Oxford Contacts
- Principal Investigator: Professor Richard Washington
- Co-Investigators: Dr Giles Wiggs and Professor David Thomas
The overall aim of this research is to collect the first dust source-area process data tailored to climate model grid-box resolution from targeted remote sensing and fieldwork in order to develop a new generation of model dust emission schemes. The research links centrally to NERC's Next Generation Science for Planet Earth (2007-2012). In particular, the proposal relates to the core science themes of: The Climate System, especially the goal of developing an improved predictive capability; and Earth System Science, examining the linkages between the component parts of the system.
We will achieve the overall aim through four linked objectives which are to:
We will pursue these objectives through a programme of 4 focused research tasks, detailed below, carried out over 3+ years by a research team comprising leading experts within their fields of the atmospheric and environmental sciences. Please see the project poster for more information.
This research is needed because:
We propose to base the field component in southern Africa (Namibia, Botswana, South Africa) for several reasons:
Type | Description | Global Example | S African Example |
---|---|---|---|
A | large dry lake beds | Bodélé Depression | Etosha, Makgadikgadi |
B | seasonally dry river valleys and sabkhas | Mali and Mauritanian sources | Kuiseb, Namibia |
C | small closed ephemerally flooded depressions | SW USA | N Cape, Karoo |
D | sand dune systems | Tenere, Niger | Namib sand sea |
We will meet the four project objectives through four linked tasks outlined sequentially below.
Task 1 aims to a) identify and b) characterise the physical attributes of dust source regions in southern Africa using remotely sensed data. Imagery from SEVIRI flying on the MSG satellite platforms will be used to identify key dust sources in southern Africa. Because of its geostationary orbit, and through the use of its infrared channels (e.g. Schepanski et al 2009), and unlike TOMS/OMI, SEVIRI can monitor dust activity over southern Africa over the full diurnal cycle, providing measurements every 15 minutes with a spatial resolution of ~ 4km. In addition to source Types A and B (which are already acknowledged), we anticipate identifying dust plumes from numerous smaller pans (Type C) and other dust related features in the Kalahari dunefield (Type D) which are analogous to the small diffuse sources in the inaccessible Western Sahara as well as the plains of the Americas. Preliminary (unfunded) work on an incomplete temporal subset of data for 2005-2008 by Oxford and partners at UCT demonstrates that SEVIRI data are capable of identifying dust events in the study area (Figure 1). In addition, algorithms developed to obtain Aerosol Optical Depth (AOD) over the Sahara from SEVIRI (Brindley, 2007; Brindley and Russell, 2009) will be employed to provide a quantitative estimate of the dust loading in the region.
Figure 1: Annual averaged OMI AOT (2005 to 2008) with MSG-SEVIRI dust trajectories (2005-2008) overlain.
From Kathryn Vickery, MSc UCT.
Once dust producing source regions have been located by SEVIRI, we will use MODIS imagery (at 250m resolution) to examine specific events in more spatial detail than provided by the c4km resolution pixels of SEVIRI in order to both fine tune site selection and pinpoint the dust sources (e.g. Baddock et al 2009) at 12x12 km scale (matching the grid box resolution of a regional climate model).
Site characterisation by remote sensing, including the spatial and temporal heterogeneity in surface properties that are most relevant for erodibility (Okin 2005; Li et al 2007), and those that are required in the regional model dust emission scheme such as topography and vegetation cover, will also be undertaken in Task 1. A remote sensing approach, using multi-date SPOT 5, HRS panchromatic stereo-image pairs, will be used to generate high resolution (5m) digital elevation data providing height, morphology and surface roughness data for each study area. We will add additional detail by extracting fractional vegetation cover from SPOT 5 HRV data (10m resolution) and surface texture from HRS resolution panchromatic images. Combining the various SPOT 5 HRS/V-derived products will allow us to quantify bare ground, grass, shrub and tree cover as well as micro to meso-scale surface parameters. Existing time series of MODIS L1B and EVI (16-day, 500m, 2000-present; held at Sheffield) will also be used to characterise sources and areas within sources that are susceptible to supply limitation through significant annual/seasonal soil moisture variability or vegetation encroachment (Bryant et al., 2007).
Summary of Deliverables: T1D1 selection of 4-5 field study areas of Types A-D 12x12 km in size (the grid box size in the regional climate model) which act as analogues for dust production in the region and elsewhere on Earth. T1D2 Spatial characterisation and temporal behaviour of source areas erodibility characteristics determined from remote sensing. T1D3 Aerosol Optical Thickness for dust events during field observations.
Using T1D1 as our starting point, we will quantify the scale-variant parameters which control dust emission at each of the 12x12 km sites through a series of targeted experiments and process monitoring over a 2-year field campaign.
