96827 Renewable Energy Resource Mapping. Solar Tanzania. Interim Solar Modeling Report February, 2015 Project: Renewable Energy Resource Mapping. Solar Tanzania. Code: 30.2495.0 Client: The World Bank Contact person: Anders Pedersen, Oliver Knight Address: 1818 H Street, NW. USA. Washington, DC 20433 Sending date: 13 February 2015 Revision history REPORT VERSION DATE SUBMITTED DATE COMMENTED Interim Solar Modeling Ver_1 February 2015 Report Elaborated by: Reviewed by: Approved by: Carlos Fernández-Peruchena, Martín Gastón, Ph.D. , KE2 – Ana Bernardos, Bsc. KE1 Ph. D. Solar Resource Solar Resource Maps Modeling Project Manager and Solar Assessment and Data Treatment, and Geo spatial treatment, Resource Assessment Expert. CENER. CENER CENER Luis Martín Ph.D. KE4 – Solar Resource Maps Modeling and Geo spatial treatment, IRSOLAV Diego Bermejo, Ph.D. Solar Resource Assessment and Data Treatment, IRSOLAV José María Vindel, DEA. Solar Resource Maps. Modeling, CIEMAT Lourdes Ramírez, Ph.D. KE3 Solar Resource Maps Modeling Expert. CIEMAT Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Table of contents 1 BACKGROUND .................................................................................................................................................... 4 2 INTRODUCTION .................................................................................................................................................. 4 2.1 CLIMATOLOGICAL DESCRIPTION OF TANZANIA .............................................................................................................. 5 3 PREVIOUS SOLAR RESOURCE ASSESSMENT AND MEASUREMENTS .................................................................... 6 3.1 SOLAR RESOURCE ASSESSMENT STUDIES ...................................................................................................................... 7 3.1.1 Related literature .......................................................................................................................................... 7 3.1.2 PVGIS ............................................................................................................................................................. 8 3.1.3 NREL .............................................................................................................................................................. 9 3.1.4 SSE NASA ....................................................................................................................................................... 9 3.2 SOLAR RADIATION MEASUREMENTS ......................................................................................................................... 10 3.2.1 Tanzania Meteorological Agency ................................................................................................................ 10 3.2.2 World Radiation Data Centre ...................................................................................................................... 11 4 METHODOLOGY OF SOLAR RESOURCE MAPS GENERATION ............................................................................12 4.1 SATELLITE-BASED SOLAR IRRADIANCE MODELLING ....................................................................................................... 13 4.2 NWP SOLAR IRRADIANCE MODELLING ...................................................................................................................... 17 4.3 ANALYSIS OF COHERENCE AND SOURCES COMBINATION ............................................................................................... 18 4.3.1 Analysis of input data ................................................................................................................................. 19 4.3.2 Coherence assessment ................................................................................................................................ 19 4.3.3 Combination and creation of the model ..................................................................................................... 23 4.4 SOLAR RESOURCE ASSESSMENT FOR MAP GENERATION................................................................................................. 26 4.4.1 Generation of GHI maps.............................................................................................................................. 26 4.4.2 Generation of DNI maps ............................................................................................................................. 26 4.4.3 Generation of DHI maps.............................................................................................................................. 26 4.4.4 Generation of GTI maps .............................................................................................................................. 26 4.4.5 Generation of PV potential maps ................................................................................................................ 28 5 SOLAR RESOURCE MAPS ...................................................................................................................................29 5.1 GLOBAL HORIZONTAL IRRADIANCE........................................................................................................................... 29 5.2 DIRECT NORMAL IRRADIANCE ................................................................................................................................. 31 5.3 DIFFUSE HORIZONTAL IRRADIANCE .......................................................................................................................... 33 6 CONCLUSIONS ...................................................................................................................................................35 7 REFERENCES ......................................................................................................................................................36 ANNEX 1. MONTHLY GLOBAL HORIZONTAL IRRADIANCE:.....................................................................................39 ANNEX 2. MONTHLY DIRECT NORMAL IRRADIANCE: ............................................................................................41 ANNEX 3. MONTHLY DIFFUSE HORIZONTAL IRRADIANCE: ....................................................................................44 Page 3 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 1 BACKGROUND This interim solar modelling report provides an overview of the results achieved by modelling the solar radiation based on satellite data and Numerical Weather Prediction Models (NWPM) in Phase 1 of the Solar Resource Mapping Project for Tanzania (WB Selection #1139235). The project comprises three Phases. The first phase comprises Project inception, preliminary modelling and implementation planning. Within this Phase 1, an un-validated solar atlas based on the synergistic combination of satellite and NWPM derived solar data for Tanzania has been carried out. The interim output of the solar atlas is presented in this report. Within Phase 2, ground-based data collection will be undertaken through a measurement campaign at sites selected from areas defined according to the results of Phase 1. Finally, in Phase 3, a resource atlas with reduced DNI/GHI/DIF uncertainty with respect to in Phase 1 will be generated from post-processing satellite and NWPM solar radiation outputs with the validated ground-based solar data collected during the measurement campaign of Phase 2. 