(Series: Drones, Data and Errors in Agricultural Environments)
Drones are often associated with urban environments, which are full of invisible sources of interference such as dense concrete structures, glass façades, large reflective surfaces, countless radio signals, electromagnetic interference, GNSS noise, mobile networks, and multipath effects caused by reflected GNSS signals. All of these factors can confuse an unmanned aircraft operating in automated mode and may even affect manual control. In contrast, agricultural fields are commonly perceived as “interference-free” spaces: wide, open green areas where a drone is assumed to encounter no obstacles or disturbances, needing only to follow a predefined route, capture imagery, and return to the designated point after collecting the required data. This assumption, however, is misleading.
From the drone’s perspective, agricultural terrain is far from a sterile environment. Vegetation itself—whether wheat, maize, or rapeseed—forms a complex and highly uniform surface, both optically and in terms of onboard sensing. Large, homogeneous, low-contrast green textures provide very limited visual reference points for navigation, particularly when GNSS signal quality degrades or when onboard software attempts to stabilise flight primarily using camera-based inputs. Moreover, open environments are not necessarily free from disturbances. GNSS signal quality may be influenced by local terrain features, soil moisture, vegetation density, ambient temperature, nearby electrical infrastructure, or even space weather. These influences are largely invisible to operators, and their magnitude cannot be reliably predicted in advance.
Drift—that is, a slow and often barely perceptible unintentional lateral displacement—is a particularly insidious phenomenon, as it is frequently detected only after a delay. The drone continues flying and performing its monitoring task while the display shows the map, grid, and position indicators, giving the impression that everything is functioning normally. During monitoring operations, the remote pilot’s attention is typically focused on task execution rather than on subtle lateral deviations or small, abnormal movements away from the planned route. At this stage, the anomaly does not yet appear significant enough to justify aborting the operation or activating emergency procedures. Initially, only minor deviations from the planned trajectory are observable; however, under certain conditions, these deviations may gradually accumulate, eventually escalating into an incident, including a crash and damage to the drone.
This series examines two real agricultural monitoring operations conducted in the same area, six days apart. Both missions began with a successful take-off and the initiation of automated flight, and in both cases anomalies were observed that were not fully aligned with the predefined flight plan. Despite these similarities, the final outcomes differed markedly. During the first operation, drift was followed by a communication failure, a complete loss of control, and impact into a wheat field. In the second flight, drift escalated and manifested along multiple axes—lateral movement, yaw, and altitude—followed by a slow descent and ground contact. Although the drone sustained damage in this case as well, it remained partially operational and recoverable. Thus, two distinct outcomes emerged under seemingly identical terrain and operational conditions.
The contrast between these two cases is particularly noteworthy because agricultural environments are rarely regarded as sources of operational risk. Yet precision agriculture today does more than produce crops—it actively challenges and exposes the limits of technology. Drones, onboard computing systems, GNSS signals, cameras, generated data, and automated decision-making processes all operate simultaneously, and disturbances do not always arise where we expect them. In the following parts of this series, we will explore what happens when a drone does not behave as its flight plan predicts, and what lessons can be drawn from a terrain long considered the opposite of the urban environment—though in reality, it is anything but.