Within Chapter2we have defined our variables; relative, absolute and steric sea level, and shown how they relate to one another physically through the derived term augmented dynamic topography (τ). Chapters 3, 5 and 4 go on to evaluate their respective trends in sea level. These trends are then combined to estimate ocean mass trends in Chapter 6.
Vertical land displacement corrections and have been defined. Set out are 3 different approaches to correct forV LD; using altimetry, a mass redistribu- tion model (ICE-5G), and the Global Positioning System. These corrections are applied in sections 3.4.4 to3.4.5 of Chapter 3.
The effects of changes in the Earth’s geopotential arising from the redistri- bution of mass are accounted for by including changes to the gravitational surface G. We apply these changes in Chapters 3, 4 and 5.
Barystatic sea level is defined in terms of both RSL and ASL and SSL
using equations (2.21, 2.23 and 2.24). Equations (2.23 and 2.24) are used in Chapter 6 to compare our mass trend anaylsis with the gravitational fingerprints (Tamisiea et al.,2001).
Tide Gauge Trends in Sea Level
and the Effects of Different
Subsets
In Chapter 3 tide gauges are used to calculate relative sea level trends. Relative sea level trends are then converted into augmented dynamic topography in order to compare with both steric and altimetry observations in Chapters 4and 5. Our analysis focuses on trends from 1993 to 2010 in order to compare our results with those from altimetry, although some longer trends within this Chapter are shown for comparison due to tide gauges having longer time series available.
Tide gauge data is of high quality, but there are gaps in both the temporal and spatial coverage. Due to this limited sampling, there are different criteria for selecting subsets of gauges, such as selecting long records or having short records with more uniform global coverage. The analysis in this Chapter shows results from 4 different sets of tide gauges: all tide gauges, long time series gauges, a set offering uniform spatial coverage and 500 gauges chosen using an algorithm to balance good spatial coverage with length of time series.
We compare 3 different corrections for Vertical Land Displacement,V LD, as there is not one global solution available. We use model based estimates for Glacial Isostatic Adjustment GIA, the difference of tide gauge and altimetry trends and Global Positioning System measurements.
Our analysis provides reasoning behind the tide gauge choices which are subse- quently taken forward into Chapter 6when estimating the mass component of sea level.
3.1
Introduction
Merrifield et al.(2009) show that the trend in sea level is regionally variable, rather
than the same as the global trend. They plot contributions from the north (25oN
- 65oN), tropics (25oS - 25oN) and south (25oS - 65oS) and show that variability
within the tropics and south are key to understanding the increase in global- average trend over the past 20 years, where sea level rates rose from around 1.7 mm yr−1in the 20th century to around 3.2 mm yr−1 between 1993-2010. Merrifield
et al. (2009) show that the variability of the trend in the tropics and south has
been out of phase for the earlier part of the time series (1955 - 1990), but became in phase after 1990, creating an increase in trend which is seen in the global mean. We explore this interpretation in more depth by further breaking down sea level trends into the steric and mass components and use altimetry (in Chapter 5) to independently verify our results.
Global mean sea level was estimated to have risen at an average rate of 1.7 mm yr−1between 1901 and 1990 (Stocker et al.,2013). The contributions to this global
mean was estimated to be 0.37 mm yr−1 from thermal expansion due to changes in
density, 0.7 mm yr−1 from glaciers and -0.11 mm yr−1 from changes in land water,
leaving a 0.5 mm yr−1 residual (potentially accounted for through changes in the
ice sheets). One problem with estimating the historical change in global mean sea level using tide gauges is that due to their scattered locations, the redistribution of water may be undetected or misdiagnosed (Pugh and Woodworth, 2014). The global mean sea level variations over 1993-2011 computed using an ensemble mean of five different analyses of altimeter data is reported as 3.2±0.6 mm yr−1 (Ablain
et al.,2015,Church et al.,2011,Leuliette and Scharroo,2010,Stocker et al.,2013)
meaning that the trends for the later years in the times series are twice the size of the earlier years. Our measurements of the contributions to this trend are therefore important for out understanding of the cause of this increase.
Tide gauges were historically set up for local needs and long time series gauges are clustered over the Northern Hemisphere. Global trends are the weighted average
of regional trends although regional variability tends to be much larger than the global trend. Tai (2011) has shown how the global mean sea level can be inferred from 237 tide gauges, but that on short time scales the trends are dominated by natural variability which gives the trend calculation a large error. A global time series of mean sea level in the IPCC AR4 (Solomon et al., 2007) compares a time series from 1950 of coastal tide gauges (Holgate and Woodworth, 2004) with the time series of mean sea level measured by both altimetry (Leuliette et al., 2004) and a reconstructed time series of sea level since 1870, (Church and White, 2006) and shows that they are in good agreement at 90% confidence intervals.
We show variability in 4 variations of the global trend using our 4 different sets of tide gauges and obtain a statistical error (σ) in section 3.6. In Chapter 6 we redefine these errors to better quantify their size. We also explore the effect that the spatial bias of tide gauge distribution has on the time series of global trends by comparing global trends calculated with all gauges (figure (3.18a)) to the time series of all northern hemisphere sea level trends (figure (3.19)).
3.1.1
Regional trends
Causes of regional trends are wide ranging. The high trends in sea level around the western tropical Pacific are now well documented using altimetry data (Cazenave
and Llovel, 2010). Merrifield et al. (2012) have shown them to be strongly linked
to trade wind variability and well correlated with the Pacific Decadal Oscillation (PDO). Regional trends from local variability can influence the global mean on annual to multi decadal time-scales to the magnitude of ∼10 mm yr−1. Merrifield
et al.(2012) do not expect this pattern to continue, but to reverse in sign over the
east and west Pacific as the Pacific Decadal Oscillation reverses. A similar region of high positive trends in sea level change is found in tide gauges on the Atlantic coast of North America. Sallenger et al. (2012), using tide gauges, show that there is evidence for a large recent decadal acceleration here. This acceleration might be accounted for by changes in the North Atlantic Oscillation, weakening of the gyre system and hence sea level gradients. These events also coincide with melting of the Greenland Ice Sheet (GIS), causing warming and/or freshening of the surface waters of the sub-polar North Atlantic and weakening the pressure gradients of the Atlantic Meridional Overturning Circulation. Calafat et al.(2012) have found that tide gauges exhibit decadal variability of up to 15cm in the North
East Atlantic and Mediterranean and are highly correlated with long-shore winds as well as with the North Atlantic Oscillation and with each other. Significant changes are reported in the Southern Ocean, such as the potential contributor to sea level through the warming of abyssal waters (Purkey and Johnson, 2010) and the upper 1000 m (Gille, 2002, Stocker et al., 2013).