2 INTRODUCTION A preliminary analysis of solar resource and its variability in Tanzania is the focus of this report. Section 3 contains a review of previous solar resource assessment and measurements in Tanzania. Section 4 presents the methodology for the solar resource assessment; using analysis of the satellite and NWPM derived solar data coherence and combination. In section 5 solar maps are presented and briefly discussed. Finally, a conclusions section outlines the current achievements and the future needs in the Solar Resource Mapping Project for Tanzania. The deployment of solar energy projects in a country needs, in a first step, precise information on the available solar resource. The solar resource information facilitates decision-making processes of the different technologies to be used, as well as appropriate policies and investments. The proper sitting of a renewable energy system is critical to its success. However, sitting a solar energy system can be particularly challenging because of the varying nature of the solar radiation. Weather fluctuations and seasonal sun position changes can have significant effects on a system's performance. The characterization of the solar resource must include its magnitude, how much solar energy is available at an area of interest over a long time period, and its variability over time. Knowledge of this variability is important for improving the design of a system and understanding the performance of a solar conversion system. It is also important to consider variability of the solar resource in space (how it varies over distance) to characterize the regional distribution of solar energy. Knowledge of the solar resource and its variability in both time and space provides critical information for determining where to conduct a measurement campaign in order to characterize the solar resource of a region. Page 4 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Solar radiation can be transmitted, absorbed, or scattered by the atmosphere in varying amounts depending on the wavelength, resulting in the following fundamental components of importance to solar energy conversion technologies: Direct Normal Irradiance (DNI) (W/m2): is the amount of solar radiation (power) received per unit area on a surface perpendicular (or normal) to solar radiation that comes directly from sun´s disc without any scattering by the atmosphere. According to WMO (World Meteorological Organization) it is measured in normal direction to the sun with a pyrheliomenter designed with about 5 degree field of view (FOV) full angle. Global Horizontal Irradiance (W/m2): is the amount of solar radiation (power) received per unit of area from a solid angle of 2π sr on a horizontal surface. This includes radiation received directly from the solid angle of sun´s disc as well as diffuse sky radiation that has been scattered in traversing the atmosphere. It is measured on the surface with a pyranometer. Diffuse Horizontal Irradiance (DHI) (W/m2): is the amount of solar radiation (power) received per unit area on a horizontal surface, after its direction has been changed by scattering by the atmosphere. It is measured on the surface with a pyranometer with a shadow device in order to avoid the direct irradiance. The three solar-irradiance components are related. On any surface, direct plus diffuse irradiance equals global irradiance. For a horizontal surface, DNI can be converted to direct horizontal using the solar- zenith angle (θ) at the time of interest: = ∙ cos + Different solar applications require different components of solar irradiance data. For example, concentrating collectors require accurate DNI estimations, while flat plate collectors require global tilt irradiance (GTI) which is derived from DNI, DHI, GHI and ground albedo. DHI and DNI are also useful for daylight applications and cooling load calculations in energy efficient buildings. Therefore, it is important to estimate all the three irradiance components i.e. GHI, DNI and DHI to make the database suitable for a wide range of solar applications. 2.1 Climatological description of Tanzania Except for a narrow coastal strip, Tanzania is dominated by highlands. The greater part of Tanzania is a central plateau ~ 900-1800 m above sea level. Tanzania has a tropical climate with regional variations due to topography; the coastal regions of Tanzania are warm and humid while the highland regions more temperate. A relatively narrow belt of heavy precipitation and very low pressure that forms near the earth’s equator (known as the Inter‐Tropical Convergence Zone, ITCZ) is the main driver of seasonal rainfall in Tanzania. The ITCZ migrates southwards through Tanzania in October to December, and reaches the south of the country in January and February, returning northwards in March, April and May. As a consequence, two wet periods are defined in the northern and eastern regions of Tanzania: the “short rains” (also known as “Vuli”) in October-December and the “long rains” (also known as “Masika”) in March-May. The central, western and southern regions of Tanzania experience one wet season that Page 5 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report ranges from October to April or May (McSweeney, New, & Lizcano, 2012). Fig 1 shows climate zones in Tanzania according to Köppen classification scheme (Peel, Finlayson, & McMahon, 2007). Fig 1 Climate zones in Tanzania. Group A zones, covering most of the country, comprises tropical climates. Tropical Savannah (Aw) and Monsoon climates (Am) both are characterized by monthly mean temperatures above 18 °C, with the Savannah climate experiencing less rainfall or more pronounced dry seasons than the Monsoon climate. Group B zones, extending diagonally from the South East to the North-West of Tanzania, represent Dry (arid and semiarid) climates. In this region, the most extensive climate is BSh (hot semi-arid), and in some isolated regions BWh (hot desert). Group C climate zones are present along the South-East border of the country and also in a small region in the North of Tanzania. This group comprises temperate climates, most commonly Cwb, a subtropical highland variety of the oceanic climate (typical of mountainous locations in some tropical countries). 3 PREVIOUS SOLAR RESOURCE ASSESSMENT AND MEASUREMENTS Solar radiation incoming on the Earth’s surface exhibits a large geographical and temporal variability due to its strong dependence on the atmospheric conditions and meteorology. The ideal way to analyse solar resource is by using ground measurements from meteorological stations; however this information presents the disadvantage of referring to a particular point and the usual scarcity of this type of data in many parts of the world. Besides the ground data, solar resource is usually modelled by satellite-based or NWPM. Meteorological stations are good at providing high frequency and accurate data (for well- maintained, high accuracy measuring equipment) for a given site. On the other hand, models provide Page 6 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report data with a lower frequency of measurement, but characterizing a long history over wide territories. Satellite or NWPM based solar data are not capable of reproducing instantaneous values with the same accuracy as ground sensors, but can provide robust aggregated values. In this report, a review of the solar radiation measurements and models in Tanzania is presented, as well as several works that have focused on the analysis of solar resource in the region of Tanzania. 3.1 Solar resource assessment studies 3.1.1 Related literature Alfayo and Uiso (Alfayo & Uiso, 2002) describe a prospect on GHI distribution in Tanzania. An empirical model based on meteorological data collected between 1965 and 1990 in Tanzania was developed to estimate GHI on horizontal surfaces. Meteorological parameters such as sunshine hours, relative humidity, air temperature and atmospheric conditions were used in the model. Measured and predicted mean monthly and mean annual GHI values as observed from the developed radiation maps and graphs indicate that Tanzania has high solar power potential. The lowest annual average radiation value in the country according to this work is 1521 kWhm2/y, while the maximum value is 2433 kWhm2/y. Besides, this work states that ~90% of the country has a high level of solar irradiation, ranging from 1825 to 2216 kWhm2/y. Fig 2 shows mean annual distribution of GHI in Tanzania, according to this work. -2 -1 Fig 2 . Mean Annual distribution of GHI (MJm d ) in Tanzania. John and Mkumbwa (John & Mkumbwa, 2011) uses solar insolation data to evaluate the economic viability of PV technology in Kondoa district. The amount of solar insolation influences the size of solar module and hence its cost (the larger the amount of solar insolation, the smaller the size of solar module to be used, and hence the lower the cost). Mean monthly GHI values corresponding to the period 1998-2008 are shown in Fig 3. Page 7 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 3 Kondoa mean monthly GHI (1998-2008) 3.1.2 PVGIS The Photovoltaic Geographical Information System (PVGIS) is a web-based knowledge distribution system that provides climate data and tools needed for the performance assessment of photovoltaic (PV) technology in Europe, Africa, and South-West Asia (Suri, Huld, Dunlop, Albuisson, & Wald, 2006). It is a part of the SOLAREC project that contributes to the implementation of renewable energy in the European Union as a sustainable and long-term energy supply by undertaking new technological developments in fields where harmonization of global PV standards is required and requested by customers. Fig 4 shows the annual average of daily GHI values for Africa in Wh/m2 (left), and the detailed map for the region of Tanzania (right). Fig 4 Left, annual average of daily