Real-time and seasonal forecasting of wheat rust epidemics to inform surveillance and control: Ethiopia as a LMIC test case

Wheat rusts are fungal diseases that can cause devastating losses to wheat crops. The fungal spores are dispersed by the wind with the potential to cause a rapid spread of a wheat rust outbreak. Whilst fungicides to control the spread of wheat rust are available they need to be applied in a timely manner in order to be effective. Failure to do so can lead to the loss of crops with the added economic costs of wasted fungicide. Wheat rust infects the crop causing the death of the plant and therefore a reduction in the wheat produced in a region; this affects both an individual growers livelihood and on a wider scale the costs incurred as a government imports wheat to supplement in-country wheat produce.

Wheat is a staple food in Ethiopia, as it is in many countries in the world. Ethiopia is also the major wheat producer in sub-Saharan Africa. Thus, outbreaks of wheat rusts have a major impact on the wheat available in the region. The Government of Ethiopia has committed to becoming self-sufficient in wheat by 2020. To meet this policy goal and ensure food security, production losses, such as those due to wheat rusts, must be minimised. The proposed work aims to contribute to this goal by implementing a wheat rust forecasting system in collaboration with the Ethiopian Institute of Agricultural Research (EIAR).

The proposed work will combine expertise on modelling plant diseases at the University of Cambridge, expertise on modelling releases of particles into the atmosphere at the UK Met Office with expertise within Ethiopia at CIMMYT and EIAR to develop a world-leading, wheat rust forecasting service. Information on the location of initial sites of wheat rusts in Ethiopia will be sent to the UK Met Office. This information will be collated from surveillance surveys and from farmer reports using mobile smart-phone technology. The UK Met Office will use forecast data from its state-of-the-art global weather forecast model together with the source information and protocols for spore dispersal developed with the University of Cambridge to forecast the large-scale wind-dispersion of the wheat rust spores. The University of Cambridge's epidemiological model will use the outputs to model the smaller-scale spread of the rusts. This, together with an assessment of the suitability of an area for wheat rust, will be used to produce a forecast risk map (where is the pathogen likely to occur?) and a hazard map (how rapidly will it spread and cause losses once introduced).

The development of products, such as wheat rust forecast risk maps, will be done in consultation with EIAR to ensure that they contain the most useful information for their onward use. EIAR will develop methods to make best use of the forecasts for advising the Ethiopian Government and farmers in order to prepare for potential wheat rust outbreaks and to enable them to take action to prevent the loss of crops. The proposed work includes workshops, exchanges and training to aid capacity building and exploitation of the wheat rust forecasting service. In doing so, strong links between the Ethiopian and UK organisations will provide a basis for further development to support other wind-borne agricultural diseases and demonstrate the value of such a system for other counties.

Puccinia graminis and P. striiformis, stem and stripe wheat rust, represent the greatest biotic constraint to wheat farmers in Ethiopia, threatening both food security and livelihoods for a large proportion of the 5 million wheat farming households. The Ethiopian Government has made control of wheat rusts a top priority in their ambition for the country to be self-sufficient in wheat by 2020. This proposal will lead to the development of an operational wheat rust forecasting capability to support the role of the Ethiopian Institute of Agricultural Research (EIAR) to help mitigate the impact of wheat rust through timely advice to farmers and the Ethiopian Government.

We propose to develop the wheat rust forecasting capability by combining the modelling strengths of (i) the Numerical Atmospheric-dispersion Modelling Environment (NAME), a Lagrangian model that can forecast the large-scale atmospheric spread of rust disease and (ii) a stochastic, highly-spatially resolved compartmental epidemiological model that can predict the spread of rust disease at field and landscape scales. The modelling will be driven by UK Met Office numerical weather prediction data at 10 km spatial resolution and will produce routine daily forecasts for up to 7 days ahead. Products of wheat rust risk generated by combining the model output with rust suitability maps will be disseminated to EIAR. The source locations of wheat used as inputs NAME will be obtained from surveillance surveys by CIMMYT and the novel use of mobile phone reported outbreak data sent by farmers.

Close collaboration between UK (University of Cambridge, Met Office) and Ethiopian (CIMMYT, EIAR) experts via exchange visits and workshops will result in the development of wheat rust forecast products tailored for in-country use with critical exchange of knowledge between the organisations.

The pathways to impact extend from core modelling activities (Cambridge and the UK Met Office), through input of data, validation and implementation (via collaborators in CIMMYT and Ethiopian Government Institutes) to direct impacts on wheat-farming families in Ethiopia (via national agencies).

The recently strengthened, Early-Warning Unit within the Ethiopian Institute of Agricultural Research (EIAR), supported by the UK institutions and CIMMYT, will be the critical interface between the model-derived information products and the national agencies responsible for rust control in Ethiopia. The Pathways allow for reciprocal flows of information, linking model forecasts and actual disease status directly with mandated national control agencies. We are confident that the arrangements are carefully designed to result in improved rust control and considerable benefits for wheat farmers in Ethiopia. Both the Epidemiology and Modelling Group at Cambridge and the Atmospheric Dispersion and Air Quality Group in the UK Met Office have experience in using models to inform and solve strategically important practical problems. CIMMYT has a mandate to improve farmers' livelihoods and the group led by Dr D. Hodson has established close contacts with EIAR, the Ministry of Agriculture and Natural Resources (MoANR) and the Agricultural Transformation Agency (ATA).

Direct Project Impacts: Core project modelling activities are seen as the drivers of the impact pathway. The feasibility of developing short-term, rust spore dispersal forecasts in Ethiopia has already been demonstrated. This project will rely on cutting-edge modelling and forecasting technologies from leading UK research Institutes to enhance and improve information products which Ethiopian partners can utilise to achieve impact on rust control. Work under Obj. 1 will result in two critical products (i) the coupling of spore dispersal with climatic suitability to develop accurate short-term forecast maps for risk of rust outbreaks (ii) development of an epidemiological model to predict likely spread of disease and a platform to assess the effectiveness of different control strategies. Initial risk maps will be developed and tested in Ethiopia during the first main cropping season (Aug-Dec 2017), then refined for the second cropping season (Aug-Dec 2018). Functional epidemiological models will be tested after 1 year and refined in the second year. Critical to achieving long-term impact from the project is developing the capacity and operationalising the rust early warning unit within EIAR (Obj. 2). Capacity building will be achieved through a series of workshops, in-country training courses and scientist exchange visits. These will take place in year 1 and the first half of year 2. Field survey and rust phone survey data will be critical inputs used to validate the forecast model outputs (Objective 3). The epidemiological models (developed in Obj. 1) will be used as a platform to test optimal control strategies using what-if scenarios in year 2.

Top-level Impact: The overall top-level impacts from the proposed work are improved and earlier rust control for wheat farmers in Ethiopia, leading to improved food security and income generation for wheat farming families. It is envisaged that direct impacts on farmers would be realised by the end of foundation project. Successful realisation requires, several factors outside of the direct influence of the project that include the procurement and efficient deployment of fungicides and capacity of the farmers to act on the improved information flows.

Long-term sustainability: Given the high political priority for rust control in Ethiopia, a functional national early warning unit with the capacity to improve rust control, optimise resources and enhance wheat productivity is likely to be institutionalised and attract continued national support.