For my project, I wanted to see how CO2 emissions could affect mosquito activity. To do this, I set up 4 water bottle traps. 2 of the traps contained a solution of tap water (1 cup), sugar (1/4 cup), and yeast (1 gram). The yeast and the sugar react to release carbon dioxide. (Fermentation) The other two bottles contained a basic solution of tap water and grass. One bottle of each type was placed in two different AOI locations.
Brief hypothesis: If mosquitoes are naturally attracted to carbon dioxide, then the CO2 emitting trap will have greater mosquito activity.
Each bottle also had rocks for stabilization and a stick to act as a breeding location.
The bottles were labeled A1, B1, A2, and B2
The traps also included a funnel to prevent larger insects from entering.
The bottles pictured on the left were the ones that would eventually be interfered with.
However, after one week, there were some experimental errors. Two of the bottles that were located in a forest area near my house* were tipped over. After investigating the situation further, I reasoned that some sort of animal must have interfered with the experiment. (Since the weather had been fairly consistent)
After thinking about the situation, I came to the conclusion that perhaps spreading the traps between two AOI locations could lead to unnecessary errors caused by animal obtrusions. So, after remaking the traps** I relocated them to the first AOI location.
Some observations/conclusions from the failed bottle traps:
I will analyze and refill the cup traps weekly. Additionally, I will make daily observations.
As for the blog, I will update it with any new observations and data.
* My house was one AOI coordinate and a forest area near it was the other
** The traps were remade to prevent timeline inconsistencies in the data. Since we didn't have any more plastic bottles, I had to use cups (Which contained the same solutions that were in the bottles.)
Update: Mon, Jun 27
During the weekly check, I found a large number of ants in and around the mosquito traps. Unfortunately, there were no larvae today.
Solving any experimental errors:
Results: 0 larvae (total)
The lack of larvae may also be due to the limited volume of each trap. If there are still no larvae next week, I may add two additional larger traps
Update: Thurs, Jul 14
Learning Through Failure
As we approach the end of the 7 weeks of citizen science experimentation, I've concluded that the cup traps were completely ineffective. It seems that the area of a mosquito trap truly does affect mosquito activity.
Despite the initial three weeks of experimental errors, the cup traps were relatively undamaged throughout the rest of the study. The only explanation for the lack of larvae has to be due to the trap's size.
In the last week, I attempted to introduce two new traps to the experiment to test if the area was really the issue.
This time, daily loss of water was no longer much of an issue. Still, I couldn't identify any larvae. However, this new idea fascinated me.
I think that overall my experiment has been a disappointment, so I plan on revisiting this question sometime later in the year, next time making sure to use larger traps.
It's disheartening to think about how I could have provided more useful data for research projects by using more effective traps. I hope that in the next round of experimentation I'll be able to provide more support for future SEES projects. Additionally, I will continue to use the MHM habitat mapper even after I'm done testing out new traps, making sure to take observations in every applicable location.
So, as I submit my 0s into the ArcGIS database, know that I will be doing all I can in the future to support scientific research through the citizen science capabilities of GLOBE.
Update: Jul 21 (Last entry)
What did I end up doing with all this data?
Using the data collected from my experiment and from thousands of similar experiments across the world, we* decided to create a research project focused on building different Convolutional Neural Network architectures to distinguish between different types of mosquito larvae.
We built, trained, and tested four different model architectures using the GLOBE mosquito mappers database:
- Consists of 5 Convolutional Layers and 3 Fully Connected Layers.
Source: Understanding AlexNet | LearnOpenCV #
- 3 convolutional layers, 2 subsampling layers and 2 fully connected layers.
Source: LeNet-5 Tutorial: Architecture, Features, and Importance | Analytics Steps (And ResearchGate Images)
- 34-layer plain network architecture inspired by VGG-19 in which then the shortcut connection is added. These shortcut connections then convert the architecture into a residual network.
Source: Residual Networks (ResNet) - Deep Learning - GeeksforGeeks (And ResearchGate Images)
- A stack of multiple (usually 1, 2, or 3) convolution layers of filter size 3 x 3, stride one, and padding 1, followed by a max-pooling layer of size 2 x 2, is the basic building block for all of these configurations. (Basic VGG architecture)
- VGG-16 was the highest performing configuration
Source: What is VGG16 - Convolutional Network for Classification and Detection (mygreatlearning.com)
About the author: Dhilan is a Junior at Lake Travis High School, Austin, TX. This blog describes a mosquito trapping experiment conducted as part of the NASA STEM Enhancement in the Earth Sciences (SEES) summer high school research internship. His virtual internship is part of a collaboration between the Institute for Global Environmental Strategies (IGES) and the NASA Texas Space Grant Consortium (TSGC) to extend the TSGC Summer Enhancement in Earth Science (SEES) internship for US high school (http://www.tsgc.utexas.edu/sees-internship/).