Software apps and online services
Hand tools and fabrication machines
Climate Change has become a global challenge and one of the biggest challenge faced by the current generation and possibly for coming generations too. Climate change is the main cause behind natural disasters such as cyclones, wildfire, super droughts, etc. One of the primary reason behind Climate change is Deforestation. Deforestation is the reduction of Forest cover due to wildfires, human settlement expansion, rapid logging down for resources, agricultural activities etc. Forests play the crucial role of Carbon sequestration. The destruction of forests and the eventual degradation of biomass like plants and soil is the 2nd largest contributor to CO2 and other Green House Gas emissions. This eventually leads to more and more Greenhouse gases getting in the atmosphere, further exacerbating the problem. Furthermore
- Deforestation has led to decrease in area of ecosystems such as Rainforests and Peatlands that are most effective in trapping atmospheric CO2,
- Between 1990 and 2016, the world lost 502, 000 square miles (1.3 million square kilometers) of forest, according to the World Bank—an area larger than South Africa,
- About 17 percent of the Amazonian rainforest has been destroyed over the past 50 years, and losses have being rising as of late,
- As of 2019, deforestation is responsible for about 11% of global greenhouse gas emissions.
It is evident that forests hold the key to combating climate change. There are many preventive methods to combat deforestation. But curative solutions are equally important. And Forestation, which includes Afforestation(creating new Forest covers) and Reforestation(replanting in areas afflicted by deforestation), is the most simple and yet important solution.What is Forestation?
Forestation is planting of trees in areas that naturally had trees(Reforestation) or naturally lacked any forests(Afforestation) in the first place. This involves identifying suitable locations, and analyzing different factors such as soil quality, climate, native biodiversity and human activity. Through meticulous analysis and following effective Forestation techniques like the famous "Miyawaki Method"(by Dr Akira Miyawaki, Osaka University), Holistic Planned Grazing (Professor Alan Savory), etc, forest covers have returned or sprout out in many places, leading to
- Increased soil fertility and ground water retention
- Increase in local biodiversity
- More natural resources, cleaner air and subsequently better health
There is a certain learning curve towards effective Forestation. Necessary scientific knowledge in the domains of Environmental Sciences, Zoology, Botany, etc, is needed to insure that Forests being grown stay for long time and do not have any adverse effects on the local biodiversity. For example, China's Green Wall initiative, that started in 1978, began to combat the desertification from Gobi dessert. Some positive effects have been seen such as
- Slowing of desertification, and
- Fall in the frequency of sandstorms
But, there have also been some issues that have resulted from improper forestation practices. Many of the trees that are planted in areas where they do not grow naturally, simply perish after a few years. Those that do survive can soak up a lot of the groundwater that the native grasses and shrubs need, causing more soil degradation. If forestation in this case continually exceeds the land’s carrying capacity, it will lead the trees to an eventual death. Therefore it is important that the necessary knowledge is readily and easily available to ensure such forestation efforts bear fruit.
Another challenge is that the technical knowledge acts a barrier for common people to take independent active role in Forestation efforts. The scientific knowledge can't be understood by everyone and might lead to hesitancy among people. The technical knowledge needed to ensure a successful forestation is not readily available to common people, who would want to independently go out to places and plant trees, with the best possible results.
There is enough data on the contributing factors for proper Forestation. By using general understanding of the science related to forestation, like soil characteristics, weather conditions, air quality, etc and using the all the related scientific data, we can build a Machine Learning based solution that can not only make forestation efforts quick and effective but also simplify the process. This solution can have software based on the Machine Learning model run against real time sensor data, such as that from PH Sensor or Moisture sensor or Air Quality sensor, to provide data driven recommendation. This recommendation will tell us that in a given environment, which kind of tree is recommended to be planted, so that the trees planted are healthy and have a positive affect on the local environment.
This solution will help
1. Professionals, by reducing their time in analysis and decision making and help them work more quickly, efficiently, and effectively, and
2.Common people, by simplifying the forestation process and hopefully motivate more people do it themselves.
This is Machine Learning based solution is the core of my project "Forestation Assistance Kit".
Forestation Assistance Kit is a ML on Edge solution that is intended to make sound data-driven recommendations on what trees to plant at a location, and as a result making Forestation simple, quick and effective.
