Range; the distribution is skewed, and there are outliers in the dataset.
The range would be a better measure of variability in this case because the dataset contains outliers, such as the values of 1.0 and 18.0, and the distribution is not symmetrical.