As we are experiencing a shift towards a better understanding of the causes of human accelerated climate change within the general population, there is a need and demand for empirical and real-time data indicative of human induced effects. One of the most defining and well known measures of this human induced change is the purported ‘carbon footprint’, which alludes to what is grouped as human-produced greenhouse gas (GHG) emissions. While the notion of emissions is qualitatively well understood by many, in order to reasonably attain a reduction in GHG emissions quantitative data from continuous real-time monitoring must be widely available. The current state of many types of GHG emission measurements rely on many low-level system estimations with varying degrees of uncertainty, notably for larger areas, for which the method of highest resolution is currently orbital atmospheric measurements. Hence, we propose that given that urban areas (and the associated industry) are the vast majority of GHG emission contributors, that a relatively low-cost real-time atmospheric spectrometry network across such areas would yield invaluable GHG monitoring.
While atmospheric spectroscopy is a well-established measurement technique, a sensor network spread throughout urban areas with the focus on collecting atmospheric column data of emissions is a comparatively novel approach. The power of such a network of sensors can be maximized by taking advantage of mass data manipulation in the form of real-time date output via Cloud computing, an evident advantage of such a network having IoT capabilities.
Monitoring/ Data Collection
Each individual sensor within the network is intended to carry a given spectroscopical setup. Within this context, it is possible to consider two separate instrument setups, namely a passive or active radiation source while adhering to the principals of atmospheric absorption spectroscopy. A ‘passive’ source takes advantage of solar light to take measurements, as seen in a spectrometer such as the Bruker EM27/SUN. Meanwhile, one can also opt to designate a specific radiation source via ‘active’ measurement, namely via infrared solar absorption spectroscopy, which would better suit higher power commercial applications, as opposed to more straightforward operation of passive systems for urban areas. Using the correct method for the given application allows for precise long-term data collection, all while avoiding excessive cost and over-generalized atmospheric measurement. Since we can select a specific type of GHG emission to focus on (i.e. CO2, CH4, CO, etc), we can better optimize the instrumentation loadout.
For long-term, autonomous monitoring, each node will be equipped with temperature and barometric pressure sensors, as these affect spectrometry performance. Units are equipped with a GPS module and antenna to validate accurate position data from geolocation based on mesh-network triangulation, critical given the differential column measurement method.
Depending on the intended measurement area of interest, data transmission follows two possible paths, with varying data security protocol options. For an exclusively urban network, devices will be powered and connected to the network of sensors as traditional IoT devices. For measurements in larger and more remote areas, an effective network mesh is composed of up to three hardware components:
2) Monitoring Sensor Node
3) Optional Relay
Sensors are controlled by the nodes internal computer/microcontroller. Depending on the area of interest, collected data is sent through either an onboard router to a local network or a connected LoRaWAN module, which uses LoRa to transmit the data 128-bit packets through the Peer-to-Peer network to reach a LoRaWAN Gateway. Such Gateways can simply be a LoRaWAN receiver connected to a Microprocessor (e.g. Rasberry Pi) connected to the internet, or some other form of proprietary Gateway. The data packets are then processed and uploaded to the database accordingly.
The fundamentals of Data Processing include extraction, preprocessing, filtering, computation, and visualization for the end-user. The raw data is downloaded from the Cloud and is separated by each sensor nodes Unique Identifier (UID). This is critical for ‘passive’ GHG column measurements which relies on the differences measured at different nodes at the same time. It is hence evident that a larger sensor node leads to a higher measurement resolution. Critically, preprocessing removes unwanted noise attributed to sensors or defective data packets. This stage also plays a role in eliminating backscattering or artifacts of passive or artificial optical sensor measurements. Filtering allows for data sorting depending on user specified requirements, which is then filter into a computation model. Computational modules are composed of sub-modules, intended to separate the nodes data into Classification modules of each individual sensor. Computationally, for larger sensor networks, given the intended long-term monitoring nature of the mesh, it is reasonable to employ a ML framework to optimize differential node calculations and possible evolutionary algorithms to allow for predictive GHG emission models. Finally, visualization takes the computed (WIP) and displays it in an insightful manner to the end user, whether this network is intended as an open-access public resource or a closed commercial monitoring network. While the collected network data may have different privacy concerns, it is good practice to adhere to a minimum of WPA3 or CoAP encryption protocols, while for one may want to adhere to AES encryption for higher value data.