Met Office Academic Partnership Summer Internship Reports:

  1. The effects of volcanic eruptions on the dynamics of the stratosphere
  1. Reconstructing AMOC variability using wind and buoyancy forcing anomalies
  1. Mechanisms of Atlantic Decadal Climate Fluctuations and Assessment of Forecast Skill
  1. Decadal-scale Climate Variability on the Central Iranian Plateau Spanning the So-called 4.2 ka BP Drought Event
  1. Controls on atmospheric blocking under climate change
  1. Improving Probabilistic Weather and Climate Predictions
  1. Improving satellite constraints on volcanic ash cloud heights
  1. Convectively-driven baroclinic flows in the laboratory as a model for baroclinic adjustment

1. The effects of volcanic eruptions on the dynamics of the stratosphere

Student: MattPatterson

Supervisor: Lesley Gray (AOPP)

Matt’s project involved investigating the effects of volcanic eruptions on the dynamics of the stratosphere. Following a sufficiently large eruption near the equator, aerosol particles are injected into stratosphere and form aerosol clouds. These clouds absorb radiation from the Sun and the Earth, heating the lower stratosphere. This temperature anomaly indirectly affects the propagation of waves in the atmosphere and Matt used diagnostics from the ERA Interim, NASA Merra, JRA55 and NCEP-CFSR datasets like EP flux and residual circulation to study this. Previous work by Graf had suggested that wave propagation from the troposphere into the stratosphere actually increased during winters following eruptions. This was surprising, but his analysis found that these more up-to-date datasets all show the same result.

Figure 1.Difference between years with and without a volcanic eruption in EP flux

Figure 1 above shows the difference between years with and without a volcanic eruption in EP flux, (F). F is displayed as vectors with div F in the background. Red areas of the plot indicate a source of wave activity. The plot shows that following volcanic eruptions, the amount of waves produced during winter in the Northern hemisphere troposphere is increased. However, the increased wave forcing doesn’t seem to strongly affect the stratospheric polar vortex. The main volcanic signal is a strengthened vortex due to the enhanced temperature gradient between the equator and poles.

Matt found the placement was a really helpful insight into how climate physicists work and gavehim a taste of and for research. He enjoyed doing genuinely original work and talking to scientists who are passionate about their subject!

As his supervisor, Lesley had a great experience working on this project with Matt, stating: “Over the summer It has been a pleasure to have Matthew working with us this summer. It has been a really useful opportunity to look at an aspect of volcanic forcing that I’d always meant to have a look at but never quite got around to. Matthew very quickly came up to speed and has found some really interesting results using more up-to-date reanalyses. It confirms a surprising result from a few years back, and has helped me to decide that it’s worth looking further, to see if models reproduce the results. So it’s been really helpful to my research, and is a great chance to help encourage potential new DPhil students – so a double plus!”

2. Reconstructing AMOC variability using wind and buoyancy forcing anomalies

Student: So Takao

Supervisors: Helen Johnson (Earth Sciences) and David Marshall (AOPP)

So Takao worked with Helen Johnson and David Marshall in the Physical Oceanography group, together with Helen Pillar at the University of Copenhagen, to reconstruct variability in the Atlantic meridional overturning circulation (AMOC).

The AMOC is a large scale ocean circulation in the Atlantic that carries warm, saline water northward and brings cold, dense water southward at depth. The phenomenon is closely related to the heat transport of the Atlantic ocean and is therefore believed to play an important role in the climate of Western Europe. The AMOC at 26°N has been continuously monitored since 2004 by oceanographers from the National Oceanographic Centre in Southampton and their US colleagues and this has resulted in an unprecedented 10 year time series of the mean AMOC strength at 26°N, which has revealed large variability on different timescales. Knowledge of this variability could be useful in reducing uncertainties in future climate predictions.

