Locusts are a major threat to existing food insecurity amongst communities, but they also greatly impact mental health and social security as well. Often, children of affected villages are recruited to scream at the pests to scare them away and are forced to skip school to help protect crops. Even governments are at maximum capacity when it comes to fighting locusts as aerial and ground control operations alone have proved insufficient.

We hope to build on already innovative techniques (such as those implemented by UNFAO) of in-situ data gathering and satellite monitoring to help communities tailor their locust responses to their specific needs. Locust Locator pursues a set of methods that can pave the way to a more efficient, safer, and sustainable response to the worsening locust problem. In pursuit of achieving this, we developed a framework consisting of three vital approaches: Track, Preempt, and Communicate (TPC).

Local communities are an indispensable asset to real-time tracking. They are mentally, physically, and economically invested in minimizing the impact of locusts on their lands and crops. We see an opportunity to enhance local efforts and channel them more effectively. Self-reporting by communities and collaboration with UNFAO’s eLocust3 tool can create widespread real-time ability to track the movement of locust swarms.

Such tracking efforts could also be used for the purpose of data collection, which would be valuable for our project’s second approach, preemption, within the TPC holistic framework. Technology such as Machine Learning (ML) can aid in building prediction models (e.g. using Random Forest) that can further help deter swarms before they arrive and damage crops. This would use longitude and latitude coordinates in real time.

Lastly, we consider the social benefits of educating people about the issue, and empowering them to use Locust Locator to tailor solutions to fit their community needs. In order to achieve an easy and comprehensible form of science communication as well as to help educate peoples in areas not commonly devastated by locusts, we plan to utilize Augmented Reality (AR). AR is able to transcend language barriers and educate through visuals, bringing culture and tradition into our understanding of local locust problems and their subsequent solutions.

In order to develop the prediction models, we apply artificial intelligence (AI), as it provides the ability to gradually learn and improve from training sets without explicitly programmed instructions. The job of the modeling algorithm is to find the most applicable mapping function from input variables to output variables and aid in the discovery of rules and patterns in the data sets, existing ones (e.g. NASA), as well as collected ones (i.e. part of our tracking approach). 

It often proves difficult or expensive to make an abundance of observations, in addition to it being challenging to gather all observations if they are expected to be made in the future. In response to such issues, we decided to, via Python, to use the ML algorithm Random forest (RF), which enables making predictions on the basis of numerous decision making ‘trees’.To classify a new object based on attributes, each tree gives a classification and a rank for that class. The classification with the highest rank (over all the trees in the forest) is chosen and in case of regression, the average of outputs by different trees is used for predictions. In Random Forest, each tree is set up as follows: N cases in the training set are defined. Samples of these cases are taken at random but with replacement. They build the training set for growing the decision tree (See graph below for an example of RF process).

As an outcome, this algorithm can help predict the routes of locust swarms, and thereby allows preemptive action to be taken in a timely manner. Our research shows that the safest, most efficient and most optimal way to battle the locust swarms, is to manage to do it during their solitariusphase, prior to their full development.

Desert locust swarms become dangerous when a set of perfect conditions are met which allow them to breed proliferously. By training ML software to track the main contributors to those conditions using satellite data, we can predict when and where locust swarms are likely to form and take the necessary precautions to reduce impact. 

Extreme wet conditions in normally dry areas create the optimal breeding environment for locusts and also abundant vegetative food sources. Rainfall and climate events can be monitored using data sets from NASA Global Land Data Assimilation System (GLDAS) and NOAA Climate Prediction Center (CPC) to predict which areas locusts are likely to breed. NASA Soil Moisture Active Passive (SMAP) data sets show moisture in surface soils and around root zones. Since females lay their eggs in moist topsoil, information from these data sets can be used to detect optimal breeding conditions. Comparing data from NASA/ NOAA’s Global Food Security-Support Analysis Data (GFSAD) Cropland Extent 2015 Africa and similar agricultural datasets can allow for the designation of high-risk areas or areas which are most susceptible to locust swarm migrations because of sufficient food sources.

Our NASA Space Apps experience was both challenging and rewarding. Through the research gathered and discussed amongst our team, we learned a lot about food (in)security from a compelling perspective. Our team includes citizens as well residents of the MENA region, therefore, we decided to focus on a subtopic that not only directly affects the UAE, but that which can indirectly affect it down the line. Locusts swarms have and continue to devastate neighboring countries such as Eritrea, Sudan, Yemen and Egypt.

Given how the majority of our team belongs to marginalized communities, and almost all of us are social scientists and scholars with a great focus on environmental matters, we found this issue particularly interesting for three reasons: 1) how it was very much overshadowed by Covid-19, 2) the little attention it receives by the media and news outlets, and 3) because it affects our regions and hometowns, which rely on its youth to address come up with new solutions.

Some of us were already vaguely familiar with the locusts as pests in the MENA region, but we had not nearly understood the gravity of the situation, and the serious risk it poses on agriculture, and had not imagined how it can additionally negatively impact mental health, children’s education, and overall social welfare. It was equally interesting as it was scary to learn about the problem of locust swarms, and even more so concerning how it is surging and becoming worse, especially due to factors such as climate change.

Our approach to tackling this challenge was first to research and gather information on the issue and its consequences, reading about locusts swarms and understanding their (biological) nature, collect and analyze existing reports and data surrounding the topic, as well as getting up to date on the current technologies and methods used to battle the swarms, while also reading expert reports that offer recommendations and suggestions on possible (new) ways to going forward in a more efficient and safe manner.

Due to our three-part solution possessing a technology-focused part that utilizes AI and ML, as social scientists and academics we experienced the challenge of needing to comprehend how such an algorithm can work. This was very tricky, but proved to be very interesting as well as valuable to complete our holistic approach. We did our own research then came together as a team to brainstorm and present our findings and our thoughts on what kind of program and algorithm we found interesting and/or best applicable to our methodology. We would like to thank our team leader, Vishnu, for getting this team together, keeping us motivated and for his drive to find innovative solutions. 

Climate Prediction Center [https://www.cpc.ncep.noaa.gov/] 

GLDAS [https://ldas.gsfc.nasa.gov/gldas] & [https://earthdata.nasa.gov/] 

FLDAS [https://ldas.gsfc.nasa.gov/fldas] & [https://earthdata.nasa.gov/] 

SMAP [https://smap.jpl.nasa.gov/] & [https://earthdata.nasa.gov/] 

GFSAD [https://croplands.org/app/map?lat=0.17578&lng=0&zoom=2], [https://www.usgs.gov/centers/wgsc/science/global-food-security-support-analysis-data-30-m-gfsad?qt-science_center_objects=0#qt-science_center_objects], [https://earthdata.nasa.gov/] 

BBC [https://www.bbc.com/future/article/20200806-the-biblical-east-african-locust-plagues-of-2020]

UN FAO [http://www.fao.org/locusts/en/], [http://www.fao.org/news/story/fr/item/1397940/icode/] 

eLocust3 [http://www.fao.org/3/i6058e/i6058e.pdf] 

United Nations https://news.un.org/en/story/2020/05/1063672

Yile Ao et al. “The linear random forest algorithm and its advantages in machine learning assisted logging regression modeling”. In: Journal of Petroleum Science and Engineering 174 (2019)

#locusts #foodsecurity #drought #famine

This project has been submitted for consideration during the Judging process.