GHI values provided by PVGIS (1984-2005); right, detailed map of Tanzania region provided by PVGIS Page 8 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 3.1.3 NREL NREL (National Renewable Energy Laboratory) develops maps for various renewable resources. The Climatological Solar Radiation (CSR) Model (George & Maxwell, 1999) uses information on cloud cover, atmospheric water vapour and trace gases, and the amount of aerosols in the atmosphere, to calculate the monthly average daily total insolation falling on a horizontal surface. The cloud cover data used as an input to the CSR model is an 8-year histogram (1985 - 1992) of monthly average cloud fraction provided for grid cells of approximately 40km x 40km in size. The data was obtained from the National Climatic Data Center in Asheville, North Carolina, and was developed from the U.S. Air Force Real Time Nephanalysis (RTNEPH) program. The modelled values are estimated to be accurate to approximately 10% of a true measured value within the grid cell. Due to terrain effects and other microclimate influences, the local cloud cover can vary significantly even within a single grid cell. Fig 5 shows annual average of daily DNI values provided by NREL (left), and the corresponding detailed map of Tanzania region (right). Fig 5 Left, annual average of daily DNI values provided by NREL; right, detailed map of Tanzania region provided by NREL (within the frame of SWERA Project) 3.1.4 SSE NASA Surface meteorology and Solar Energy (SSE) data were obtained from the NASA Science Mission Directorate's satellite and re-analysis research programs of meteorological and satellite data from 1983 to 2005. The SSE estimation of solar irradiation is based on satellite derived electromagnetic energy incident on earth surface which has been reanalyzed together with meteorological data by NASA to produce a global estimation. The analysis of the SSE NASA solar resource interpolated data in conjunction to ground measurements in Tanzania has been carried out by Hammar (Hammar, 2011), the results of which are presented in this section. According to this work, the SSE NASA solar resource estimation may be a useful indicator of insolation in the region, and it indicates substantial differences in insolation within the region, with highest values in a broad stretch throughout central Tanzania. The SSE NASA solar data also indicates a higher insolation along the northern coast. However, this pattern is not evidently in other assessments. In terms of seasonal variations, SSE NASA indicates that central-western Tanzania gains Page 9 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report substantially more insolation during May to September, while most other areas of the region peak during November–March. Seasonal variations are also high in western Tanzania. 3.2 Solar radiation measurements 3.2.1 Tanzania Meteorological Agency The Tanzania Meteorological Agency (TMA) is a national Government Agency established in December 1999 (http://www.meteo.go.tz/). Prior to this, meteorological services were provided by the Directorate of Meteorology (from 1978) under the Tanzanian Government, operated by the East Africa (E.A) Meteorological Department of the former E.A. Community. TMA organize and manage surface and upper air observation networks and record the climate conditions of the United Republic of Tanzania. TMA also collects, processes, stores and disseminates meteorological information and forecasts. Fig 6 shows TMA network stations. Fig 6 . TMA network stations (Osima, 2014). The available instrumental meteorological observations in Tanzania measure all common parameters including air temperature, precipitation, air pressure, surface radiation budget, wind speed and direction and water vapour. Synoptic stations make hourly observations of various elements including: wind direction and speed, solar radiation, relative humidity of the air, wind run, and cloud type. Agro- meteorological stations make meteorological observations twice daily, at 06.00 hours and at 12.00 hours UTC. Observations include: daily rainfall, wet bulb and dry bulb temperatures, maximum and minimum temperatures, hours of bright sunshine, relative air humidity, class-A pan evaporation, soil water, soil temperature and crop growth and development. Climatological stations measure rainfall, temperature, humidity and sunshine. Currently, TMA is involved in a plan for the improvement of Meteorological Services in Tanzania in order to improve Meteorological Infrastructure, capacity building, data processing and archiving (Table 1). Page 10 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Table 1 . TMA observation network current and required stations. Number of stations Description Current Needed Conventional surface synoptic stations 28 32 AWS Surface synoptic stations 15 80 Agrometeorological stations 14 35 Ordinary climate stations 130 250 Rainfall stations 1549 2500 Weather radar 1 7 In the current phase of the project (preliminary modelling), monthly GHI values measured by TMA have been used, where possible by extraction from technical reports or research articles: Kilimanjaro, Dodoma, Morogoro, Dar es Salaam and Kondoa (see section 4.3.1 for more details). Unfortunately, no hourly radiation data has been accessible and thus considered in this phase of the solar resource assessment in Tanzania. 3.2.2 World Radiation Data Centre The World Radiation Data Centre, WRDC (http://wrdc.mgo.rssi.ru/), is a recognized World Data Center sponsored by the World Meteorological Organization (WMO). The WRDC centrally collects and archives radiometric data from around the world to ensure the availability of these data for research by the international scientific community. The WRDC is a laboratory of the Voeikov Main Geophysical Observatory, the Russian Federal Service for Hydrometeorology and Environmental Monitoring (formerly USSR State Committee for Hydrometeorology), and is located in St. Petersburg. The WRDC regularly issues the publication "Solar Radiation and Radiation Balance Data (The World Network)" with the purpose of providing the users with data on solar radiation, radiation balance and sunshine duration in convenient and readily accessible form. The WRDC began as the World Radiation Data Centre in 1964 and produced its first data publication of Solar Radiation and Radiation Balance Data (The World Network) in 1965. In Tanzania, WRDC provides solar radiation from 1964 to 1977, measured with Bellani pyranometers. The list of stations is shown in Table 2, and their GHI monthly mean values are represented in Fig 7. Page 11 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Table 2 WRDC stations in Tanzania. Elevation Period Station WMO index Latitude Longitude (m) (years) Bukoba 63729 1°20'S 31°49'E 1137 1964-1976 Musoma 63733 1°30'S 33°48'E 1147 1970-1973 Mwanza 63756 2°28'S 32°55'E 1139 1965-1976 Arusha 63789 3°20'S 36°37'E 1387 1973-1976 Kilimanjaro Arpt 63791 3°25'S 37°04'E 891 1971-1976 Same 63816 4°05'S 37°43'E 872 1969-1976 Kigoma 63801 4°53'S 29°38'E 882 1970-1976 Tabora Arpt 63832 5°05'S 32°50'E 1181 1965-1975 Dodoma 63862 6°10'S 35°46'E 1119 1965-1976 Zanzibar / Kisauni 63870 6°13'S 39°13'E 15 1964-1974 Morogoro 63866 6°50'S 37°39'E 526 1970-1974 Dar es Salaam Arpt 63894 6°53'S 39°12'E 55 1964-1976 Iringa 63887 7°40'S 35°45'E 1426 1964-1977 Mtwara 63971 10°16'S 40°11'E 113 1969-1976 Songea 63962 10°41'S 35°35'E 1067 1968-1976 Fig 7 . Monthly means values of GHI measured in WRDC stations in Tanzania. 