So’s summer project focused on reconstructing the time series of AMOC variability from a range of observed forcing datasets, by projecting surface wind and buoyancy forcing anomalies onto linear sensitivity patterns, which were calculated using the adjoint of the MITgcm ocean circulation model. Our reconstruction using NCEP reanalysis II (1979-2015)

forcing successfully reproduces the AMOC variability on short time scales, which can be attributed to wind forcing, but diverges from the observed AMOC on longer time scales when the response to surface heat fluxes becomes more important. We believe that this divergence is caused by limitations of our linear sensitivity approach, which prevents us from capturing the effects of forcing anomalies on time scales longer than 15 years, and on differences between the decadal variability of the model and the real ocean.

We have used this reconstruction to “predict” the AMOC for the 15 month period (April 2014 to June 2015) for which observational data at 26N are not yet available. When the RAPID team of oceanographers recover their moorings this autumn, we expect that they will find a mean AMOC over this period roughly equal to that over the past few years (a change of within 0.3 Sv from the 2009-2014 mean which was 15.6 Sv), and with no significant weakening events over the past winter. This prediction forms the basis of two blog articles (on the RAPID and OSNAP project websites).


By using more historic surface forcing data such as that available via the NCEP reanalysis I (1948-2015) and 20th century reanalysis (1851-2011) products, we were also able to obtain longer time series of AMOC variability, which opens up the possibility of understanding past behaviour of the AMOC. We intend to write a paper comparing the AMOC variability that results from different reanalysis products; investigating how much the AMOC anomalies they generate deviate from one another will help to understand the accuracy of these products. In the meantime, So has started a Masters degree in Applied Maths at Imperial College.

3. Mechanisms of Atlantic Decadal Climate Fluctuations and Assessment of Forecast Skill

Student: Ben Huddart

Supervisors: Aneesh Subramanian (AOPP), Laure Zanna (AOPP), Tim Palmer(AOPP)

The sea surface temperatures (SSTs) and climate in the Atlantic sector are well known to vary on seasonal to decadal timescales with a large impact on rainfall and temperature over Europe, Eastern US and Africa. However the mechanisms controlling these long-term climate variations are still unknown. Understanding the mechanisms associated with these long-term oceanic and atmospheric fluctuations in the Atlantic sector can potentially lead to more accurate assessments of climate predictions and associated uncertainties. Seasonal and decadal predictability of Atlantic climate and SSTs is of particular interest to the U. K. Met Office and Europe in general, due to its downstream influence over the U.K. weather and its direct influence on climate in Europe. This leads to the primary focus of our project, which is to identify the mechanisms leading to predictive skill of seasonal and decadal modes of the Atlantic Ocean system.

There is clearly a gap in our understanding of drivers controlling Atlantic seasonal and decadal climate variability and its predictability, what timescales and regional variability contributes most to the predictability. We focused on using sophisticated statistical, yet dynamically based, models (Linear Inverse Models, LIM) to identify statistical relations among variables, diagnose physical processes, and isolate potentially predictable components of the flows.

The LIM is a simplified reduced stochastic-dynamic climate model (Majda et al., 2009). The governing dynamics of the system in LIM is modeled as:

dx = (Lx + ξ) dt (1)

where x represents an appropriate system state vector, L is a linear dynamical operator matrix, and ξ is a vector of stochastic Gaussian noise. For such a system, L can be estimated from observational estimates of covariance. Eq. 1 can then be solved analytically for different lead times τ: x(t + τ) = G(τ)x(t) + ε where G(τ) = exp(Lτ) represents the decaying, predictable signals at forecast lead time τ and ε is the nonlinear stochastic unpredictable component. The dynamical operator, L, contains summarizing information about the damped and oscillatory behaviors of the interacting modes in x, while the scaling covariance matrix ξ summarizes unresolved high-frequency dynamics, parameterized as multivariate normal noise. We analysed the LIM interaction matrices in novel ways for seasonal and decadal mode diagnostics to study the patterns and regions that control the coherent behaviours in the decadal modes.