Each 12x12 km site will be divided into landscape units as defined by erosion susceptibility and determined with reference to Task 1 SPOT image analysis of vegetation and topography and also by expert analysis on the ground. Within each landscape unit (preliminary work suggests 3-5 such units per site) an equipment array will be erected to collect 1-minute time series data on friction velocity (6 m mast), wind direction, downward short-wave radiative flux and saltation impact (using a Sensit); hourly data on 10 cm and 2 cm depth soil moisture (using Theta probes) and dust concentration (using a DustTrak) and physical and chemical characteristics including size distribution (SEM and aerosol filter); and bi-monthly measures of horizontal dust flux (using a vertical array of BSNE samplers) and depositional flux (using 'frisbee' depositional traps). Sampling locations will reflect evident spatial variability in emission and deposition. Analysis of these data will provide the necessary information on the temporal and spatial variation of erosivity/erodibility, roughness length, wind erosion threshold (Stout 2004) and dust quantities within and between each landscape unit (contained in each 12x12 km site). On the primary dust producing landscape unit within each site, additional instrumentation will be erected to measure precipitation, vertical temperature gradient and aerosol optical thickness (AOT - using a Cimel photometer).
In addition, each 12x12 km site will be divided into 1x1 km grids. Within each of the 144 grid cells data will be collected for analysis of vegetation structure and fractional cover, surface particle size distribution and non-erodible surface fraction (including surface crusting). The survey will be repeated with regard to surface cover (vegetation, non-erodible and crusted fractions) for both a wet and dry season. Variation in these surface properties within each site will impact upon erosion thresholds and subsequent variation in dust emission. Such variation will be quantified using a PI-SWERL portable wind tunnel (Etyemezian et al., 2007) within areas of similar measured surface properties. Where large roughness elements (e.g. vegetation) are prevalent the PI-SWERL will be used on surfaces between the elements and in combination with a mobile anemometer/Sensit array within the roughness elements. An analysis of these data using shear stress partitioning (King et al., 2005) will allow the determination of the threshold for erosion on these rougher surfaces (Sweeney, 2009). Data from the 144 grid cells will be used to ground truth the remotely sensed surface characteristics (T1D2).
The field campaign will generate novel data on erosivity, erodibility, erosion threshold and dust quantity variation across key dust emitting surfaces at a variety of scales. It will provide the necessary data to derive an optimised dust emission model (Task 3) founded on the Marticorena & Bergametti (1995) emission scheme used in the Met Office climate model (HadGEM3-RA), to test it against field-measured dust quantities, and to analyse its sensitivity to changing spatial and temporal scales of inputs with regard to erosive force and surface resistance. It will also focus on gaining data recognised as of potentially critical importance to the dust emission system (e.g. the soil particle size distribution and 1-2 cm depth soil moisture, Darmenova et al., 2009) and allow comparative analyses of the Raupach et al (1993) and Okin (2008) erosion threshold models.
Summary of Deliverables: T2D1 Quantification of the spatial and temporal variability in erodibility/erosivity parameters and aeolian dust quantities and impacts (concentration, AOT and horizontal and vertical flux) within each of the study sites. T2D2 Entrainment thresholds in relation to changing surface characteristics. T2D3 Performance assessment of the Marticorena & Bergametti (1995), Raupach et al. (1993) and Okin (2008) threshold and emission schemes.
The aim of this proposal is to lay the groundwork for the development of a new generation of model dust emission schemes with the benefit of observational data sets that exactly match the scale of a regional climate model. Tasks 1-2 all serve as inputs to this modelling process. We propose to approach the problem in two steps.
Step 1: Box Model: We will start by setting up the modified Marticorena & Bergametti (1995) emission scheme, as used in the Met Office model (HadGEM3-RA), as an off-line model. The scheme is currently an integral part of the model and uses the model's prognostic variables. Each off-line box model will exactly match one of the 12km field sites. This approach is designed to limit the difficulties of spatially scaling up measures and processes from either point field data or idealised wind-tunnel experiments (Darmenova et al 2009). The parameters required within this scheme in the current version of HadGEM3-RA include friction velocity, soil moisture in top 10 cm, vegetation-fraction and soil particle size distribution. As a joint exercise between modellers and fieldworkers, we will optimise the emission scheme for the emission source types, first attending to where the emission scheme has previously been limited by field observations, missing controls and/or processes and/or computational requirements (e.g. roughness length, non-erodible gridbox fraction, limited number of particle size bins etc). Parameter measures will be derived from the field observations (Deliverables T2D1-2). Compared with previous commendable attempts to implement new emission schemes in models, which have involved interpolation of data such as particle size over many hundreds of kilometers (e.g. Laurent et al 2008), we are well positioned to limit the uncertainty associated with parameter measures. However, a key research problem involves specifying the box model parameter measurements based on multiple measurements from within the 12 km field sites to cope with factors such as space (and even time) dependent entrainment thresholds which most dust models take to be invariant. We will start by using a simple spatial average of parameter measures across the domain before testing the gains from weighting measurements based on the distribution of factors such as vegetation and non-erodible fraction. It will be the first time that an emission scheme can be so optimised. We will assess the degree of difference between the optimised emission scheme for the dust source types and, as a further step, attempt to develop one model which best represents all source types.
