Assessing the Skill of a High-Resolution Marine Biophysical Model Using Geostatistical Analysis of Mesoscale Ocean Chlorophyll Variability From Field Observations and Remote Sensing

Rachel Eveleth,Alison P Chase,Ivan D Lima,Matthew C Long,David M Glover,Scott C Doney

doi:10.3389/fmars.2021.612764

Abstract

High-resolution ocean biophysical models are now routinely being conducted at basin and global-scale, opening opportunities to deepen our understanding of the mechanistic coupling of physical and biological processes at the mesoscale. Prior to using these models to test scientific questions, we need to assess their skill. While progress has been made in validating the mean field, little work has been done to evaluate skill of the simulated mesoscale variability. Here we use geostatistical 2-D variograms to quantify the magnitude and spatial scale of chlorophyll a patchiness in a 1/10th-degree eddy-resolving coupled Community Earth System Model simulation. We compare results from satellite remote sensing and ship underway observations in the North Atlantic Ocean, where there is a large seasonal phytoplankton bloom. The coefficients of variation, i.e., the arithmetic standard deviation divided by the mean, from the two observational data sets are approximately invariant across a large range of mean chlorophyllavalues from oligotrophic and winter to subpolar bloom conditions. This relationship between the chlorophyllamesoscale variability and the mean field appears to reflect an emergent property of marine biophysics, and the high-resolution simulation does poorly in capturing this skill metric, with the model underestimating observed variability under low chlorophyllaconditions such as in the subtropics.

Highlights

Mesoscale variability of phytoplankton biomass and chlorophyll a, including “patchiness,” has long been observed in the global ocean, where mesoscale is defined here as O(10–100 km) and weeksmonths (e.g., Mackas et al, 1985; McGillicuddy, 2016)
We focus here on analyzing the skill of a high-resolution, eddy-resolving biophysical ocean simulation in capturing magnitude and length scales of observed phytoplankton variability using structure function analysis (Journel and Huijbregts, 1978; Clark, 1987)
Improved measures of mesoscale biological variability in the surface ocean are needed both to address scientific questions and test the skill of eddy-resolving marine biophysical simulations. Geostatistical approaches, such as the variogram techniques used here, have some distinct advantages in that they can be applied to both 1-D and 2-D fields with data gaps, partition resolved spatial variability from unresolved instrument and algorithm noise and sub sample-resolution scale biophysical processes, and provide a measure of the spatial correlation scale or range (Doney et al, 2003; Glover et al, 2018)

Summary

Introduction

Mesoscale variability of phytoplankton biomass and chlorophyll a, including “patchiness,” has long been observed in the global ocean, where mesoscale is defined here as O(10–100 km) and weeksmonths (e.g., Mackas et al, 1985; McGillicuddy, 2016). At global or even basin scales, resolving mesoscale ocean processes has been limited by computing power, and the physical ramifications of mesoscale eddies often have been accounted for by parameterizations of isopycnal mixing and eddy-induced transport in lowresolution simulations (Gent and Mcwilliams, 1990). These low-resolution ocean models have been used for most multi-decadal hindcasts of marine ecosystem and ocean biogeochemical dynamics, as well as almost all century-scale coupled Earth System Model projections of future climate change (e.g., Bonan and Doney, 2018). With the emergence of routine, high-resolution, global-scale simulations comes new opportunities to test the impact of parameterizations in more computationally accessible low-resolution simulations and to investigate specific science questions on mesoscale biophysical dynamics

Methods

Results

Conclusion