Interpreting Earth's top-of-the-atmosphere broadband radiation flux variability using observations and models

Abstract

Observed broadband radiation fluxes at the top-of-the atmosphere (TOA) and at the Earth’s surface are determined by a complex network of atmospheric and surface processes. It is imperative that climate models are able to accurately simulate these observed variations and relationships in order to provide confidence in projections of our future climate. In this thesis I use a combination of observations, reanalysis fields and output from global circulation models (GCMs) to interpret radiation flux variability with respect to atmospheric properties and processes, in particular clouds, atmospheric water vapour and aerosols. I use observations and models in two ways. In Chapters 3 and 4 I evaluate model output using observations from satellite instruments and surface measurement stations to characterise the model ability to 1) recreate observed variability and 2) contrast TOA and surface radiation flux co-variability with atmospheric properties. In Chapter 5 I use satellite observations of atmospheric temperature and humidity profiles, as well as broadband radiation flux, to assess evidence of physical mechanisms which have recently been hypothesised using output from GCMs. The chapters are based on two regions of the tropics. I focus on the first of these, a region in western Africa, partly due to the presence of aerosols, such as Saharan mineral dust, and also the west African monsoon. Both of these factors have large impacts on the radiation balance and therefore make this region interesting from a radiation perspective. Additionally, west Africa is a region vulnerable to changes in climate, having already suffered from extended droughts in the last decades. My second focus region is the tropical ocean, where changes in tropical low clouds play an important role in the TOA radiation balance, and has therefore been linked to climate model sensitivity. The spatial and temporal scales used in the studies vary dramatically, which determines both the model output evaluated and also the methods I employ. In Chapter 3, I exploit the 2006 high frequency observational data at Niamey, Niger from the Atmospheric Radiation Measurement (ARM) Mobile Facility, the Geostationary Earth Radiation Budget (GERB) and Spinning Enhanced Visible and Infrared Imager (SEVIRI) instruments, and products from the Climate Mon itoring Satellites Applications Facility (CMSAF) to evaluate daily output from the European Centre for Medium-Range Weather Forecasts (ECWMF) Integrated Forecasting System 43r1. The data available include surface, atmospheric pro file and TOA measurements. By constructing multi-variate linear models of each component in the energy budget, I test their sensitivity to changes in atmo spheric properties, including 2m air temperature, aerosol optical depth (AOD), cloud properties and total column water vapour (TCWV). I find that the lack of ice in clouds, manifested as a reduced ice water path (IWP) in 43r1 with respect to the estimate from CMSAF, results in too much shortwave radiation passing through the atmosphere in 43r1, and therefore too much downwelling shortwave radiation (DSR) at the surface and too little reflected shortwave radiation (RSR) at the TOA. I also identify the use of an aerosol climatology in 43r1 as a cause of discrepancy between the observation and the model in the surface fluxes, with the lower aersol loading in the model leading to a reduction in downwelling longwave radiation (DLR) and an increase in DSR. This work is published in Atmospheric Chemistry and Physics as Mackie et al. (2017). In Chapter 4, I examine a wider region in western Africa, which I refer to as ‘west Africa’, which encompasses three distinct sub regions: the Sahel, the Sahara and the south-western coastal region. As observational references, I use a range of radiation data from the TOA and surface from satellite products and surface station measurements to construct mean annual cycles with which to evaluate output from GCMs submitted to the Intergovernmental Panel on Climate Change’s Coupled Model Intercomparison Project Phase 5 (CMIP5). This chapter has two aims: firstly, to compare the reference data and to establish the observational range in the targeted metrics, and secondly to evaluate how the CMIP5 multi-model mean and range fit with this range. By contrasting coupled and atmosphere-only model output, I link differences in radiation at the TOA to the models’ tendency to model the west African monsoon onset too late and to model the limit of its northwards progression to too far south. By contrasting the sensitivity of the models to changes in AOD and TCWV to that of the Clouds and the Earth’s Radiant Energy System Energy Balanced and Filled (CERES EBAF) product, I find some indication that DSR in the CMIP5 models may be too sensitive to changes in water vapour, and not sensitive enough to changes in AOD. This work is under review at the AGU journal Earth and Space Science. In Chapter 5, I evaluate observational evidence for a model-based hypotheses which links tropospheric temperature and humidity changes to patterns in tropical sea surface temperature (SST) warming. The hypothesis states that if SSTs in regions of strong ascent warm relative to the tropical ocean mean, the warming is efficiently lofted to the upper troposphere. In contrast, if warming is concentrated in regions of subsidence, the effects are limited to below the inversion which is characteristic of these regions. The subsequent effects of SST warming patterns are hypothesised to be key in determining the feedbacks from low cloud, and has thus been linked to climate sensitivity. To test this hypothesis I use co located Atmospheric Infrared Sounder (AIRS) temperature and humidity profiles and CERES radiation data, including window region data, and subset these data using vertical velocity at 500 hPa from ECMWF’s ERA-Interim reanalysis. I find some evidence which supports the hypothesised mechanism, specifically that if subsiding regions warm preferentially, there is a strong decrease in low cloud, with associated decrease in reflected shortwave radiation (RSR), and evidence that temperature increases are suppressed above the inversion. I also find small, but statistically significant, increases in humidity above the boundary layer inversion, though the origin of this is not clear. If regions of convection preferentially warm, the observations suggest that changes in relative humidity in the upper troposphere are due to changes in specific humidity rather than temperature, with temperatures in the upper troposphere relatively insensitive to relative warming. The largest changes in TOA radiation are in the longwave, which I hypothesise are linked to the observed increase in high cloud. This work is being prepared for publication.