Bibliography (for preliminary research only – used for the research plan)

References

Aelion/University of South Carolina, C. M. (n.d.). Soil Contamination Monitoring. UNESCO - ENCYCLOPEDIA OF LIFE SUPPORT SYSTEMS (EOLSS). https://www.eolss.net/Sample-Chapters/C09/E6-38A-05-02B.pdf

Boedeker, W., Watts, M., Clausing, P. et al. The global distribution of acute unintentional pesticide poisoning: estimations based on a systematic review. BMC Public Health 20, 1875 (2020). https://doi.org/10.1186/s12889-020-09939-0https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-020-09939-0

Characterization and Monitoring Technology Guides for Cleaning Up Contaminated Sites. (2023, November 9). U.S. Environmental Protection Agency. https://www.epa.gov/remedytech/characterization-and-monitoring-technology-guides-cleaning-contaminated-sites

Lindwall, C. (2022, July 21). Industrial Agricultural Pollution 101. Natural Resources Defense Council (NRDC). https://www.nrdc.org/stories/industrial-agricultural-pollution-101#causes

University of Sydney. (2021, March 30). 64% of global agricultural land at risk of pesticide pollution? ScienceDaily. https://www.sciencedaily.com/releases/2021/03/210330092530.htm

Bibliography (for in-depth research – used for the research paper)

References

Caruso, A., Chessa, S., Escolar, S., Barba, J., & López, J. C. (2021, November 15). Collection of Data with Drones in Precision Agriculture: Analytical Model and LoRa Case Study. IEEE Xplore. https://ieeexplore.ieee.org/document/9416287

Chavel, I. (2001). Isoperimetric inequalities: Differential geometric and analytic perspectives. Cambridge University Press.

Farooq, M. S., Riaz, S., Helou, M. A., Khan, F. S., Abid, A., & Alvi, A. (2022, April 11). Internet of Things in Greenhouse Agriculture: A Survey on Enabling Technologies, Applications, and Protocols. IEEE Xplore. https://ieeexplore.ieee.org/document/9755156

Gunstone, T., Cornelisse, T., Klein, K., Dubey, A., & Donley, N. (2021, May 4). Pesticides and Soil Invertebrates: A Hazard Assessment. Frontiers. https://www.frontiersin.org/articles/10.3389/fenvs.2021.643847/full

Kumaran, S., Shankar, M. G., Krithikraj, P., & Karthik, M. (2024, January 26). An Intelligent Framework for Wireless Sensor Networks in Environmental Monitoring. IEEE Xplore. https://ieeexplore.ieee.org/document/10404602

Liu, L., Huan, H., Zhang, M., Shao, X., Zhao, B., Cui, X., & Zhu, L. (2019, April). Photoacoustic Spectrometric Evaluation of Soil Heavy Metal Contaminants. IEEE Xplore. https://ieeexplore.ieee.org/document/8664575

Rayhana, R., Xiao, G., & Liu, Z. (2020, March 31). Internet of Things Empowered Smart Greenhouse Farming. IEEE Xplore. https://ieeexplore.ieee.org/document/9051987

Shafique, K., Khawaja, B., Sabir, F., Qazi, S., & Mustaqim, M. (2020, January 28). Internet of Things (IoT) for Next-Generation Smart Systems: A Review of Current Challenges, Future Trends and Prospects for Emerging 5G-IoT Scenarios. IEEE Xplore. https://ieeexplore.ieee.org/document/8972389

Shaikh, F. K., Karim, S., Zeadally, S., & Nebhen, J. (2022, September 27). Recent Trends in Internet-of-Things-Enabled Sensor Technologies for Smart Agriculture. IEEE Xplore. https://ieeexplore.ieee.org/document/9903855

Sänket, R. J., Baviskar, A., Ahire, S. S., & Mapare, S. V. (2018, February 22). Development of a low cost nitrate detection soil sensor. IEEE Xplore. https://ieeexplore.ieee.org/document/8299968

Tang, F. H., Lenzen, M., McBratney, A., & Maggi, F. (2021, March 29). Risk of pesticide pollution at the global scale. Nature. https://www.nature.com/articles/s41561-021-00712-5

Xue, D., & Huang, W. (2020, November 18). Smart Agriculture Wireless Sensor Routing Protocol and Node Location Algorithm Based on Internet of Things Technology. IEEE Xplore. https://ieeexplore.ieee.org/document/9263378

Yu, X., Ergun, K., Cherkasova, L., & Rosing, T. S. (2020, October 2). Optimizing Sensor Deployment and Maintenance Costs for Large-Scale Environmental Monitoring. IEEE Xplore. https://ieeexplore.ieee.org/document/9211489

