Tufa Dinku^1*, Augustine Kanemba², Barbara Platzer³ and Madeleine C. Thomson^1,4

1 International Research Institute for Climate and Society, Earth Institute, Columbia University- Lamont Campus, Palisades, New York, USA
2 Tanzania Meteorological Agency, Dar Es Salaam, Tanzania
3 Columbia Global Center-Africa, Nairobi, Kenya
4 Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, USA

xThis article introduces the latest tools and services piloted by the Tanzania Meteorological Agency (TMA), with technical support from the International Research Institute for Climate and Society (IRI), in service to the national health community. A project called ‰ÛÏEnhancing National Climate Services‰Û (ENACTS) has been implemented in Tanzania with focus on users from the public health community.

Abstract:

While the potential usefulness of climate in health decision-making is increasingly recognized, few countries in Africa have the capability to routinely provide the health community with relevant, accurate and timely information that can readily be integrated into decision-support tools. This stalemate is beginning to change as new high-quality information services are being established in some African countries using a blend of quality controlled national observations and the best available remote sensing and other products. This new approach, ‰ÛÏEnhancing National Climate Services‰Û (ENACTS), designed to improve the availability, access and use of climate data, includes generating historical rainfall and temperature data that have the potential to transform the capacities of national meteorological agencies in partnership with stakeholders and research collaborators. This article introduces the latest tools and services piloted by the Tanzania Meteorological Agency (TMA), with technical support from the International Research Institute for Climate and Society (IRI), in service to the national health community.

Introduction:

Public health policymakers and practitioners are increasingly concerned about the potential impact of climate, environmental and social changes on the effectiveness of current and future vector-borne disease control and elimination programs. Yet, while climate change adaptation programs are increasing in scope and resourcing, there remains an identified gap in research and professional capacity to use climate information in decision-making. In the health sector in particular, many control programs of climate sensitive diseases (such as malaria) are not informed by grounded knowledge and information on the climate. This is because few public-health institutions or practitioners are equipped to understand or manage the effects of a changing climate, despite major advances in recent years in alerting the health community to its risks. A key challenge that has been identified is ‰Û÷market atrophy,’ a comparative lack of demand from the health sector for climate services coupled with a lack of supply of relevant, actionable information (as there is often no clear demand).

Malaria remains a major cause of death and illness worldwide with more than 500,000 deaths each year and more than 200 million cases. It is widely identified and studied as the most climate sensitive vector-borne disease, which occurs in geographic areas conducive to the malaria parasite and its mosquito vector. åÊIn the absence of control, the spatial and seasonal risk of the disease, as well as differences from one year to another and in long-term trends, are often governed by climatic factors such as rainfall, temperature and humidity.

It is clear that climate is only one of many important drivers of malaria (e.g., education, migration, land use change, control measures). Climate information is different, however, in that it is ideally suited for integration into information systems for the health sector. This is owed to both the nature of climate (its climatology, seasonality, diurnal rhythm and potential predictability at multiple time scales) and the fact that it is routinely measured in a systematic way by land observations, remote sensing and global model outputs all around the world. Climate information therefore has the potential to inform a wide range of health decisions and improve our understanding of the following:

‰Û¢ Mechanisms of Disease Transmission: to help identify new opportunities for intervention

‰Û¢ åÑ Spatial Risk: to help identify populations at risk for better targeting of interventions

‰Û¢åÑåÊ Seasonal Risk: to inform the timing of routine interventions

‰Û¢ Sub-seasonal and Year-to-Year Changes in Risk: to identify when changes in epidemic risk are likely to occur to initiate appropriate prevention and response strategies

åщۢ Trends in Risk: to identify long-term drivers of disease occurrence (including changes in the climate) to plan for and support future prevention and response strategies

‰Û¢åÑ Assessment of the Impacts of Interventions: to evaluate the role of climate as it enables or limits disease transmission.

Figure 3: Comparison of station measurements (left), satellite estimates (middle) and merged products (right) for a given 10-day period. The top panel is for operational stations while the bottom one is for all available stations.