We use LIM to build a reduced climate state statistical model using observed Atlantic SST anomalies between latitudes 20°S and 66°N from 1870 to present. Additionally, subjective data pre-filtering are performed in order to isolate climate signals of interest. We refer to this model as the filtered LIM. The filtered LIM analysis involves attempting to identify time-scale interactions among decadal modes and between decadal and interannual to intraannual modes. This is achieved through spectral analysis and analysis of LIM modal interactions. The skill and mechanisms are then evaluated against the state-of-the-art ECMWF seasonal forecasting system for seasonal SST anomalies and Met Office forecast system for decadal SST anomalies (SSTa) to understand the seasonal and decadal skill in the Atlantic Sector.

On the seasonal timescale, the LIM forecasts show skill for lead times of up to 4 months in spring time (Apr-Jun) and have the least skill in winter (Jan-Apr) and summer (Jul-Sept) of only about 2-3 months before the forecast root mean square error exceeds climatological forecast error (Fig. 1).

Figure 1. RMS error relative to climatological forecasts for two year forecasts initialised at different months. The dotted line indicates the average over all months. A relative error of 1 indicates that the forecast error is of similar magnitude as assuming the climatological SSTa as forecast. Inset: Initial growth rate of absolute RMSE of forecast SSTa.

We also evaluate the forecast skill using an anomaly correlation metric in time at every grid point. We then show that the correlation deteriorates most rapidly in the sub-polar gyre region, with very little correlation between the forecasted and the observed SST anomalies beyond 3-4 months in this region. This forecast skill, measured as anomaly correlation, is highest in the Tropics, with reasonable skill (positive correlation) upto 12 months lead time. The error growth rate in LIM forecasts are similar to that seen in ECMWF seasonal forecasts, with the RMSE increasing by about 0.3oC over the first six months.

On the decadal timescales, we use the filtered LIM to explore the predictability of the decadal component of the SSTa. Forecast skill is lost in time periods of 3-4 years over the entire North Atlantic basin. The spatial maps of temporal anomaly correlations show that the sub-polar gyre has the highest predictability compared to other regions in the N. Atlantic on decadal timescales (Fig. 2). Using the filtered LIM with only decadal components of the signal as opposed to using higher frequency signals (inter-annual and intra-annual) tends to reduce the skill over all regions in the basin (Fig. 3).

Figure 2 Temporal correlation over the North Atlantic region for a range of lead times using the filtered LIM with only the decadal component of SSTa.
Figure 3 Temporal correlation over the North Atlantic region for a range of lead times using the filtered LIM with both the decadal component and the higher frequency (inter-annual and intra- annual) components of SSTa.

Hence, the higher frequency SSTa signal tends to help improve forecast skill on decadal timescales over all regions in the basin, and most in the tropical and sub- tropical regions of the basin. Coupling was mainly observed between interannual and decadal modes of variability with very little or no coupling between the intraannual modes and decadal modes for the decadal forecast experiments.

We show from this study that there are modes of variability in the decadal and interannual frequencies both in the tropics and extra-tropics, which influence the predictability of SST anomalies on seasonal to decadal timescales. Further analysis will be required to understand the physical mechanisms by which these modes couple and how they influence the predictability on these timescales. This and other aspects of the system which influence predictability over the North Atlantic region will be explored in future studies to help improve weather and climate prediction in this region.

Reference:

Huddart, B., Subramanian, A., Zanna, L. & Palmer, T. N., 2015: Seasonal and decadal forecasts of Atlantic SST using a Linear Inverse Model. J. Clim.,In Prep

4. Decadal-scale Climate Variability on the Central Iranian Plateau Spanning the So-called 4.2 ka BP Drought Event

Student: Luke Maxfield

Supervisor: Stacy Carolin (Dept. Earth Sciences)

Luke’s summer placement was with the Isotopes and Climate research group at Oxford's Department of Earth Sciences, supervised by Dr Stacy Carolin, a palaeoclimatologist in the department.

