Evaluating seasonal loading models and their impact on global and regional reference frame alignment

Abstract

Seasonal variations are observed in GPS time series, but are not included in the International Terrestrial Reference Frame (ITRF) models. Unmodeled seasonal variations at sites used for reference frame alignment are aliased into the reference frame parameters and bias all coordinates in the transformed solution. We augment ITRF2008 with seasonal loading models based either on Gravity Recovery and Climate Experiment (GRACE) measurements or a suite of models for atmospheric pressure, continental hydrology, and nontidal ocean loading. We model the seasonal components using either annual and semiannual terms or a nonparametric approach. When we include a seasonal variation model, the weighted root-mean-square misfit after seven-parameter transformation decreases for 70-90% of the daily GPS solutions depending on the network and seasonal model used, relative to a baseline case using ITRF2008. When seasonal variations are included in the reference frame solution, the observed seasonal variations are more consistent with the GRACE-based model at 80-85% of the GPS sites that were not used in the frame alignment. The suite of forward models performs nearly as well as the GRACE-based model for North America, but substantially worse for other parts of the world. We interpret these findings to mean that the use of ITRF2008 without seasonal terms causes the amplitude of seasonal variations in the coordinate time series to be damped down relative to the true loading deformation and that the observed GPS time series are more consistent with a TRF model that includes seasonal variations. At present, a seasonal model derived from GRACE captures seasonal variations more faithfully than one based on hydrologic models.

Quantifying deformation in North Borneo with GPS

“…Always some aliasing of (small) position offsets in the Helmert transformations will take place during the (numerical) matching of the predicted (IGS08) and computed GPS position of each included IGS08 reference station. This is somewhat similar to the more extreme case of exclusively using a regional IGS sub-network where seasonal signals are reduced as these are aliased in the reference frame transformation parameters as has been observed by several other studies (Freymueller 2009;Legrand et al 2012;Zou et al 2014). In our case this aliasing seems to be more of a (seasonal) position misfit balancing nature that depends on the size and geometry of the chosen global IGS08 sub-network.…”

“…The shorter 2009-2011 data period is more susceptible to this especially as it continues beyond the latest data inputs into ITRF08/IGS08. However, mainly a larger spread/variation in the velocity differences in each direction is observed here (Table K) The seasonal variations remain similar using either method for both data periods ( Legrand et al 2012;Zou et al 2014). …”

Comparison of observed and modeled seasonal crustal vertical displacements derived from multi‐institution GPS and GRACE solutions

“…However, it has been reported that annual signals in GPS observations only decrease slightly when applying GRACE‐modeled CVD corrections to GPS time series at stations spread worldwide [ Tregoning et al , ; Horwath et al , ; Tesmer et al , ; Zou et al , ]. Errors or limitations in both the GPS and GRACE systems may result in low root‐mean‐square (RMS) reduction.…”

A Full‐Spectrum Bedrock Thermal Expansion Model and Its Impact on the Global Positioning System Height Time Series

“…Nontidal ocean mass loads are derived from the 3-hourly ocean bottom pressure on a regular 1.0°× 1.0°global grid from the MPIOM ocean model (van Dam et al, 2012). Notably, the environment loads vary among different data centers (Jiang et al, 2013;Zou et al, 2014). Since our main purpose is to study the impact of bedrock thermal expansion on the GPS height time series, only the Deutsche GeoForschungsZentrum (GFZ) products are selected here for reference (Dill & Dobslaw, 2013).…”