Zhou, L. (2022, November 22). Environmental Sensor Network Construction and Real-Time Monitoring System for Rural Environmental Pollution Control. IEEE Xplore. https://ieeexplore.ieee.org/document/9952030

Research Plan

Rationale

Anthropogenic contamination has been steadily on the rise for the past centuries. A key industry that has been impacted is agriculture, something that inherently relies on the environment. Additionally, using pesticides, fertilizers, and other chemicals are common practices that have been used to protect crop quality. However, now, 64% of all global farmland is at risk of pesticide pollution. Not only does this pose a severe danger to crop growth and quality, but it also directly impacts consumer health. Around 3.3 million deaths each year are caused by inhaled pollutant particles alone. Note that this does not account for harmful pollutants that may enter the body through other ways, such as ingestion and absorption. Furthermore, farmland contamination can also be devastating to ecosystems, harming organisms and resulting in a ripple effect. For example, it can harm animals who eat the contaminated crops, and then move up the food chain to its predators as well. Afterward, entire ecosystems can be devastated and left ruined by such contamination.

This project proposes a new, innovative soil contamination monitoring system. In the past, technology such as fiber optic chemical sensors, gas chromatography, test kits, mass spectrometry, and X-ray fluorescence have already been incorporated into this process. However, these methods still require manual deployment of sensors or an operator in the field throughout the entire process. Soil monitoring is often a tedious process that is redundant and time-consuming. It requires continuously collecting data over long time periods. Instead of wasting time and energy, machines can help to facilitate this entire process. Furthermore, another downside of manual monitoring is that long-term exposure to chemicals within pesticides can be very harmful. A 2020 study found that 44% of farmers are poisoned by pesticides every year. Therefore, reducing the possible human contact with dangerous chemicals is necessary to protect the health of farmers and other professionals. A safer and more efficient solution is sensors that are programmed to regularly produce readouts. The sensors are connected to drones within a system, with the drones deploying them as well as collecting their data. Eventually, this system can evolve into an Internet of Things (IoT), including sensors, drones, gateways, servers, data storages, and ultimately, the Internet. Such a platform can automate the entire process, with the only human involvement needed being at the final analysis step. With the data stored remotely, researchers and scientists can access it on the Internet and analyze it without having to go through the long and potentially dangerous process of collecting the data on their own.

Research Questions

1. What is the optimal number of sensors needed to maximize land coverage?
2. What is a sufficient number of readouts needed to analyze the data accurately? 
3. What is a realistic detection area of one individual sensor?
4. How much monitoring time is needed to accumulate reliable readouts?
5. How can the need for human manual labor be minimized throughout this process? 

Engineering Goal

A major goal of this research is to determine an ideal number of sensors needed for accurate and efficient monitoring of an area of farmland. This process would be automated by drones, which deploy the sensors and collect their data. To do this, numerous factors such as the farmland perimeter, farmland area, number of deployed sensors, the detection area of each sensor, monitoring time, the time needed for a sensor to produce a readout, total number of readouts, and minimum number of readouts needed for analysis must be considered. Ultimately, these will be used to specifically determine the minimum number of sensors to cover the land while also producing no less than a certain number of readouts.

Procedure

1. Evaluate the size of the area of farmland that will be monitored
   1. Use historical data from records for a general idea to start from
   2. Use drone monitoring and aerial mapping for a more accurate value
   3. If determining the exact area is difficult, the perimeter can also be sufficient
2. Determine the ideal sensor arrangement to maximize the land coverage while minimizing the number of sensors needed
   1. Consider factors such as the farmland topography, soil composition, and potential sources of contamination
   2. Prioritize that all areas are covered
   3. After, try reducing the overlap between each sensor’s detection areas
3. Deploy the sensors in the arrangement from (2) using drones
   1. Although drones can overcome terrain challenges, keep in mind possible weather conditions that may hinder the deployment process, such as wind and precipitation
   2. Consider real-time weather monitoring and contingency plans in case of such conditions
4. Collect a sufficient number of readouts using the sensors
   1. For a pre-scheduled duration (ex. 1 week) each sensor will regularly produce a readout over a set time (ex. every 2 hours)
   2. Over this duration, the sensors must produce a total number of readouts that is greater than the minimum number of readouts needed to accurately evaluate the area
5. Collect the sensors’ data using the drone
   1. Fly out the drone to wirelessly collect data after the pre-scheduled duration
   2. Allows for a quick turnaround between the monitoring and analysis processes
6. Return the drones to their charging station
   1. There, forward the data collected from the drones to IoT gateways and servers
   2. The data will travel through these gateways and servers until it reaches its final storage destination
7. Store the data in IoT data storages
   1. Ensures secure and reliable data storage
8. Able to remotely access the data with an Internet connection
   1. Useful for analysis conducted by environmental scientists and agricultural professionals