4 METHODOLOGY OF SOLAR RESOURCE MAPS GENERATION As a meteorological variable measured directly on the ground, solar radiation is measured in relatively few ground stations and for short and, in most cases, discontinuous periods of time. It is very common to find, for a specific site of interest, lack of reliable information on solar radiation. Moreover when seeking coverage of a given region as opposed to a specific site, the poor spatial variability provided by historical databases from ground measurements means that the projection or assessment of solar power systems becomes a seriously difficult task. In fact, when it comes to characterizing the solar resource of a given specific site, the following considerations must be taken into account: Page 12 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report • Data from nearby stations can be used in areas of relatively flat terrain and when distances are less than 10 km from the site. In the case of complex terrain or greater distances the use of radiation data from other geographical points is not possible. • Interpolation of surrounding measurements is possible only for areas with a high density of stations and for average distances between stations of about 20-50 km (Pérez, Seals, & Zelenka, 1997; Zelenka, Perez, Seals, & Renne, 1999). In this context, modelled solar radiation data can provide robust aggregated values in the solar resource assessment, and a long history over wide territories. The most widely used methodologies for modelling radiation are satellite-based solar irradiance modelling and Numerical Weather Prediction Models (NWPM): • Satellite-based solar irradiance modelling is based mainly on information relating to cloud cover characteristics. • NWPM reproduces the behaviour and evolution of the atmosphere, taking into account meteorological information about the attenuating components of the atmosphere which affects solar radiation The synergistic combination of both methodologies offers the advantage of including complementary information, combining solar radiation estimated from the cloud images and the atmospheric dynamic from the meteorological modelling point of view. This combination is currently the most suitable approach in solar resource assessment (Gastón, Pagola, Fernández-Peruchena, & Blanco, 2012). This section describes the inputs used in the solar resource assessment (section 4.1 and section 4.2), as well as the analysis of their coherence and their combination to derive GHI monthly data (section 4.3). Finally, methodologies used in the solar resource assessment for map generation are presented (section 4.4). 4.1 Satellite-based solar irradiance modelling Satellite based solar estimation provides information on the spatial distribution of solar radiation; its great advantage is that a long temporal range that can be achieved for a single site. The methodologies developed for this purpose are widely accepted by the scientific community (McArthur, 1998). The methodology followed for satellite-based solar irradiance modelling is described below. This methodology is based on the concept of cloud index (n) as used to explain solar radiation variability, the concept being used in numerous peer-reviewed publications (K. F. Dagestad & Olseth, 2007; Mueller et al., 2004; Rigollier, Lefèvre, & Wald, 2004; Schillings, Meyer, & Mannstein, 2004; Suri, Huld, & Dunlop, 2005): − = − where is the reflectivity observed by the satellite sensor, and and are the reflectivities corresponding to clear sky and overcast conditions (ground and cloud albedos), respectively. For the Page 13 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Heliosat-3 model (K.F. Dagestad, 2005), we have developed our own methodology, below (Polo, Zarzalejo, Martin, Navarro, & Marchante, 2009; Zarzalejo, Polo, Martín, Ramírez, & Espinar, 2009). In order to obtain the cloud index, our methodology takes into account the fact that the reflectivity detected by the satellite sensor depends upon the so-called co-scattering angle, since part of the irradiation measured by the satellite comes from backscattering within the layers of the atmosphere. Ground albedo is estimated from a multil-day mobile window; in this study, the window used covers 20 days. Within this mobile window, the 4th percentile of the ground albedo distribution is estimated and fitted to a polynomial expression in order to find the relationship between ground albedo and co- scattering angle [Polo et al., 2012, 2013]. Separately, cloud albedo is estimated by using a set of polynomial expressions that governs the relationship between maximum reflectivity (as determined by the 95th percentile) and co-scattering angle for each month and for each satellite/sensor. Finally, global irradiance is estimated by means of the empirical relationship between the clear sky index and the cloud index (Rigollier & Wald, 1998). < −0.2 , = 1.2 −0.2 ≤ < 0.8 , =1− $ 0.8 ≤ < 1.1 , = 2.0667 − 3.6667 · + 1.6667 · ≥ 1.1 , = 0.05 = where GHI is the global horizontal irradiation and GHICS the global horizontal irradiation for clear sky conditions. The GHICS is determined using the REST2 model (Gueymard, 2008). REST2 (Gueymard, 2010) is a two-broadband (visible and infrared) model that has been well parameterized and validated (2% RMSE), and has the following properties: • REST2 accounts for radiation extinction: scattering and absorption • As inputs, REST2 uses: pressure, ground albedo, aerosol optical depth, ozone and NO2 vertical pathlengths, aerosol single-scattering albedo, Angstrom’s turbidity coefficient, precipitable water, etc. • As outputs, REST2 provides: GHI, DNI, DHI. Direct normal irradiation is estimated by using the DirInt model (Perez, Ineichen, Maxwell, Seals, & Zelenka, 1992). In addition, a specific algorithm is applied in order to detect clear sky conditions (Polo et al., 2009), and when such conditions are met, the estimated irradiation is substituted for the clear sky value yielded by REST2. Fig 8 shows the general outline of our methodology. Page 14 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 8 .Outline of the methodology for satellite-based solar irradiance modeling. DHI is automatically derived from the estimated GHI and DNI and the cosine of the solar zenith angle (θ) for each instant of the day, through the relations between these variables (section 2). Our methodology also accounts for aerosol and water vapour information which are required by the clear sky and direct normal irradiation models. These data are ingested as daily values of aerosol optical depth (AOD) at 550nm and column of water vapour. The sources of this data used in this study are MACC and NCEP, respectively. MACC (Monitoring atmospheric composition and climate) is a project from the European Center of Medium range Weather Forecasting (ECMWF) that provides reanalysis data, in particular AOD, with a spatial resolution of 1.125 degrees for the period 2003 to 2013. For previous years to 2003, dedicated averaged values are considered. NCEP (National Centers for Environmental Prediction, from the National Oceanic and Atmospheric Administration, NOAA) provides reanalysis data and, in this case, water vapour column is taken for the whole period of satellite images used in this work. Topography is also taken into account by our methodology, which is obtained from the Digital Elevation Model (DEM) from the Shuttle Radar Topographic Mission (SRTM), with a spatial resolution up to 90m. Initially, it was proposed that satellite images were used from the Meteosat First Generation (1994-2005) and Second Generation (2005-2014) geostationary satellites, which operate at 0º Longitude (Fig 9, left). However, this study has ultimately used images obtained by the IODC/MFG geostationary satellite, located over Indian Ocean (Fig 9, right). This satellite is Meteosat First Generation technology (MFG-5 from 1999 to 2007, MFG-7 from 2007 to present), which provides visible images every half an hour with the finest spatial resolution of 5 km X 5 km. For the first period (1999-2007), the footprint of the satellite is placed at 63ºE degrees of longitude, and for the second period (2007 to present), the footprint is placed at 57ºE degrees. Page 15 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 9 . Left panel: 0º MSG image. Right panel: IODC/MFG image. Red square highlights Tanzania, Rwanda and Burundi. The selection of this satellite over the one in Europe arises from several reasons. First, it is worth highlighting that the satellite centred over Europe shows an abrupt transition from First to Second Generation technology in 2005/2006, leaving a period of gaps that is not found in the IODC/MFG technology. Second, the position of Tanzania is close to the limb in the MSG images (more so than in IODC/MFG images); hence limb distortions are expected to be less intense than in the satellite centred over Europe. Third, it has been found in MSG’s high resolution images from 0º degrees that they have been trimmed at times close to sunset causing a losing information at the end of the day over parts of Tanzania (Fig 10). Fig 10 . Sequence of MSG images showing the trims at times close to the sunset. From 1999 to 2013, both years included, satellite estimations have been performed for the region of Tanzania, with a spatial resolution of 5 x 5 km, and covering the spatial domain of 12ºS to 0º latitude, and 28ºE to 44ºE longitude. Page 16 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 4.2 NWP solar irradiance modelling Mesoscale Numerical Weather Prediction models (NWPM) consist of a set of mathematical equations that reproduce the atmosphere’s behaviour and evolution. These models are normally used to obtain meteorological forecasts in the Short/Mid Term, but they can also be used to reproduce the long term meteorology in a concrete domain. This is possible via the execution of a historical