Last year our team travelled to the Philippines to test the CRW platform in a wide range of realistic flood conditions. During our time there we had to opportunity to deploy on Lake Taal, a volcanic lake home to a large segment of the Philippines aquaculture industry. We learned that the crop of fish had recently been devastated by an unexpected rise in water temperature and subsequent drop in dissolved oxygen content (read more in some of our previous posts). This experience first motivated our investigation into using the CRW platform for distributed monitoring of dissolved oxygen content in bodies of water that are used for aquaculture. Most of the aquaculture industry use aerated ponds where a diffused air system is used to aerate the pond multiple times a day to replenish the oxygen. Farm operators usually take hand measurements at a few edges of the pond and estimate the disolved oxygen in the entire pond. This is highly inaccurate as there is some amount of spatial variation and most of the time operators end up over-aerating the pond. If we can estimate the spatial variation of disolved oxygen in the pond, the operators can more accurately estimate when to aerate and potentially save huge amounts of electricity consumed by the diffused air system.
Early this year our team headed to Shelby Fish Farm in western Ohio to put the boats to the test and see how well the system could map spatial and temporal variations of dissolved oxygen content in the water. Equipped with the Atlas-Scientific dissolved oxygen sensor, the airboats gathered data autonomously using random, lawnmower, and highest uncertainty sampling patterns. Some of the data gathered is presented below.
Analysis on the data revealed that the sensor suffered from the hysteresis effect in which the sensor responds slower than the rate of change of the variable that is being observed in the field. Spatial measurements obtained using most electrode type sensors exhibit this effect. This effect can be compensated by intelligent sampling techniques that predict the rate of change and take measurements in an adaptive way. It was also observed that some of the data collected was corrupted with a lot of random noise. Further investigation revealed that the noise was introduced into the system due to the mechanical shock caused by the sensors bumping against hulls in the boats' wakes. However in the non-corrupted data, we observed spatial variations in dissolved oxygen throughout the ponds. This suggests a potential opportunity for using the CRW platform to monitor large aquaculture areas to efficiently generate up-to-date maps of dissolved oxygen content and identify areas in need of re-oxygenation. By mounting the sensors rigidly below the waterline, we hope to eliminate the noise caused repeated shocks to the sensor and verify our findings. We will test this theory when we head back to Shelby Fish Farm this weekend to gather more data and test the additional sampling algorithms that we have developed to compensate for the hysteresis effect.
Above is a video of 4 boats autonomously collecting D.O data.
Heres a video below on a manual sampling technique by the State Water Resources Control Board.