The ‰ÛÏEnhancing National Climate Services‰Û (ENACTS) approach aims at simultaneously improving the availability, access and use of climate information [3]. It has the potential to transform the capacities of national meteorological services to respond to and invest in the research and operational interests highlighted above. Now available for Tanzania, Ethiopia, Madagascar and at the regional level in West Africa, this approach has been designed to overcome challenges due to the decline in the number and quality of weather stations in many parts of Africa and the fact that available stations are often unevenly distributed with most of the stations located along the main roads. These twin challenges impose severe limitations to the availability of climate information and services to communities where these services are often needed most. Where observations are taken, they suffer from gaps and poor quality and, because of policy and staff constraints, are often unavailable beyond the respective national meteorological services.

Data availability can, however, be improved by combining available local observations with satellite and other proxies. Access and use of this information can further be improved by making information products openly and readily available, as well as by working with stakeholders to better understand and use the information products. This offers hope in alleviating the challenges outlined above and underpins a new approach to improved climate resilience in Africa. This report describes how ENACTS has been implemented in Tanzania with focus on users from the public health community.

Building Bridges between the Climate and Health Communities in Tanzania

Leveraging the recently piloted ENACTS products, the Tanzania Meteorological Agency (TMA) recently hosted a stakeholder meeting in Dar es Salaam from Oct. 16-18, 2013, to discuss opportunities for the use of climate information specifically in malaria prevention and control. A number of different opportunities for using climate information in Tanzania were discussed, including the use of the new ENACTS products in malaria impact assessments.

Figure 1 presents how change in climate suitability for malaria transmission might impact intervention efforts. Central to the evaluation of development interventions for malaria reduction and prevention is to use a baseline year or period in order to measure changes in outcomes, such as malaria incidence. If the baseline year (or period) was unusually wet or dry (warm or cool) for the particular outcome, then achieving change relative to that baseline may be confounded as a result of variability or trends in the climate. For example, if during a relatively dry baseline year or period mosquito nets are introduced and malaria incidence declines, it may be tempting to attribute the decline of malaria incidence to the introduction of mosquito nets alone without considering the impact of climate. However, should the rains return and malaria incidence increase, the program managers might be blamed. Better understanding of the underlying climate baseline is critical to evaluating the effectiveness of different interventions for malaria.

Figure 2 shows a drought index called the ‰ÛÏWeighted Anomaly Standardized Precipitation (WASP)‰Û index for Tanzania. This index uses the remote sensing product Merged Analysis Product (CMAP) developed by the Climate Prediction Center, which incorporates only those stations available to the Global Telecommunication System (GTS) (left panel) and for the ENACTS blended product, which incorporates all suitable national meteorological stations (right panel). The ENACTS product improves the assessment of rainfall extremes and provides a much better estimate of the impact of the 1997/8 El Nino [6]. The improved quality of the ENACTS products over the cruder CMAP product means that they can be used at national and sub-national (including district) levels. The ENACTS products recently launched in Tanzania have been used to assess whether climate was likely to have impacted malaria transmission during the pre- and post-malaria intervention period, marked ‰ÛÏb‰Û (1995-1999) and ‰ÛÏa‰Û respectively (2000-2010). In this example, the baseline period includes a year of very high rainfall, whereas the malaria intervention period includes three major droughts.

Comparative Advantage of ENACTS Products for Climate Information and Analysis

The ENACTS framework (Box 1) and products are designed to dramatically increase the availability, access and use of climate information for decision-making.

The ENACTS products made available by TMA are unique and leverage cutting-edge technologies to improve availability, access and use of climate information for Tanzania. This effort has focused on the creation of reliable climate information that is suitable for national and local decision-making. Data availability is improved by blending data from the national observation network and satellite and other proxies such as elevation. The main advantage of the proxies is good spatial coverage: satellite data are available over most parts of the world at increasingly improved spatial and temporal resolutions. Satellite rainfall estimates also now go back more than 30 years. Combining ground-based observations with satellite and other proxies helps to overcome the spatial and temporal gaps in station data. Additional information on the methodology behind these tools is outlined below.