simulation of the atmosphere’s evolution. The local execution of a mesoscale NWP model allows its parameters and executions to be controlled and so to adapt it in order to generate meteorological forecasts in any domain over the world. This provides similar information in terms of resolution and reliability to those derived from satellite images. But together with the solar radiation, it includes other meteorological data such as temperature and atmospheric pressure, which allows solar radiation and its related variables to be estimated in places where limited measured terrestrial data is available. Skiron is a mesoscale NWP model based on the Eta prediction model, using input data from the Global Forecast System (GFS) (Kallos et al., 1997). It was developed for operational use in several stages by Athens University, the Helenic National Meteorological Service (HNMS) and the United States National Center of Environmental Prediction (NCEP). Skiron is run at CENER using GFS data as an input, providing hourly GHI series data with a spatial resolution of 5 x 5 km, which can be achieved both by the execution of the model at that resolution, and by means of spatial interpolation (from lower resolution model executions). The GFS data describes the global situation of the atmosphere and from them, the mesoscale model solves the atmospheric equations that drive its evolution, and thereby, the magnitudes of the different meteorological variables. The existence of historical GFS data allows us to have both the data necessary to map the variables of interest, and a virtual series at fixed points. This database is therefore usable for estimating solar resource. For Tanzania, Skiron has been executed covering a 10 year period (2003- 2013), with a geographic domain covering between latitudes -36º and 13º north and longitudes 5º and 52º east, with a spatial resolution of 5 x 5 km. This domain is shown in Fig 11 where estimations of cloud cover are shown. Page 17 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 11 Cloud cover estimated by Skiron Given the lack of DNI measurements or regional GHI to DNI conversion models, DHI estimates from Skiron derived GHI series have been derived by means of a general model applicable to the region (Ruiz- Arias, Alsamamra, Tovar-Pescador, & Pozo-Vázquez, 2010). This is a regressive model for the estimation of hourly diffuse solar irradiation under all sky conditions, and is based on the sigmoid function using clearness index ((the ratio of GHI to top-of-atmosphere solar irradiance on the same plane) and relative optical mass as its predictors. This model provides relative root mean square error values ranging 20– 35% and a relative mean bias error ranging from 5% to 12%. Additionally, this model shows some advantages compared to other evaluated models: the sigmoid behaviour of this model is able to provide physically reliable estimates for extreme values of the clearness index in spite of using less parameters than other tested models. DNI is automatically derived from the estimated GHI and DHI and the cosine of the solar zenith angle, θ, for each instant of the day, through the relations between these variables (section 2). 4.3 Analysis of coherence and sources combination In this section, the solar radiation series derived from satellite and NWPM methodologies is analyzed in order to assure their validity. Subsequently, these datasets are compared in order to assess the similarity of their solar radiation outputs, providing a comprehensive view of the accuracy of different solar modelling approaches in Tanzania. Finally, the best combination of both sources (satellite and NWPM) for reproducing solar irradiation measurements is investigated, in order to provide a synergistic combination of both methodologies. This combination offers the advantage of including complementary information, combining solar radiation as estimated from cloud images and the atmospheric dynamic from the meteorological modelling point of view. Page 18 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 4.3.1 Analysis of input data Due to magnitude of solar radiation data generated at a resolution of 0.05° from satellite data and the Skiron model, in this section we have analysed the data at a resolution of 1º so as to facilitate information management. A global window of the following coordinates has been established for the project: • Latitude: from 1º S up to 12 S. • Longitude: from 28° E up to 43° E Stringent quality control was applied on all the data available from satellite and NWPM derived solar radiation series, in accordance with BSRN quality check procedures (Ohmura et al., 1998) in order to assure the validity of the data. Regarding ground measurements, it must be stressed that there are no hourly data sets and thus preliminary validation can only be made using monthly long-term values. Table 3 presents the selected stations from the section “3 Previous Solar Resource Assessment and Measurements”, summarizing the corresponding source (1: World Radiation Data Center; 2: Alfayo and Uliso, 2002; 3: John and Mukumba, 2011), the annual values, and the ID assigned to the station for the following analysis. Table 3 Information related to the ground stations used in the coherence assessment. Station Source GHI (kWh/m2) ID station Bukoba 1 4.58 1 Musoma 1 5.34 2 Mwanza 1 5.51 3 Arusha 1 5.65 4 Kilimanjaro Arpt 2 4.56 5 Same 1 4.72 6 Kigoma 1 4.6 7 Kondoa 3 5.89 8 Tabora Arpt 1 5.65 9 Dodoma 2 5.93 10 Zanzibar / Kisauni 1 4.91 11 Morogoro 2 4.51 12 Dar es Salaam Arpt 2 4.91 13 Iringa 1 6.24 14 Mtwara 1 4.62 15 Songea 1 4.18 16 4.3.2 Coherence assessment Temporal coherence The temporal coherence of satellite and NWPM derived solar radiation series has been analyzed. In general, NWPM series tends to provide clear sky conditions more frequently than satellite, as demonstrated in Fig 12, showing a specific day’s results. In that graph it can also be verified that data synchronization between both datasets is correct. Page 19 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 2 Fig 12 Satellite (sat) and NWPM (skr) derived hourly solar radiation series for a given day (W/m ). A comparison between GHI measured data (section 3.2) and the available estimated series (satellite, section 4.1, and NWPM, section 4.2) has been carried out. In assessing the strengths of combining these data, the following limitations must be taken into account: • Estimations (satellite and NWPM methodologies) require a local fitting. • Perhaps not all measured sources are representative of long-term (some stations only provide 3 years). • The uncertainty of available measurements is unknown. • There are no details of maintenance procedures and instrumentation related to TMA stations. Fig 13 shows the correlations of satellite and NWPM solar radiation GHI monthly estimations vs. measurements, in kWh/m2d. Fig 13 Monthly GHI scatterplot between satellite (left) and NWPM (right) derived data and measurements. As a first approach for the coherence and combination assessment station by station, a linear regression from the two estimates was carried out in the form: = '( + ') *+, + '$ *-. + '/ 01 Page 20 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Where Dec is the declination that was included in order to obtain a better fitting. The obtained R2 have values at least of 0.5 (except for the stations 1 and 2) and in occasions, their value overcome 0.9. Due to the misbehaviour showed for the stations 1 and 2, they have been removed of the later analysis. Spatial coherence Annual maps from each of the two methodologies (satellite, Fig 14 left; NWPM, Fig 14, centre), as well as the differences between them (Fig 14, right), are shown below. It can be observed that the differences are fairly constant throughout the country, and of a similar order to those found in the station by station analysis. The greatest similarities between the two sources are found in central Tanzania. 