Spatial Coverage of Monitoring Networks: A Climate Observing System Simulation Experiment

Abstract

Observing systems consisting of a finite number of in situ monitoring stations can provide high-quality measurements with the ability to quality assure both the instruments and the data but offer limited information over larger geographic areas. This paper quantifies the spatial coverage represented by a finite set of monitoring stations by using global data-data that are possibly of lower resolution and quality. For illustration purposes, merged satellite temperature data from Microwave Sounding Units are used to estimate the representativeness of the Global Climate Observing System Reference Upper-Air Network (GRUAN). While many metrics exist for evaluating the representativeness of a site, the ability to have highly accurate monthly averaged data is essential for both trend detection and climatology evaluation. The calculated correlations of the monthly averaged upper-troposphere satellite-derived temperatures over the GRUAN stations with all other pixels around the globe show that the current 9 certified GRUAN stations have moderate correlations (r $ 0.7) for approximately 10% of the earth, but an expanded network incorporating another 15 stations would result in moderate correlations for just over 60% of the earth. This analysis indicates that the value of additional stations can be quantified by using historical, satellite, or model data and can be used to reveal critical gaps in current monitoring capabilities. Evaluating the value of potential additional stations and prioritizing their initiation can optimize networks. The expansion of networks can be evaluated in a manner that allows for optimal benefit on the basis of optimization theory and economic analyses.

Multi-decadal surface ozone trends at globally distributed remote locations

“…Full understanding of global surface ozone trends based on observations would require a major expansion of the surface ozone observation network. Based on newly developed methods [Sofen et al, 2016;Weatherhead et al, 2017;Chang et al, 2019] the locations of these additional sites could be carefully planned to maximize their spatial representativeness.…”

The GEWEX Water Vapor Assessment: Overview and Introduction to Results and Recommendations

“…The GRUAN network expansion carefully considers satellite validation and climate monitoring requirements. However, certain areas of the Earth will remain to be underrepresented [57] and reprocessing activities of historical archives are not covered by GRUAN's mandate. In order to further enhance the value of radiosonde data in the validation of satellite-based CDRs and to increase the number of collocations between satellite and radiosonde, G-VAP recommends to bias-correct and reprocess stable multi-station radiosonde archives of humidity and temperature going back to the 1970s (see also [33]).…”

Statistical regularization for trend detection: An integrated approach for detecting long-term trends from sparse tropospheric ozone profiles

“…We choose the thin-plate smoothing spline as the spatial basis function because it is computationally efficient, and because it avoids the problem of choosing "knot locations" (Wood, 2003). Note that the focus of this study is jointly modeling multiple correlated time series at different heights above the same region, which differs from a trend study of a series of measurements with uneven temporal sampling at a single altitude (Tiao et al, 1990;Rehfeld et al, 2011;Weatherhead et al, 2017), or irregularly distributed observations from a monitoring network (Stein, 1999;Chang et al, 2015Chang et al, , 2017. Our problem can be seen as decomposing a two dimensional signal according to the associated variability on a regular grid cell (i.e., based on the monthly and annual index at a fixed altitude bin).…”

“…Time series are well correlated (∼ 0.6 − 0.9) within a vertical range of 100 hPa, except for the top layer where correlations decrease with distance at twice the rate of the lowest layers. As noted byWeatherhead et al (2017), two time series can be highly correlated even if they have very different mean values or magnitudes of variability. Because our technique takes advantage of the nearby layers with moderate or high correlation, we expect an improved quantification of the absolute variability and resulting trend estimate.The smooth seasonal and interannual components in units of ppbv (parts per billion by volume) derived from Eq (2) are shown inFigure 2(a) and (b), respectively.…”