The climate is changing fast, especially in the Arctic Ocean where sea-ice extent continues to decease. Understanding the effects of the changing climate on the fragile Arctic ecosystem is of utmost importance, as changes in sea-ice cover will have direct consequences on the ecosystem.
The team that left Longyearbyen was a diverse group, consisting of scientists from oceanography, marine technology, meteorology, as well as an outreach group.
To understand ocean processes, it is necessary to collect measurements of the water column. The standard method for getting the measurements is to use moorings and point samples of conductivity and temperature at different depths (CTD). The point samples provide a sparse set of spatially distributed measurements, while the moorings provide excellent temporal coverage and resolution, but are very limited in spatial distribution.
Gliders are a type of unmanned underwater vehicles (UUVs) that utilize changes in buoyancy to produce forward thrust, and these vehicles have been used extensively to understand large-scale ocean processes. During the Legacy cruise, two such gliders were deployed. One would collect measurements for several months; the other glider was recovered after 9 days out in the ocean.
Glider measurements have a limited resolution since gliders are often moving slower than the water masses they are sampling. Autonomous underwater vehicles (AUVs) are a class of UUVs that are propeller driven, which have the potential to deliver high-resolution, spatially distributed samples of the water masses. However, the ocean is big, and care must be taken to utilize the resource in an optimal manner.
The goal of the group from the Applied Underwater Robotics laboratory (AUR-lab) at NTNU (Norwegian University of Science and Technology) was to test and collect data using a method for adaptively sampling fronts. Fronts are characterized by large temperature and salinity changes over relatively short distances, which can indicate layers of different water masses. Tracking and sampling of fronts are important to increase the understanding of different oceanographic processes, but time and resources are limited, and the use of the available platform must therefore be optimized.
This is where adaptive path planning comes into play – instead of providing a complete coverage of an area, the vehicle will be attracted towards areas of interest, and increase the time spent sampling the important processes. Usually, the path of the robot is planned prior to the launch of the vehicle, and the vehicle will follow this path until the mission is complete.
This is inefficient, since the robot acquires new knowledge as the mission progress. By utilizing this new information to alter the path of the robot, a more efficient path, with respect to the scientific objective, can be achieved. This is called adaptive path planning – the robot adapts its plan based on new information. This allows us to define scientific objectives, attracting the robot towards areas of interest, rather than defining large areas that the robot should cover, which in turn allows increased utilization of time and sparse resources.
The developed method was applied to track an Arctic front at 82º north, where melt water from the ice and warmer Atlantic water mixed. The front that was tracked was characterized by a large temperature gradient in the upper part of the water column, with temperatures down to zero degrees.
Normally, an AUV mission consist of preprogrammed waypoints, and the vehicle follows a deterministic path. In adaptive path planning, the mission is generated online based on what the vehicle is sensing. This complicates the operation, since it induces uncertainty into the behavior of the vehicle – it is not easy to identify if the vehicle is behaving as it should, or if something has gone wrong.
An operation at Arctic latitudes is further complicated by harsh environmental conditions, severely limited communication, as well as the possibility of drifting sea ice entering the operational area. Nonetheless, the AUV was successful in tracking the front and an illustration of the front is shown in Figure 3.
Further development of methods like this will save time, money, and may help researchers answer important questions, which has previously suffered from undersampling.