30° E 35° E 40° E 30° E 35° E 40° E 30° E 35° E 40° E 0° 0° 0° 1 2 1 2 1 2 3 3 3 45 45 45 6 6 6 5° S 7 9 8 5° S 7 9 8 5° S 7 9 8 10 11 10 11 10 11 12 13 12 13 12 13 14 14 14 10° S 15 10° S 15 10° S 15 16 16 16 4 4.5 5 5.5 6 6.5 7 4 4.5 5 5.5 6 6.5 7 -1.5 -1 -0.5 0 0.5 1 1.5 GHI (SAT) in kW h/m2 dia GHI (SKR) in kWh/m2 dia DIF (SAT-SKR) in kWh/m2 dia Fig 14 GHI annual map for Tanzania from satellite (left) and NWPM (center) methodologies and their differences (right). An additional analysis of spatial coherence for each of the sources has been performed, based on the fact that adjacent areas should have similar irradiation and, therefore, irradiation changes between a point and its neighbour should be modest. In terms of probability, there should be more small variations than larger ones. In order to perform this analysis, irradiation values are divided into 10 categories at the percentiles boundaries: 10, 20, 30, 40, 50, 60, 70, 80 and 90. Next, the variable "changes" (or "jumps"), and their probability distribution are calculated. The values of that variable (in a range from -9 up to 9) as well as their corresponding probabilities (PDFs) by months are shown in Fig 15 for satellite (left) and NWPM (right) derived solar data. Accordingly, Fig 15 shows the probabilities corresponding to the different jumps between radiation categories when a shift of one degree occurs (latitude→upper graphs; longitude → lower graphs). Page 21 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 15 PDFs to latitudinal and longitudinal variations for satellite (left) and NWPM (right) data. NWPM data seems to show a greater spatial homogeneity. It is likely that the actual behaviour of spatial variability will lie between the results of the satellite and NWPM, and an appropriate combination of both will provide the closest analogue to the real behaviour. Page 22 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 4.3.3 Combination and creation of the model Cluster analysis is a commonly used method of combining data sources, as in (Diabate, Blanc, & Wald, 2004)’s clustering of climatic zones in Africa. In this study a cluster analysis based on the k-means method has been applied in order to combine the solar radiation modelled data, using the sequences of monthly GHI values measured at the selected stations. Following pre-analysis of individual modelling of each station, stations 1 and 2 (section 4.3.1) were not considered in the model as they did not show a relationship with the variables being used as estimators. The steps followed for performing the combination model are described below: 1. Establishment of a general regression model. The variables needed to adjust the model. Three different models were tested and the following variables were selected: a. Explanatory variables i. Monthly long-term GHI values estimated by the satellite analysis (GHIsat). ii. Monthly long-term GHI values estimated by NWPM (GHIskr). iii. Monthly mean values of the declination (changing per month) (Dec). iv. Latitude values of each location (Lat). v. Longitude values of each location (Lon). b. Dependent variable vi. Monthly long-term values measured at each station (GHImonth). Thus, the linear multi-regression model proposed to be applied in each cluster is: 23456 = '( + ') 785 + '$ 79: + '/ 01 + '; <=> + '? <@ 2. Clustering process Due to number of stations available, modelling with 3 to 4 clusters is most appropriate, allowing each cluster to incorporate on average three to four stations. Since the clustering process is dependent on the initial selection (different cluster classification can be obtained for a same number of clusters depending on the selected seed), the selection process was repeated 100 times (one order of magnitude higher than number of potential seeds, that is to say, the number of stations - 14). For each iteration of the clustering process, the proposed model was fitted in each cluster and the R2 value calculated. The clustering iteration that produced the greatest R2 values across all clusters was selected. The outcome of the process was the selection of three clusters. Whilst the use of 4 clusters was analyzed, in those iterations with the best fit, always had station 4 was always alone in one of the clusters; hence it was excluded from the analysis. Thus, the output shown below (Table 4) is based on the three clusters finally selected. Page 23 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Table 4 Clustering of the ground stations. GHI STATION Lat (S) Lon (E) ID CLUSTER (kWh/m2) Mwanza 2.47 32.92 5.51 3 3 Kilimanjaro 3.42 37.07 4.76 5 1 Same 4.08 37.72 4.72 6 1 Kigoma 4.88 29.63 4.60 7 2 Kondoa 5.00 35.75 5.89 8 3 Tabora Arpt 5.08 32.83 5.65 9 3 Dodoma 6.17 35.77 5.88 10 3 Zanzibar 6.22 39.22 4.91 11 2 Morogoro 6.83 37.65 4.40 12 1 Dar Es Salaam 6.88 39.20 4.88 13 1 Iringa 7.67 35.75 6.24 14 3 Mtwara 10.27 40.18 4.62 15 2 Songea 10.68 35.58 4.18 16 2 3. Results of the fitted model in each cluster. Accordingly, the regression results for the selected model in the selected clustering iteration are shown: ) = −6.42 + 0,25 785 + 0.16 79: − 1.62 01 + 4.48 <=> + 13.88 <@ R2=0.84 $ = 1.37 + 0,25 785 + 0.11 79: − 0.17 01 + 7.69 <=> + 3.66 <@ R2=0.74 / = −2.67 + 0,18 785 + 0.41 79: + 0.06 01 − 0.52 <=> + 8.11 <@ R2=0.73 In general, longitude is observed to have a strong impact in all clusters. In cluster 1, encompassing stations in eastern Tanzania, the impact of longitude is the most evident, with a bias towards higher values than those measured, as reflected in the origin ordinate. In cluster 2, encompassing the stations of the eastern, south and west of the country, there is a greater dependence on latitude than on longitude. The contributions of satellite and Skiron modelling results are very similar as in cluster 1, though with a slightly stronger contribution in the case of the satellite. In cluster 3, encompassing the rest of stations in the central area of Tanzania, a lesser dependence on latitude than in the previous clusters is observed. Here Skiron shows a greater contribution than the satellite. 4. Generation of the coefficients layers and model behaviour From the values of the model coefficients for each station (depending on the assigned cluster), layers related to each coefficient (from β 0 to β5) of the general model are created so that they can be applied to the whole of the territory at the available resolution of the input variables: 0.05° (Fig 16). These coefficients are extrapolated using a computing method based on the Delaunay triangulation and on the interpolation by the "Natural neighbour interpolation". Page 24 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 16 Spatial distribution of the estimated model’s coefficients. In Fig 17 a scatterplot of the model behaviour is shown. Fig 17 Behaviour of the final model. Page 25 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 4.4 Solar resource assessment for map generation 4.4.1 Generation of GHI maps The combination of satellite and NWP models reproduces the monthly GHI measured values more closely than any individual model used (section 4.3). The underlying idea behind the combined approach is to use each method’s unique features to capture different patterns in the data, offering the possibility to take advantage of the strongest points of different stand-alone assessment techniques. Therefore, GHI monthly values used in the generation of the maps have been computed by means of the combination model (section 4.3). Annual GHI maps have been computed from monthly maps. 4.4.2 Generation of DNI maps DNI monthly values for the map generation have been computed from GHI monthly values calculated by means of the combination model (section 4.3). Annual DNI values for map generation have been computed from the corresponding monthly series. The expression used to calculate monthly DNI values was developed by Ari Rabl using 26 SOLMET stations located in the United States (Rabl, 1981). In this work, Rabl found a relation between yearly average DNI (IDNI), and yearly average clearness index, Kt: CDE = 1.37 ∙ -5 − 0.34 In this study, this expression is used to describe the relationship between monthly average DNI and monthly average Kt, whose physical manifestation is similar (Moreno et al., 2010). 4.4.3 Generation of DHI maps DHI monthly values for map generation have been computed from GHI monthly values calculated by means of the combination model (section 4.3). Annual DHI values for maps generation have been computed from the corresponding monthly series. The expression used to calculate monthly DHI values was developed by Page (Page, 1961), based on regression analysis of data measured from locations between 40ºN and 40ºS latitudes. In this work, Page derived the following linear dimensionless equation: -F = = + G ∙ -5 where Kd is the monthly diffuse fraction (the DHI to GHI ratio), and a and b the climatologically determined regression coefficients. The mean values for a widely scattered set of stations across the world found by Page were a = 1.00 and b = 1.13. 