Introduction: The wildfires in California and throughout the western USA have been devastating, especially over the past 3 years, causing loss of lives, incredible property losses, and the bankruptcy of a giant utility [Pacific Gas and Electric] with over $50B of debt. And it is not just the USA that suffers from such fiery disasters but there are devastating fires throughout Australia, Siberia, and around the world. The premise is that early detection could provide avenues for emergency response and the limitation of lost lives and property damage. Current methods to detect these fires ranges from deployment of local forest rangers, satellite optical and thermal IR imaging, some local monitoring, and public education. But none of these are adequate as satellites may not have resolution required for small startup fires and can be obscured by weather. Since fires produce emissions that can provide an atmospheric signature including gases and particles as well as thermal plumes, the hypothesis is that detection of these “fire” signatures with sensitivity could provide early warning if a low-cost distributed network of sensors is deployed.The case can be made for sensing atmospheric emissions in that Carbon Monoxide is produced in high concentrations especially in smoldering fires and has been used for early warning of fires in coal mines for half a century or more. Figure 1 illustrates the use of CO to map the impact of 2018 fires in California through IR imaging through 18000 feet of the atmosphere1 at 200 ppb and lower concentrations. Clearly CO is related to the fire activity but how early can a fire be sensed? What confounding variables will be present such as T, P, RH, and wind velocity or local activities and the natural variability of the chemical composition of the air at low levels? Are CO or particles the best emission parameters to monitor or are better variables available? The answers to these questions are key to the design and deployment of an early warning system based on measurement of atmospheric conditions and atmospheric fire signature emissions! Methods The approach has two objectives in this early research: 1] to understand the emission signatures from fires as recorded by low cost environmental sensors, and 2] to evaluate the feasibility of a practical early warning fire detector based on emission signatures in addition to changes in T, RH, and wind direction. Clearly, we know CO2 (~1500g/kg forest fuel) and CO (~100g/kg) are the largest gaseous emissions by >2 orders of magnitude, but other variables include particles (~20-50g/kg), as well as gases hazardous at very low concentrations like O3, NO2, SO2 and other gases such as methane, CH2O and other NMVOCs2 that could come from different sources.A monitoring system was designed made from low cost COTS [commercially off the shelf] sensors and electronics that could be used to gather field data. The system is complete with sensor hardware, communication alternatives, and computation capability. SPEC sensors3 and KWJ4 electronic hardware and algorithms were selected to partner with commercial CO2, particulate and environmental [T, P, RH] sensors5 in order to build a system that could address Objective 1. Evaluation of actual field data will then be used to achieve Objective 2 as it is clear that a deployed sensor system must have low cost, near zero power, and long-term stability as well as be free from interfering signals. Results The construction of the prototype integrated monitoring and analysis system node [at left] was completed and included SPEC Sensors for CO, NO2, O3; Alphasense SO2 sensor, Sensirion SPD30 (NDIR) CO2 and SPS30 particulate sensor as well as Wind Speed and Direction, a 4.4 A∙h LiPo Battery, two 3.4 W adjustable solar panels, and USB Connection for power/communications.Data from a deployed SPEC printed CO sensor - taken by a third party (SCAQMD) - is illustrated at right. The SPEC CO sensor was only calibrated once at the start of the deployment and the data from it are plotted vs a co-located EPA reference instrument. The agreement over months supports the hypothesis that an ultra-low-cost screen-printed electrochemical sensor can possess the stability that will be needed in a practical application scenario. There is significant agreement over 0-2000ppb between a frequently calibrated instrument with 100 ppb resolution reading 1/hr and the hourly average of a tiny low-cost real-time sensor! Concluding Remarks: The systems and sensor and the application data will be discussed further in the presentation. The data supports the possibilities for complete systems and remote nodes that can both detect wildfires and then better understand wildfire emissions. Early fire warnings though emission monitoring will become one of the methods contributing to the protection of lives, property and health. Study of the application and improving sensor systems is ongoing.
Read full abstract