Pioneering New Approaches to Improve Availability

The ENACTS products merge observational data with existing and openly available products using new techniques. For rainfall, estimates from the TAMSAT (Tropical Applications of Meteorology using Satellite data and ground-based observations) from the University of Reading have been used. The TAMSAT rainfall product [7, 8], derived from only thermal infrared satellite data, provides temporally consistent rainfall estimates going back to 1983. The data is free and is openly available through the University of Reading’s website. åÊAs for temperature, there is no reliable satellite data going back 30 years. As a result, the Moderate Resolution Imaging Radiometer (MODIS) land-surface temperature estimates and digital elevation model (DEM) are used for merging with station minimum and maximum temperature measurements. MODIS land surface temperature (LST) estimates are available at a spatial resolution of 1 kilometer, starting from 2002. This is not a long-enough period for many climatological analyses. However, the average of the 10-year data can be used as climatological background to interpolate station temperature measurements. Thus, the average of MODIS LST from 2003 to 2011 is used for merging with station observations and elevation data.

After exploring different approaches, Regression Kriging [9, 10, 11] was selected for merging station data with satellite and other proxies. Regression gridding is a two-step process. It models the value of a variable at a desired location as the sum of the deterministic and stochastic components. The deterministic component is obtained through regression on an auxiliary variable and the stochastic components are interpolated residuals. The main advantage of this method is that it could be extended to a broader range of regression techniques and allows separate interpretation of the two components.

These enhanced national climate time series overcome traditional barriers in data quality and availability. The spatially and temporally continuous datasets allow for characterization of climate risks at a local scale and offer a low-cost, high impact opportunity with major potential to support climate resilient development. There are two different products for the rainfall data.

The first merged time series includes a few operational stations and is used mainly for monitoring purposes. Operational stations are those stations that report daily. This product is updated every 10 days. The data from the other stations are received at TMA headquarters much later after observation. The second version of the merged product uses all available stations. This is the standard product and is updated about once a year. The merged temperature uses all stations that report temperature, but these are much fewer than the rainfall stations.

Figure 3 and Figure 4 present sample products for rainfall and temperature, respectively. The left panel in Figure 3 represents station measurements for a 10-day period. The middle and right panels are satellite estimate and combined gauge-satellite products, respectively. The station data is assumed to represent the ‰ÛÏtrue‰Û rainfall, but there are no stations over many parts of the country, particularly for the operational product. The satellite product covers the entire country, but tends to underestimate rainfall amounts over most parts of the country. The merged products overcome the lack of station coverage as well as the underestimation by the satellite product. This is accomplished by combining the spatial information from the satellite estimates with the point measurements at station locations. There is a significant difference between the two versions of the merged rainfall products.åÊ However, even the product with operational stations is a significant improvement over the satellite estimate.

The same is true for temperature shown in Figure 4.åÊ Here, elevation is shown instead of satellite data.åÊ A comparison of the elevation map with station measurements and the merged product shows the influence of elevation on the spatial variation of temperature in Tanzania.

Investing in Access and Use

Improved data availability, however, may not necessarily lead to improved data access and use. Dedicated efforts also need to be made to improve access to the data and its operational utility. Access to information through the ENACTS products are provided through virtual ‰ÛÏmap rooms‰Û. TMA’s online tool currently includes three map rooms for: Climate Analysis, Climate Monitoring, and Climate Forecast (Figure 5). The Climate Analysis Map Room provides information on the mean climate (in terms of rainfall and temperatures) at any point and at national and sub-national levels defined by administrative boundaries. It can also be used to explore the performance of the rainfall for a specific season over the years as compared to the mean. The Climate Monitoring Map Room enables monitoring of the current season in terms of rainfall. Different maps and graphs compare the latest 10-day period with the mean or values for recent years. This information also can be extracted at any point or for any administrative boundary. Data is updated every 10 days, thus enabling close monitoring of the season. Extracting and presenting information at any administrative level enables focusing on specific areas of interest. The Climate Forecast Map Room translates TMA’s seasonal rainfall forecasts to values that can easily be understood by users. It presents the forecasts in the context of historical rainfall data.

Improving the use of climate information involves an iterative process of engaging with a wide range of stakeholders and users. This includes involving users directly in the development of information products and training policymakers, researchers and practitioners on how to best use the derived information products and tools.åÊ The workshop organized by TMA in 2013 was the beginning of the engagement process with the health community in Tanzania. The primary objectives of the workshop were to showcase TMA’s new data and information products, demonstrate examples of how they can be used for disease stratification, improved early warning systems and impact assessments, as well as to solicit critical feedback from the health community on their needs for climate, environmental and epidemiological information, in particular for use in malaria decision-making. The workshop participants offered useful recommendations on how to improve existing information products and develop new ones.

