Random error analysis of marine xCO2 measurements in a coastal upwelling region

Abstract

Quantifying and identifying measurement error is an ongoing challenge for carbon cycle science to constrain measurable uncertainty related to the sources and sinks of CO2. One source of uncertainty in measurements is derived from random errors (ε); thus, it is important to quantify their magnitude and their relationship to environmental variability in order to constrain local-to-global carbon budgets. We applied a paired-observation method to determine ε associated with marine xCO2 in a coastal upwelling zone of an eastern boundary current. Continuous data (3-h resolution) from a mooring platform during upwelling and non-upwelling seasons was analyzed off of northern Baja California in the California Current. To test the rigor of the algorithm to calculate ε we propose a method for determining daily mean time series values that may be affected by ε. To do this we used either two or three variables in the function, but no significant differences for ε mean values were found due to the large variability in ε (−0.088±27ppm for two variables and −0.057±28ppm for three variables). Mean ε values were centered on zero, with low values of ε more frequent than greater values, and follow a double exponential distribution. Random error variability increased with higher magnitudes of xCO2, and in general, ε variability increased in relation to upwelling conditions (up to ∼9% of measurements). Increased ε during upwelling suggests the importance of meso-scale processes on ε variability and could have a large influence seasonal to annual CO2 estimates. This approach could be extended and modified to other marine carbonate system variables as part of data quality assurance/quality control and to quantify uncertainty (due to ε) from a wide variety of continuous oceanographic monitoring platforms.