Using MAX to identify and prevent sample error
In semi-intensive shrimp farming, the shrimp are distributed unevenly throughout the pond. Additionally, as they are living animals they move around the pond. The location to sample depends on a variety of factors - where the feed is being distributed and also the depth of the pond. This makes it difficult to get a representative sample of the pond. Before we get into the details of how xpertSea helps farmers improve sampling, let’s take a look at what sampling error is and how it can occur.
What is sampling error?
Let’s take a look at two examples of how sample error can occur. If we have a bag with 16 marbles in it, 8 of which are grey, and 8 of which are blue, sampling error can happen extremely easily.
Taking a handful out, we get 3 grey and 1 blue, leading us to infer that the bag has 75% grey marbles and 25% blue marbles.
Someone else could take a different handful of marbles out, which has 2 grey marbles, and 3 blue marbles, leading that person to infer there are 40% grey marbles and 60% blue marbles.
As can be seen, when a sample is taken that is not representative it leads to incorrect results.
Sampling error in shrimp farming
The same issue occurs in shrimp ponds, which can have a significant impact on biomass estimations. Let’s take an example more specific to shrimp farming and see how it can lead to issues.
When calculating the biomass to plan feed, the average weight of animals in the pond is very important. In this example, the production manager would be alarmed that the pond’s growth has slowed significantly and spend time investigating whether there is an issue in the pond, when in reality there is no problem at all. Being off by 0.2g may not seem significant, but when growth is being used to identify issues it can lead to a lot of wasted time and effort that would be better spent on more important things.
How can xpertSea help?
xpertSea has developed a model using artificial intelligence, which predicts the average weight of shrimp in the pond on a daily basis. With every monitoring, the model learns how the pond is performing and gives an accurate representation of the shrimp population both on the current date, but also every day until harvest.
While xpertSea can help identify sample errors, the reason for these sample errors is often due to biological factors. The sampling process may yield different results even if samples are taken in the same locations in the pond due to biological and external factors. The amount of shrimp that comes out in a sample is dependent on things such as the lunar phase, the water quality, and the activity of shrimp.
Below is an example of a farm that has a strong sampling protocol, where the line represents the growth of the pond, and the dots show each individual monitoring. All of the dots are quite close to the line, and by taking two monitorings per week it can be seen that this farm does not experience significant issues from sampling error. However, there are still cases where the farm does have a monitoring that is not where they would expect, as is the case on May 8th. The model sample was 0.6g greater than the expected weight, and this can be an indicator to the farm that this sample was likely not a good representation of the pond.
However, there are cases where the last monitoring is above where it was expected to be, and the most recent monitoring is below. This is an example where it is likely two monitorings in a row had some sampling error, and the model is a better representation of the pond. The recommended action here is to take an additional sample with extra attention to the location of the sample to identify if there was in fact an issue with the pond that caused the growth to stop, or if it was only a sampling error.
For more information or to bring xpertSea MAX to your farm, contact us at info@xpertsea.com