A series of 'denial' experiments is then planned. Having created an optimised emissions model, we will degrade the emission scheme parameter measures to 'climatological' values typical of those currently used in the emission scheme for this region which are not constrained by direct field observations. We will note the sensitivity of the model emissions resulting from the introduction of each realistic erodibility parameter measure ranging from the erodibility data retrieved from remote sensing only (T1D2) to the full suite of parameters obtained from the field observations (T2D1). To allow for non-linearities, these experiments will be repeated with the parameters introduced in different sequences. The errors introduced by erosivity measures will be assessed by driving the emission scheme with observed wind, simulated wind from the regional model driven by ERA-Interim reanalysis data and then the still coarser resolution reanalysis wind. This will help to isolate the role of erosivity characteristics such as gustiness. The contribution of erosivity and erodibility parameter measures to model dust emission will therefore be assessed on the basis of scale-matched field observations for the first time.
Step 2: Regional Model: We will transfer the emission scheme(s) developed in Step 1 to the regional model HadGEM3-RA which will have been set up at 12km resolution over southern Africa (including enhanced ancillary files such as vegetation; see Letter of Support, UCT) to exactly match the field site locations. We will make the necessary non-trivial adjustments to related parts of the model dust scheme (e.g. new particle size bins and radiation code dependence, soil moisture treatment and specification). We will then test and optimise the emission scheme against observed dust emission measures including: AOD, near surface dust concentrations (mg.m-3) and clear-sky downward shortwave flux (W.m-2) with bracketed uncertainties for the parameter measures and evaluation. Simulations using ERA-Interim boundary conditions with the optimised source area specifications will be compared with the standard model setup and the gain of using the optimised specifications will be assessed. As in Step 1, we will repeat the denial experiments to determine the sensitivity of the emission scheme to the parameter measures.
Summary of Deliverables: T3D1 Optimised emission scheme based on observed and remote sensing data. T3D2 Assessment of the contribution of erosivity and erodibility parameters to emission simulation.
Collectively these experiments will provide the basis for evaluating the relative merits of investing in a variety of more precise, high resolution data of erodibility and erosivity for the purposes of optimising dust models elsewhere. We propose to deliver a quantified strategy where the best returns on investment with respect to high resolution data (erodibilty versus erosivity) can be made. We will also be in a position to determine the relative gain of fieldwork over remotely sensed products and to establish the degree to which a 'one-size-fits-all' representation of parameters is adequate. These evaluations will be facilitated by a final Workshop to which additional stakeholders will be invited.
Summary of Deliverables: T4D1 Quantified benefits of source area data to dust emission schemes and strategy for gaining data on source areas.
The diagram below indicates the linkages and progression between the 4 research tasks of the research programme and the timing of activities by the three PDRAS (PDRA 1 linked to Task 1, PDRA 2 linked to Task 2, PDRA 3 linked to Tasks 2-5).
We will hold two week-long project workshops with the key researchers as participants: the first at the end of year 1 to focus on field study area data and parameterisation, the second six months before the close of the project to the outputs of the numerical modelling. While we anticipate the presentation of findings at many meetings and conferences, we highlight the EGU and the AGU as especially significant meetings at which to showcase key outcomes. Definitive findings will also be presented at the 4-yearly ICAR meeting in 2014 (after the project).
Principal management of the project will be undertaken by the PI and Co-Is in Oxford (Washington will be on leave for 2011) with specific specialist tasks undertaken by our PIs at Imperial and Sheffield and sub-contracted collaborators at the Hadley Centre and the Department for Environment and Geographical Sciences at the University of Cape Town. Resources will be managed from Oxford: the PI has considerable experience of managing major research projects. The project involves 3 PDRAs who may have emerged from quite focused and specialised doctoral research and they will benefit from the inter-disciplinarity of the proposed project. PDRA1 will work on remote sensing but interface with both fieldwork and modelling. PDRA2 will receive supplementary training in remote sensing techniques, advanced field skills, aeolian sediment transport modelling, and meteorological analysis. PDRA3 will receive any necessary additional training in climate modelling and will have the opportunity to work directly with the source code of one of the finest climate models in the world in a setting that combines field work and numerical modelling - a rare but necessary combination for progress in this field. While each PDRA will play to their own strengths in each task, they will gain an appreciation for the challenges of quantifying the Earth System and the requirements and uncertainties involved in predictive modelling at different scales. Linked to this, OUCE has many postdoctoral researchers and an excellent record of them gaining further employment beyond the projects on which they have worked. This, we believe, will equip them well for successful research careers. Project management will take the form of bi-weekly progress reports, monthly Skype calls, 4 day long meetings for UK partners as well as the scheduled workshops. Field data will be available from BADC.
Key: MS1 = Cape Town Workshop. MS2 = Oxford Workshop. MS3 = Publication Outputs Completed.
CF1 = EGU 2012. CF2 = AGU 2012. CF3 = AGU 2013.