His project was a palaeoclimate study of Western Asia, and involved geochemical analysis of a speleothem. Cave speleothems can provide extremely good records of past climates, by encoding climatic information in their isotopic composition as they grow. The speleothemhe was working on, GZ14-1, was sampled from Gol-e zard cave in northern Iran. Northern Iran is an interesting area to investigate past climate change as the regional atmospheric circulation is thought to be sensitive to a number of Earth's climate systems, such as the Indian Summer Monsoon (ISM) and the North Atlantic Oscillation (NAO), though these relations are currently not well understood.

He first used a MicroMill to drill out and store over 400 calcite powder samples at 250 or 500 μm intervals along the growth axis of the stalagmite. The age model of this stalagmite shows that this sampling interval produces samples at ~5-yr resolution. From these samples, he extracted 30-60 μg calcite powder, which he placed into glass autosampler vials in preparation for the analysis of stable oxygen and carbon isotopes on the Delta V isotope ratio mass spectrometer (IRMS). He also extracted an additional 70-90 μg of calcite powder from the same samples in preparation for trace element analysis on the Element inductively coupled plasma mass spectrometer (ICP-MS). Overall, he assisted in preparing and running over 100 samples on both instruments.

Over the eight weeks, I discussed relevant scientific literature with my supervisor to provide some context for the work I was doing. This introduced me to an interesting application of palaeoclimate studies: understanding how changes in climate have influenced the development of human civilisations. A particularly interesting topic of debate is a 4.2ka BP ‘drought event', which has been linked to the collapse of the Akkadian Empire (Staubwasser & Weiss, 2006). My preliminary stable oxygen isotope results reveal a large 40-yr excursion in the record at 4440 BP and a smaller 30-yr excursion at 4250 BP, based on the tentative age model (Figure 1).

Figure 1 Iran stalagmite GZ14-1 stable isotope record, inferred as a rainfall proxy with more positive (down) indicating drier conditions. Yellow bars highlight stable isotope excursions.

Currently, it is difficult to interpret the results based on the limited number of measurements that have been analysed on the instruments, however the samples that I drilled and prepared are set to reveal exciting new climate data from this region that is absent in the literature to-date.

What he found most valuable about the placement was learning about the scientific process, from deciding a research question to processing data and writing a paper. Luke found it a great experience and is now looking forward to seeing future results. The completed dataset will be presented at the 2015 American Geophysical Union (AGU) Fall Meeting in San Francisco this December.

5. Controls on atmospheric blocking under climate change

Student: Daniel Kennedy.

Supervisors: Tim Woollings(AOPP) and Tess Parker (AOPP)

Blocking is a high impact weather pattern in which the usual westerly flow in midlatitudes is blocked by a persistent, quasi-stationary anticyclone, leading to severe cold in winter and heatwaves in summer. These complex dynamical events have long been a challenge for weather and climate prediction models, and so confidence in projected blocking changes remains low.

In this project we diagnosed blocking events in a set of HadGAM1 model experiments designed to probe uncertainties in the storm track response to climate change. These experiments perturbed the upper and lower level equator-to-pole temperature gradients separately to assess the competing effect of these changes on the storm track. The main result is that projected blocking changes are more sensitive to changes in the upper level temperature gradient, and hence to uncertainties in the tropical warming rather than the level of polar amplification.

Second order effects also emerged, for example the temperature anomalies experienced blocking are projected to change. The strongest example of this is in winter, when the extent of the cold anomaly is dependent on the amount of Arctic warming. If the Arctic warms strongly, the area of Eurasia affected by extreme cold temperatures during blocking is greatly reduced.

6. Improving Probabilistic Weather and Climate Predictions

Student: Michael Varley

Supervisor: Peter Watson (AOPP)

The focus of Michael’s project was weather forecasting, and throughout the summer he was working with a numerical 'toy model' of the atmosphere called the Lorenz '96 system. This system has been used to test other proposals for improving atmospheric models in the past, before they were subsequently implemented in Numerical Weather Prediction Models (NWPs).