Evaluation of the Performance of ENACTS MAP-ROOM Products over Tanzania

Tanzania has inadequate weather stations (28-synoptic weather stations), which are sparsely distributed over complex topographic terrain. Many places, especially rural areas, have no stations to monitor weather and climate. In this study, we evaluate the performance of ENACT-MAPROOM products over Tanzania with the aim of assessing their potential to supplement observed weather and climate data, especially over areas where there is limited number of weather stations. Monthly rainfall total and monthly averaged minimum and maximum temperatures from ENACT-MAPROOM are evaluated against observed data from 23 weather stations. The evaluation is limited to analyze how well the ENACT-MAPROOM products reproduce climatological trends, annual cycles and inter-annual variability of rainfall, minimum and maximum temperatures. Statistical analysis recommended by the World Meteorological Organization (WMO) that includes that correlation and trend analysis are used. It is found that ENACT-MAPROOM products reproduce the climatological trends, annual cycles and inter-annual variability of rainfall, minimum and maximum temperatures over most stations. The statistical relationship between ENACT-MAPROOM products against observed data from 23 weather stations using Pearson correlation coefficient indicates that ENACT-MAPROOM products bear strong and statistically significant correlation coefficient to observed data. The overall evaluation here finds high skills of ENACT-MAPROOM products in representing rainfall and temperature over Tanzania, suggesting their potential use in planning and decision making especially over areas with limited number of weather stations.


Introduction
Climate change and variability depends on long-term observational climate datasets [1]. These datasets, especially in developing countries are observed by ground based weather stations. However, the number of ground based weather stations in developing countries are not adequate to provide climate dataset at high temporal and spatial resolution [2] [3]. For instance, Tanzania has few weather stations (28-synoptic weather stations) which are sparsely distributed over complex topographical terrain [4]. Many places, especially in rural areas where farming practices are predominant, have no weather stations to monitor climate change and variability. In order to overcome some of the challenges of data from ground based weather stations (large data gaps, sparse stations networks), satellite based gridded climate data are generated at high temporal and spatial resolutions.
The gridded climate data are now available at high temporal and spatial resolutions [5]. These include the Global Precipitation Climatology Project (GPCP at 1˚ by 1˚ spatial resolution), Tropical Rainfall Measuring Mission (TRMM, at 0.25˚ by 0.25˚ spatial resolution) and the Famine Early Warning System (FEW at 0.1˚ by 0.1˚ spatial resolution) [6]. However, the quality of gridded climate datasets depends on the number of ground weather observations used in merging and interpolation processes [7]. In developing countries, climate data are regarded as proprietary and not freely shared. This makes few numbers of weather stations that participate in the generation of global merged or interpolated climate products, and make them of inadequate quality to support adaptation decision in developing countries.
The Tanzania Meteorological Agency (TMA) has been working with the International Research Institute for Climate and Society (IRI) through Enhancing National Climate Services (ENACTS) initiative to produce country specific gridded climate data library (ENACTS MAPROOM). These merged climate datasets use all observed climate records available in the country and are designed to bring climate knowledge into national decision making by improving availability, access to, and use of climate information [8].
The climate datasets produced by IRI through ENACTS initiative are readily available and can be accessed by different stakeholders such as planners and decision makers. However, before providing knowledge and information to promote the use of ENACTS-MAPROOM products to different stakeholders for day to day and longer term planning, it is important to evaluate their performance to reproduce historical climate over different stations. In this study, the performance of MAPROOM product to reproduce present climate condition over different regions of Tanzania is evaluated.

Study Area
Tanzania is located in East African region between latitudes 1˚S and 12˚S and

Data and Analysis
Decadal total rainfall, and decadal mean of minimum and maximum temperatures for the period of 1983 to 2010 from 23 weather stations available from the CLIDATA archive at the Tanzania Meteorological Agency (TMA) were used to merge with decal total rainfall, and decadal mean of minimum and maximum temperatures for the period of 1983 to 2010 from satellite observations. The merged rainfall and temperatures are then interpolated across space and time into balanced panel of observations on fixed spatial scale or grid of 4 km by 4 km spatial resolution. Figures 1-3 respectively indicate spatial distribution of decadal total rainfall and minimum and maximum temperatures averaged over time from 1983 to 2010. These merged and interpolated rainfall and temperature outputs are known as MAP-ROOM products. The MAP-ROOM products gridded at 4 km by 4 km spatial resolutions were evaluated against actual measurement from weather stations using the interpolation method where the gridded MAP-ROOM products were interpolated to the location of weather station and the results were compared with observed station data. There are different interpolation techniques, the simplest is the nearest neighbor interpolation method that assume climate or weather variable at a given weather station is equal to that at the closest grid point [11] [12]. In this study nearest neighbor interpolation technique is used to interpolate MAP-ROOM products to the location of weather stations.
There are several criteria to evaluate the performance of MAP-ROOM products to represent observed weather or climate data [13]. However, there is no   individual evaluation technique or performance is considered superior, rather, it is combination of many techniques or measures that provide comprehensive evaluation. In this study, MAP-ROOM products are evaluated against observation data using a statistical measure recommended by the World Meteorological Organization (WMO) as reported by [14]. This statistic is the Pearson correlation coefficient defined as:

Annual Cycle
The performance of MAP-ROOM products to reproduce annual cycle of rainfall and temperature is presented in Figure 4.   In the bimodal regions Figure 4(a) observed data exhibit two peaks of rainfall, a primary maximum in April and a secondary one in November and low minimum in June, July and August. This pattern is reproduced by MAP ROOM rainfall estimates, although they general overestimate rainfall from April to July.
In the unimodal regions Figure 4

Conclusion
The study evaluated the performance of MAP ROOM products to reproduce annual cycles and inter-annual time series. The key objective was to assess if the MAP ROOM products can supplement observation in areas with limited number of weather stations. It is found that MAP ROOM products represent the an-