Investigating the Uncertainty in Global SST Trends Due to Internal Variations Using an Improved Trend Estimator

Abstract

A new criterion was proposed recently to measure the influence of internal variations on secular trends in a time series. When the magnitude of the trend is greater than a theoretical threshold that scales the influence from internal variations, the sign of the estimated trend can be interpreted as the underlying long‐term change. Otherwise, the sign may depend on the period chosen. An improved least squares method is developed here to further reduce the theoretical threshold and is applied to eight sea surface temperature (SST) data sets covering the period 1881–2013 to investigate whether there are robust trends in global SSTs. It is found that the warming trends in the western boundary regions, the South Atlantic, and the tropical and southern‐most Indian Ocean are robust. However, robust trends are not found in the North Pacific, the North Atlantic, or the South Indian Ocean. The globally averaged SST and Indian Ocean Dipole indices are found to have robustly increased, whereas trends in the zonal SST gradient across the equatorial Pacific, Niño 3.4 SST, and the Atlantic Multidecadal Oscillation indices are within the uncertainty range associated with internal variations. These results indicate that great care is required when interpreting SST trends using the available records in certain regions and indices. It is worth noting that the theoretical threshold can be strongly influenced by low‐frequency oscillations, and the above conclusions are based on the assumption that trends are linear. Caution should be exercised when applying the theoretical threshold criterion to real data.

Global ocean primary production trends in the modern ocean color satellite record (1998–2015)

“…Although it is established that global sea surface temperature has increased over the past several decades (Hausfather et al 2017, Huang et al 2017, Lian et al 2018, our analysis shows no significant trend (after end-point bias correction). This is because our observational period 1998-2015, which is defined by the availability of modern ocean color satellite data and established data processing, unfortunately falls in a period bracketed by two major ENSO events, as The trends reported here are not corrected for autocorrelation.…”

A Global, Multiproduct Analysis of Coastal Marine Heatwaves: Distribution, Characteristics, and Long‐Term Trends

“…Shortening our time series to a common 25‐year time period from the 35 years available in most SST products had little impact on the estimation of the trends (not shown). Additional bias can be created in regions where the amplitude of low frequency internal variability is larger than the amplitude of the trend (Lian et al., 2018). The cooling observed in regions of the east Pacific and south Atlantic from the Multiproduct SST trend (Figure 2b) was also observed in long‐term trends computed from the NOAA OISST v2 product, with similar magnitudes (Lima & Wethey, 2012; Oliver et al., 2018).…”

“…Such cooling has been attributed to multidecadal modes of variability related to a persistent La Niña state in the Pacific Ocean, which has been shown to have caused the recent hiatus in global average surface air temperatures during the first decade of the 21st century (England et al., 2014; Kosaka & Xie, 2013). The underlying warming trend in the eastern Pacific, visible in long‐term SST reconstruction datasets (Lian et al., 2018), is therefore not captured within the recent 25 years. Despite this limitation, our results showed that internal variability was the main driver of MHW trends in the eastern Pacific (Figure 5) over the last 25 years.…”

“…Shortening our time series to a common 25-year time period from the 35 years available in most SST products had little impact on the estimation of the trends (not shown). Additional bias can be created in regions where the amplitude of low frequency internal variability is larger than the amplitude of the trend (Lian et al, 2018).…”