4.4.4 Generation of GTI maps Tilt angles have a decisive impact on fixed tilted PV power plant design. Modules are tilted for optimized annual irradiation on the modules’ surface. When it comes to photovoltaic power plants, most solar systems are oriented and tilted, while solar resource data is typically available only for the horizontal plane. It is therefore necessary to transpose irradiance from horizontal to the tilted and oriented plane: the global tilted irradiation (GTI). GTI is derived once the GHI, DNI and DHI have been estimated, and Page 26 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report following purely geometric arguments, related to the position of the sun with respect to the site (θ, solar zenith angle; φ: solar azimuth angle; γ: surface azimuth angle (deviation with respect to the South direction)) and the plane’s inclination with respect to the ground (β). Hence, global irradiation (GTI) over an inclined surface is calculated as: 1 F8HIJ 1 , = · · K1 − cos 'L · + .M · · cos + · · K1 + cos 'L 2 2 depending upon GHI, DNI, DHI, daily ground albedo and the geometry, with cos ' · cos + Kcos N · cos O − sin N · sin OL · sin · sin ' .M = cos The evaluation of the best adapted tilt angles for optimized irradiation on modules shows that at latitudes near the equator, irradiation cannot be significantly improved by tilt angles, whereas in latitudes lower than 30°S and higher than 30°N a significant increase in irradiation on module surface can be expected. As a consequence, a preliminary analysis of the optimal tilt has been evaluated as a function of the GHI to GTI ratio for the whole period available (Fig 18). Fig 18 Mean GHI to GTI ratio in Tanzania for the period 2003-2013, shown in the map (left) and represented in a frequency histogram (left) Mean GHI to GTI ratio shown in Fig 18, left, suggests a latitudinal dependence: ~ 1 near the equator increasing with increasing latitude values to ~ 1.05. This is due to a combination of seasonal trends of solar irradiation in conjunction with solar geometry throughout the year. The quartile analysis of GHI to GTI ratio indicates that this ratio is above 0.98 in 75% of months. Moreover, it has been found that in 60% of months GHI is greater than GTI. As a consequence of the analysis, the optimal tilt angle proposed in Tanzania is 0º, thus GTI maps are effectively equivalent to GHI maps. Page 27 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 4.4.5 Generation of PV potential maps The knowledge on photovoltaic potential (PV potential) is essential to predict the economic performance of a PV system. If we consider the amount of electric energy produced by a PV system calculated as: R = S9 · S. · (4.1.4.1) where S9 is the unit peak power, S. is the system performance ratio, and is the global irradiation on a horizontal, vertical or inclined plane of the PV module (kWh/m2) (Súri et al., 2007), then, the PV potential is defined as the electric energy produced per peak power installed: R ST = = S. · (4.1.4.2) S9 The peak power characterizes the nominal power output of the PV modules at Standard Test Conditions (STC) (IEC/TS 61836, 1997), i.e., when the irradiance in the plane of the PV modules is 1000 Wm-2 and the temperature of the modules is 25ºC and an air mass of 1.5. Having the PV potential in these units, there is no need to know about the PV conversion efficiency or the module area. STC are ideal conditions that produces PR=1 but, in practice, the output of a PV system is lower than the peak power, for example, when the operating temperature is higher than 25ºC. The performance ratio (IEC 61724, 1998) accounts for the causes that affect the performance of a PV system, so it is essential to quantify it in order to estimate the best as possible the behaviour of a PV plant. The most important causes that lead to a reduced PR are module temperature, attenuation of incoming radiation due to soiling, dirt and dust, mismatch between connected modules, ohmic resistance in the system wiring and losses due to DC to AC conversion (Woyte et al., 2013). Nowadays, PR has reached values over 0.80 (van Shark et al., 2012, Reich et al., 2012). Based on several studies on performance of PV systems (Marion et al., 2005, Monedero et al., 2007), in this work we have characterized the PR on monthly basis, by factorizing it in two terms: 2 S. 2 = S.UV2W · S. 2 U (4.1.4.3) where S. 2 U represents an overall typical derate factor at STC conditions that includes all possible effects except the module temperature, and is set to a value of 0.82 according to reviewed bibliography, 2 and S.UV2W accounts for the performance due to the module temperature, expressed as: 2 2 (4.1.4.4) S.UV2W = 1 − ,X − 25ºZ ZU 2 where ZU is the module power temperature coefficient, set to a typical value of 0.0043ºC-1, and ,X is the monthly averaged module temperature, calculated as: Page 28 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report 2 (4.1.4.5) 2 2 \Z, − 20ºZ ,X = ,82[ + · 2 ] ℎ 800 $ ^ where NOTC is the Normal Operation Temperature, set to a typical crystalline silicon module value of 2 46ºC, ,82[ is the ambient temperature, which is taken from SKIRON model output, ℎ2 is the mean day 2 `6 duration of month m in hours and the tilted irradiance in . 2a Therefore, the PV potential on monthly basis shall be calculated as: 2 ST 2 = 0.82 · S.UV2W · 2 (4.1.4.6) 5 SOLAR RESOURCE MAPS The maps developed in this first phase of the project are delivered at 0.05º spatial resolution in formats suitable for GIS (Geographic Information System) software; in Raster (GeoTiff) and Vector format, based in the WGS84 projection system and with angular unit decimal precision. The data delivered are compatible with the Global Atlas for Solar & Wind from IRENA (International Renewable Energy Agency). As previously explained, the GTI variable is assumed as equivalent to GHI, hence four outputs are delivered as surface information: GHI, DNI and DHI, expressed in kWh/m2d, and PV potential expressed in kWh/kWpeak. For each output the following layers have been implemented. 1. Annual mean value (one layer for each output). 2. Monthly mean value (12 layers for each output). 3. Available monthly data (132 layers for each output corresponding to the time period with Satellite and NWPM information). The principal characteristics observed in the different solar radiation outputs are explained further below. 5.1 Global Horizontal Irradiance This section shows the results related to the annual GHI and inter-annual GHI variability maps, following the procedures described in section 4.3. Monthly mean maps can be seen in Annex 1. GHI shows highest values in central areas of Tanzania, gradually decreasing in all directions of the country, with minimums in southwest. Maximum values of 6.21 kWh/m²d are reached around latitude 6.25S and longitude 35.75E (Fig 19). Page 29 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 19 Long-term estimation of annual GHI in kWh/m2d Fig 20 shows the frequency histogram of annual GHI values in Tanzania. 83% of Tanzania’s land area has annual average GHI above 4.5 kWh/m²d and 14% is above 5.5 kWh/m²d. The minimum GHI value estimated is 4 kWh/m2d. Fig 20 Histogram of annual values of GHI Page 30 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 21 shows the variance estimated for annual GHI values. A low interannual GHI variability is expected in all the territory. Higher values of variance are found in the north and central eastern region of the country, being in all cases below 0.5 kWh/m2d. 2 2 Fig 21 Annual variance of GHI (kWh/m d) 5.2 Direct Normal Irradiance This section shows annual DNI and inter-annual DNI variability maps, following the procedures described in section 4.4. Monthly DNI maps can be seen in Annex 2. DNI shows highest values in central area of Tanzania, gradually decreasing around the country with minimums in south. Maximum DNI values of 6 kWh/m²d are reached in the central regions of Tanzania (Fig 22). Page 31 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 22 Long-time estimation of annual DNI in kWh/m2d Fig 23 shows the frequency histogram of annual DNI values in Tanzania. 44% of Tanzania’s land area has an annual average DNI of4 kWh/m²d and above. Fig 23 Histogram of annual values of DNI Page 32 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 24 shows annual variance of DNI (expressed in (kWh/m2d)2). DNI has higher variance values than GHI, reaching 0.7 (kWh/m2d)2; variance is also distributed more evenly across the country. 2 2 Fig 24 Annual variance of DNI (kWh/m d) 5.3 Diffuse Horizontal Irradiance This section shows the annual DHI and inter-annual DHI variability maps, following the procedures described in section 4.4. Monthly means are can be found in Annex 3. DHI shows highest values in the peripheral regions of Tanzania, gradually decreasing towards the central regions of the country. Annual DHI values occur within a narrow range, between 1.82 and 2.2 kWh/m²d (Fig 25). Page 33 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 25 Long–term estimation of annual DHI in kWh/m2d Regarding the variability of annual DHI values (Fig 25), the maximum values are found in the central regions of the country, decaying in the peripheral regions (especially in the north). Fig 26 shows the frequency histogram of annual DHI values in Tanzania. 