Simulation Methods 

In the simulations, two relationships will be considered. First, the effect of each sensor’s detection area on the number of deployed sensors, and second, the effect of each sensor’s detection area on the total number of readouts. Therefore, two sets of simulations will be conducted, with each set running more than 100 data points. Other variables will be held constant, such as a needed amount of 20,000 readouts, a land perimeter of 100 meters, a monitoring time of 1 week (or 168 hours), and a needed time of 2 hours for a sensor to produce one readout. The values of these variables held constant will be determined from the typical parameters encountered in real-life agricultural situations based on various sources. These specific variables will be held constant to highlight the impact of more significant factors, namely the sensor detection area. 

Data Analysis

After the data points have been plotted and connected on a graph, it is first important to ensure that it is accurate and readable. This may include standardizing the format, fixing any missing data, and possibly correcting any outliers. Afterward, the general pattern of the graph can be analyzed through visualizations. Then, more specific statistical measures can be examined, such as certain measures of central tendency (mean, median, and mode). The data’s maximum, minimum, range, standard deviation, quartiles or percentiles, and more can also be determined and provide an easy basis for comparison. Next, it is important to communicate and compare your results with others. Finally, the data can be continuously monitored and updated, leading to revisions in the analysis, in order to improve the realisticness of the results.

A hypothesis is that in both graphs, the sensor detection area (as the independent variable) will cause the graph to eventually level out. At this point, no matter how much you increase the value, the dependent variable will still come out the same. Therefore, this value in both graphs will be the resulting optimal detection area of each sensor that maximizes the number of deployed sensors and the total number of readouts in graphs 1 and 2, respectively. A potential reason for this occurrence is that beyond the minimum number of readouts to analyze the land, the total number of collected readouts (or, in the case of graph 1, the number of deployed sensors which directly relates to the readout amount) is already sufficient. In fact, increasing the number of readouts is unnecessary and will waste resources.

Risk and Safety

No specific safety measures need to be followed for this project.

Project Summary

            To properly evaluate the effect of each sensor’s detection area on both the number of deployed sensors and the total number of readouts, other factors needed to be held constant. For example, in both simulations, the land perimeter was set at 100 m, the minimum number of readouts needed was 20,000 readouts, the total monitoring time was 1 week or 168 hrs, and the time needed for each individual sensor to produce one readout was 2 hours. The values of these constants were determined from extensive research of historical records and based on realistic agricultural situations. Finally, the independent variable of each sensor’s detection area was changed within the range of 0.5 to 4.1 m² to assess its effect on the dependent variables.

In the first set of simulations, the impact of each sensor’s detection area on the number of sensors needed to be deployed was evaluated. In general, their relationship was inversely proportional, with the number of deployed sensors decreasing as each sensor’s detection area increased. This is because as each sensor detects a larger area, fewer sensors are needed to cover the entire area. The relationship appears to be exponential, with the declining slope being significantly steeper in the first half than in the second. However, on closer examination, the values for the number of deployed sensors eventually always equaled 239 instead of gradually growing closer and closer to a limit, or horizontal asymptote. This phenomenon occurred beyond the threshold of 3.4 m², where increasing each sensor’s detection area did not have any impact on the number of needed sensors. Therefore, the graph is a piecewise function, with one part being exponential and the other being linear. This shows how the function is dynamic, and using it to create strategies to tackle farmland contamination may be complex.

In the second set of simulations, the impact of each sensor’s detection area on the total number of readouts was evaluated. The graph’s trend was very similar to the simulation before, with the relationship appearing to be inversely proportional and exponential. The smaller each sensor’s detection area was, the more sensors it needed to cover the entire area, and therefore the more readouts they produced. However, the more readouts did not necessarily mean the better, as it only needed to be more than the minimum number of readouts. Otherwise, it is just using unnecessary resources. Therefore, an optimal sensor detection area is 3.4 m², where 20,076 readouts are generated by 239 sensors. Not only is this more than the required number of readouts, but it is also not significantly over the limit by too much. 

In ongoing research, the other factors’ impacts will also be tested against each other in similar simulations. Furthermore, new and more complex factors will be included to improve the simulation’s realisticness, such as the sensor battery life, drone detection power, drone flying height above ground, and more. Adding these will help to accurately depict the complexity of this issue. Finally, creating a model to determine the ideal sensor arrangement will enhance the current work by maximizing the efficiency of resource usage.