Summary and Next Steps

TMA, in collaboration with partners including the IRI, has made significant progress in enhancing its national climate services. These enhancements include a more-than-30 yearåÊ time series of spatially and temporally complete climate data, online facilities to make information products available to users and dedicated engagements with stakeholders. The first outreach to users included the public health community. Further investments are still needed to respond to needs articulated by public health practitioners and researchers invited to initial stakeholder consultations and to build capacity to meet demands across additional sectors moving forward. TMA will need to continue to demonstrate its commitment to iteratively improving its service in the provision of climate data and products by pioneering and ensuring uptake of relevant, accurate and timely information that can readily be integrated into decision-support tools for improved resilience.

Acknowledgements:

This research was funded through grant support from NASA SERVIRåÊandåÊthe World Meteorological Organization (WMO),åÊas well as support from the U.S. Agency for International Development (USAID)åÊthrough the President’s Malaria Initiative (PMI). Support to the ENACTS developmentåÊhas also been provided byåÊåÊthe CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS). The authors would like to acknowledge the contribution of Dr. Pietro Ceccato as principal investigator of the NASA funded project.åÊThe authors also acknowledge Rene Salgado, Achuyt Bhattarai, Christie Hershey and Carrie Nielsen of PMI for helpful discussions on the needs and uses of climate and malaria data in conducting evaluations of the impact of malaria control interventions.åÊThe authors also thank the countless partners that have contributed to this work through partnership and leveraged resources.

References

1. A Gap Analysis for the Implementation of the Global Climate Observing System Programme in Africa. Palisades, NY, International Research Institute for Climate and Society, 2006.
2. WHO/UNICEF (2013). “The World Malaria Report 2012.”
3. T. Dinku, K. Asefa, K. Hailemariam, D. Grimes, and S. J. Connor, ‰ÛÏImproving Availability, Access and Use of Climate Information,‰Û WMO Bulletin, vol. 60, no 2, pp 80-86, 2011.
4. T. Dinku, K. Hailemariam, R. Maidement, E. Tarnavsky, and S. J. Connor, ‰ÛÏCombined Use of Satellite Estimates and Raingauge Observations to Produce High-Quality Historical Rainfall Time Series Over Ethiopia,‰ÛåÊ International Journal of Climatology, DOI: 10.1002/joc.3855, 2013.
5. B. Lyon, “The strength of El Ni̱o and the spatial extent of tropical drought.” Geophysical Research Letters 31(L21204), 2004.
6. M. C. Thomson, F. Zadravecz, B. Lyon, G. Mantilla, D. Willis, P. Ceccato and T. Dinku, ‰ÛÏDevelopment of Climate Analysis Section for the President’s Malaria Initiative Impact Evaluation: Reports for Ethiopia and Tanzania,‰Û President’s Malaria Initiative-USAID Report, IRI, 2012.
7. D. F. Grimes, E. Pardo and R. Bonifacio, R,åÊ ‰ÛÏOptimal areal rainfall estimation using raingauges and satellite data, ‰Û J. Hydrology, vol. 222, pp 93-108, 1999.
8. V. Thorne , P. Coakeley, D. Grimes, and G. Dugdale, ‰ÛÏComparison of TAMSAT and CPC rainfall estimates with raingauges, for southern Africa,‰Û Int. J. Remote Sensing, vol. 22, no. 10, pp 1951-1974, 2001.
9. I. Odeh, A. McBratney , D. Chittleborough, ‰ÛÏFurther results on prediction of soil properties from terrain attributes: heterotopic cokriging and regression-kriging,‰Û Geoderma, vol. 67, pp 215‰ÛÒ226, 1995.
10. T. Hengl,åÊ G. M. B. Heuvelink, and A. Stein, ‰ÛÏA generic framework for spatial prediction of soil variables based on regressionkriging,‰Û Geoderma, vol. 122, pp 75‰ÛÒ93, 2004.
11. T. Hengl,åÊ G. M. B. Heuvelink, D. G. Rossiter, ‰ÛÏAbout regression-kriging: From equations to case studies,‰Û Computers & Geosciences, vol. 33, pp 1301‰ÛÒ1315, 2007.