Remake of the low cost carbon dioxide sensor of the carbon dioxide network deployed by INVOLCAN in the urban areas of Puerto Naos and La Bombilla, La Palma, Canary Islands

An anomalous CO₂ degassing appeared by the end of Tajogaite eruption (North-West flank of Cumbre Vieja volcano ridge, La Palma, Canary Islands), in the neighborhoods of La Bombilla and Puerto Naos at about 6 km distance from the volcanic vent. The areas affected by the anomalous CO₂ degassing were not directly affected by lava flows during the eruptive period. After the eruption, and due to this strong volcanic-hydrothermal carbon dioxide emissions (CO₂>5-20%)  were included in the exclusion zone. CO₂ is an invisible toxic gas, as well as asphyxiating, and may be lethal when is present in concentrations higher than 14%. During the post-eruptive period, INVOLCAN deployed its own indoor and outdoor CO₂ monitoring networks in collaboration with other institutions, with the aim of delimitating the anomalous CO₂ degassing areas, paying attention to those areas where CO₂ air concentration exceeds hazardous thresholds. The number of monitoring stations were increasing to cover most of the homes, garages, basements, and local businesses. The first monitoring network were based on a LILYGO® TTGO T-SIM7000G electroniccard, previously programmed with an unstable algorithm that caused problems during the measurements. After some implementations to enhance the stability of the sensor, a new algorithm was developed that consists of the acquisition of ambient values every 5 seconds, applying a Moving Average Filter in every measurement to avoid outliers. The SIM card integrated in the hardware allows the data transmission to an MQTT broker where the values are published every 5 minutes, recollecting them in a unique Raspberry Pi 4 Model B located at the INVOLCAN headquarters, that reads and stores the data in two databases (InfluxDB and Google Sheets). The visualization of the values are done through Grafana Cloud, recollecting the data from InfluxDB and showing them distributed as tables and a geographic map that illustrates the concentration in the measurement points. The difference between this and the last storing is the flexibility when visualizing the data, that can be transformed to different kind of plots as mentioned. Moreover, an API for the management of each subsystem is created using PyQT, allowing to the user the calibration of the sensors in remote, as well as executing a soft reboot, or the integration of deeper parameters like the sensor mode (manual polling, streaming or command mode) or pressure data. Two of the 20 devices have been successfully installed and they are working correctly in La Palma, meanwhile an amount of 18 devices are being tested and recollecting properly with better stability in CO₂ concentration measurements at our laboratory and will be installed indoor in different locations soon. The remaking of the algorithm allows to forget previous problems of wrong data and disconnections, obtaining accurate data compared to commercial sensors and helping the operator to configure and control the sensors without moving to conflicting locations.