Low-cost drone system for monitoring ammonia and GHG emissions in dairy barns: a case study in southern Italy

Bottom-up approaches for estimating greenhouse gas and ammonia (NH3) emissions from livestock farms require gas concentration measurements at the farm level. A major limit to their applicability across diverse farming and environmental contexts is the reliance on high-cost standard instruments, with several operational limitations [1]. This study aimed to validate a prototype low-cost sensing system for ground and aerial measurements, designed to monitor concentrations and estimate NH3 and CO2 emissions of a dairy barn in the Mediterranean area. The experiment was conducted over one week during the hot season at a facility housing 50 lactating cows (Fig. 1). The system, previously described in [2], consisted of two prototype devices equipped with low-cost sensors to monitor gas concentrations and environmental parameters. One device was installed at a fixed ground station downwind of the main airflow (Fig.1, point 4), while the second was mounted on a mobile drone (point 5). Wind speed and direction were measured using a 2D anemometer placed inside the barn (point 6). Additionally, measurements from the low-cost devices (point 3) were acquired simultaneously with those from an infrared photoacoustic gas monitor (point 2), used as a reference instrument, to derive correction equations for both NH3 and CO2 concentrations. Finally, the corrected measurements from the low-cost devices were used to estimate NH3 and CO2 emissions during and after specific barn operation (i.e., slurry removal and feed delivery) using the WindTrax dispersion model. Estimated emissions were compared using the Mann-Whitney U test. NH3 emissions during slurry removal (0.00037 g m-2 s-1) were significantly higher (p < 0.001) compared to those at the end of cleaning procedures (0.00020 g m-2 s-1), while the opposite trend (p < 0.001) was observed when comparing NH3 emissions during (0.00025 g m-2 s-1) and after (0.00034 g m-2 s-1) feed distribution. Regarding CO2 emissions, higher values (p < 0.001) were estimated during cleaning operations (0.1386 g m-2 s-1) than after slurry removal (0.0487 g m-2 s-1), while no significant differences were found (p = 0.379) between emissions during (0.0340 g m-2 s-1) and after (0.0392 g m-2 s-1) feed distribution. Overall, the system showed positive performance in measuring gas concentrations after correction against a reference instrument and in capturing emissions fluctuations across diverse phases of barn management. Developing automated, low-cost systems thus represents a promising solution for both tailoring emission factors to local conditions and providing immediate feedback to farmers, enabling them to adopt timely mitigation measures.