Exponential leap-forward gradient scheme for determining the isothermal layer depth from profile data

Two distinct layers usually exist in the upper ocean. The first has a near-zero vertical gradient in temperature (or density) from the surface and is called the isothermal layer (or mixed layer). Beneath that is a layer with a strong vertical gradient in temperature (or density), called the thermocline (or pycnocline). The isothermal layer depth (ILD) or mixed layer depth (MLD) for the same profile varies depending on the method used to determine it. Also, whether they are subjective or objective, existing methods of determining the ILD do not estimate the thermocline (pycnocline) gradient. Here, we propose a new exponential leap-forward gradient (ELG) method of determining the ILD that retains the strengths of subjective (simplicity) and objective (gradient change) methods and avoids their weaknesses (subjective methods are threshold-sensitive and objective methods are computationally intensive). This new method involves two steps: (1) the estimation of the thermocline gradient G _th for an individual temperature profile, and (2) the computation of the vertical gradient by averaging over gradients using exponential leap-forward steps. Such averaging can filter out noise in the profile data. Five existing methods of determining the ILD (difference, gradient, maximum curvature, maximum angle, and optimal linear fitting methods) as well as the proposed ELG method were verified using global expendable bathythermograph (XBT) temperature and conductivity–temperature–depth (CTD) datasets. Among all the methods considered, the ELG method yielded the highest skill score and the lowest Shannon information entropy (i.e., the lowest uncertainty).