The prototype that I have built uses the Quicklogic's Quickfeather and Particle Photon as the brains. The three sensors being used are - a Garden PH meter(couldn't get a proper PH kit due to lockdown), a Capacitive Soil Moisture sensor and a MQ135 Air Quality Sensor. I utilized the Garden PH Meter by pulling out the analog signal pins and sending it to the Photon for processing. I'm also using a Adafruit ADS1015 ADC to send analog sensor data over I2C to the Quickfeather. The whole system is powered by a 9V battery, regulated down to a 5V via a 7805 transistor, with a rocker switch and is perfect for the used case as we are only aiming for limited operation of 2-3 mins until we receive the recommendation.
For my used case, I am targeting 3 trees - Banyan, Neem and Peepal. Banyan is very famous Indian tree, having religious significance. Neem has many herbal benefits and is known to be a air purifying kind of tree. Peepal is also a very widely found tree in India. These trees are common in my area, so that is why I targeted them. I went ahead and did some research about what kind of soil and air conditions are best fit for each plant. From my findings, Banyan needs high moisture, Neem is good for areas with bad air quality, and Peepal grows well in Alkaline soil.
A similar process can be followed for your choice of plant.Hardware Assembly
- Place Photon and Quickfeather on breadboard and connect ground pins between them
- Connect 3.3V and Gnd from Photon to Moisture and Gas sensor and ADS1015
- Connect analog pins from moisture and gas sensors to A1 and A2 pins of the ADC respectively
- Pull out analog pins of the PH meter via opening the back and soldering some wires. Connect the analog positive to A0 pin on the Photon. Then connect the DAC pin of Photon to the A0 pin of the ADC.
- Connect the I2C pins from ADC to the Quickfeather..
- Connect the UART pins of the Quickfeather to the UART1 port(TX and RX pins) of the Photon.
- For power, place a 7405 on the breadboard with a capacitor in parallel to the input-ground pins. Then connect the output and ground pins to the Vin and ground pins on the Photon.
- Connect the 7405 Output to VBat pin on the Quickfeather
- Take a 9V battery connector and connect the positive to a tap switch. Connect other end of tap switch to the input pin of the 7405. Connect the grounds of battery connector and 7405.
Over here I am using jumper connectors to connect the power part and the main electronics. And this is the result!
Now onto the software work!Software Work
For the Photon, the code can be found in the linked Github repo. It is suggested to be programmed via the Particle Web IDE, details for which are here. The code is pretty self-explanatory. Basically, the program sets up some variables, waits for 2 mins for the MQ135 sensor to heat-up, then reads analog readings from the Garden PH meter. It then converts it into one of the 2 analog PH values(100 or 4000) to indicate the nature of the soil. It then outputs that signal via it's DAC pin. Then it reads the result from the ML model running on the Quickfeather, processes it and outputs the name of the tree being recommended by the ML model which can be read via the Particle IOT App or via the user console of your particle account.
For the Quickfeather side of the software, it is necessary that you go through the guides provided by the Quicklogic people here. Then go through this Hackster project to practically try out how to work with the software for the Quickfeather. Futher, to understand the SensiML software better, check their tutorial series.
- First download the Qorc-Sdk as given and perform complete setup. We will be using the qf_ssi_ai_app project in the qf_apps folder.
- Next step is to perform the data collection and labeling for the Tree identification ML model. We first need to make certain modifications in the files of qf_ssi_ai_app project. You can simply replace the files I have provided in the GitHub below following the directions.
- Following the changes, building the executable and flashing it on the Quickfeather (as given in the detailed GitHub document here by the fine folks at Quicklogic).
- No changes in the Device plugin file is needed as we are using 3 analog channels, same as the on-board IMU.
- Attach a USB-TTL(CP2101 given in the kit) convertor to the UART pins of the Quickfeather to stream sensor data to the SensiML datacapture lab.
- Once the sensor data is being received, go on to record the data and save it in your SensiML project.
- Go ahead and label the data for the different types of trees that u are targeting.
- Then go ahead and open up your Analytics Studio and according to the SensiML documentation, prepare your data, build your model and then download the final generated knowledge pack in library format.
- Take this library, and replace the existing knowledge pack in your qf_ssi_ai_app directory. Then build the project again and flash the executable on the Quickfeather.
- The end result will look something like this
You can then remove the USB-TTL convertor and connect the UART ports of the Quickfeather and Photon to finish the procedure.In Action
The completed prototype
And a working demoFurther Work
Here are somethings I think can be improved upon
- Addition of GPS into the ML model prediction
- Better quality sensors
- More data and wider parameter sets so as to make the kit more versatile