75% of Tanzania’s land area has an annual average DHI above 2.17 kWh/m²d, and 90% is above 2.18 kWh/m²d. Page 34 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Fig 26 Histogram of annual values of DHI 6 CONCLUSIONS Satellite or NWPM based solar radiation data by themselves are not capable of reproducing instantaneous solar irradiance values or local patterns with the same accuracy as ground sensors, but they can provide robust aggregated values over a long history and wide territories. In this study, the combination of satellite and NWP models has been used to reproduce monthly GHI measured values that are closer than individual models are capable of. Accordingly, the GHI monthly values used in the generation of this study’s maps have been computed by means of the combination model developed over the course of the study. DNI and DHI monthly values have been calculated from these GHI monthly values using well established methodologies. Maps for monthly and annual GHI values were constructed using the best combination (section 4.3.3) of the available modelled sources (satellite and NWPM, sections 4.1 and 4.2 respectively) and ground measurements (section 3.2). Long-term estimates of solar irradiation (section 5) indicate that Tanzania has a high solar potential. The highest annual GHI values found are in the central regions of the country (Fig 19), where annual average GHI values exceed 6 kWh/m2d. There are also some northern regions where annual average GHI values are greater than 5.5 kWh/m2d. Similarly the highest annual average DNI values (6 kWh/m2d) are in the central regions of the country (Fig 22), and some parts of the northern region (5.5 kWh/m2d). A certain amount of correlation is found between high irradiation zones and arid climates regions (section 2.1). Southern regions of Tanzania, which have shown the lowest irradiation values, are likely influenced by the wet season, which ranges from October to April or May. The coastal regions of Tanzania, warm and humid, also show low irradiation values. The interannual variability in solar radiation is mainly caused by changes in cloudiness and aerosol loading, driven by Page 35 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report changes in regional atmospheric circulation patterns. These changes, especially in clouds, have a dramatic impact on DNI. Solar radiation scattered in the atmosphere increases DHI component, which has effect on GHI. This behaviour can be seen in Fig 21andFig 24, where annual variances of GHI and DNI are shown. Higher irradiation variability is found in a small region of north-eastern Tanzania where a Monsoon climate prevails (section 2.1). This is the high mountain region where Kilimanjaro (5895 m above sea level) is located. These results are in qualitative agreement with previous solar resource assessments in Tanzania. The work of Alfayo & Uiso (section 3.1.1) assigns to the central regions of the country high GHI annual values (6.5 kWh/m2d), and also in some of the northern regions (up to 6.1 kWh/m2d). The work of Alfayo & Uiso, however, also assigns higher irradiation values at a region slightly displaced to the south eastern regions of the country, in contrast to the results of this study. PVGIS and NREL show also similar numeric and spatial irradiation patterns to those found in the present work, but they assign slightly higher values in the northern part of the country in comparison with the maximum values found in the central regions. It is worth highlighting that the solar resource models used in this report are not validated with ground- based coincident solar data, measured using well-maintained, high accuracy measuring equipment. Therefore, an unknown deviation between the actual solar resource in Tanzania and the modelled data may exist. The solar renewable energy community depends on radiometric measurements to develop and validate solar radiation models (Gueymard & Myers, 2009). The correct planning and execution of a high quality measurement campaign (with well-maintained, high accuracy measuring equipment) is essential for minimizing uncertainties in solar resource assessment. The solar resource assessment and variability analysis presented here provides the means by which measurement networks may most efficiently characterize the solar resource in Tanzania. 7 REFERENCES Alfayo, R., & Uiso, C. B. S. (2002). Global solar radiation distribution and available solar energy potential in Tanzania. Physica Scripta, 97, 91-98. Dagestad, K. F. (2005). Estimating global radiation at ground level from satellite images. Dagestad, K. F., & Olseth, J. A. (2007). A modified algorithm for calculating the cloud index. Solar Energy, 81, 280-289. Diabate, L., Blanc, P., & Wald, L. (2004). Solar radiation climate in Africa. Solar Energy, 76(6), 733-744. Gastón, M., Pagola, I., Fernández-Peruchena, C. M., & Blanco, M. (2012). Using learning machines to create solar radiation maps from numerical weather prediction models, ground measurements and satellite images. Paper presented at the Proceedings of the SolarPACES. George, R., & Maxwell, E. (1999). High-Resolution Maps of Solar Collector Performance Using A Climatological Solar Radiation Model. Paper presented at the Proceedings of the 1999 Annual Conference, American Solar Energy Society. Page 36 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Gueymard, C. A. (2008). REST2: High-performance solar radiation model for cloudless-sky irradiance, illuminance, and photosynthetically active radiation - Validation with a benchmark dataset. Solar Energy, 82, 272-285. Gueymard, C. A. (2010). Progress in direct irradiance modeling and validation. Paper presented at the Solar 2010 Conf., Gueymard, C. A., & Myers, D. R. (2009). Evaluation of conventional and high-performance routine solar radiation measurements for improved solar resource, climatological trends, and radiative modeling. Solar Energy, 83(2), 171-185. Hammar, L. (2011). Distribution of Wind and Solar Energy Resources in Tanzania and Mozambique: Chalmers University of Technology, ISSN 1404-8167. John, J., & Mkumbwa, M. (2011). A Cost Effective Solar PV Power Solution for Rural Household in Tanzania, Renewable and sustainable energy. Kallos, G., Nickovic, S., Papadopoulos, A., Jovic, D., Kakaliagou, O., Misirlis, N., et al. (1997). The regional weather forecasting system SKIRON: An overview. Paper presented at the Proceedings of the symposium on regional weather prediction on parallel computer environments. McArthur, L. J. B. (1998). Baseline Surface Radiation Network (BSRN). Operations manual V1.0. Geneva (Switzerland)o. Document Number) McSweeney, C., New, M., & Lizcano, G. (2012). UNDP Climate Change Country Profiles. Tanzania o. Document Number) Moreno, S., Ramírez, L., Fernández-Peruchena, C. M., Elena, P., Bernardos, A., & Manuel, S. (2010). Long Term Estimation of DNI From Solar Global Irradiation. a Model to Estimate DNI From Monthly Clearness Index. Paper presented at the Proceedings of the SolarPACES, Perpignan, France. Mueller, R. W., Dagestad, K. F., Ineichen, P., Schroedter-Homscheidt, M., Cros, S., Dumortier, D., et al. (2004). Rethinking satellite-based solar irradiance modelling: The SOLIS clear-sky module. Remote Sensing of Environment, 91, 160-174. Ohmura, A., Gilgen, H., Hegner, H., Müller, G., Wild, M., Dutton, E. G., et al. (1998). Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate research. Bulletin of the American Meteorological Society, 79(10), 2115-2136. Osima, S. E. (2014). Observed climate data in Tanzania. Rwanda, Kigalio. Document Number) Page, J. K. (1961). The estimation of monthly mean values of daily total short-wave radiation on vertical surfaces from sunshine records for latitude 40 N-40 S. Paper presented at the UK Conf. on New Sources of Energy, Paper No. S/98. Peel, M. C., Finlayson, B. L., & McMahon, T. A. (2007). Updated world map of the Köppen-Geiger climate classification. Hydrology and earth system sciences discussions, 4(2), 439-473. Perez, R., Ineichen, P., Maxwell, E., Seals, R., & Zelenka, A. (1992). Dynamic global-to-direct irradiance conversion models. ASHRAE Transactions, 98, 354-369. Page 37 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Pérez, R., Seals, R., & Zelenka, A. (1997). Comparing satellite remote sensing and ground network measurements for the production of site time specific irradiance data. Solar Energy, 60, 89-96. Polo, J., Zarzalejo, L. F., Martin, L., Navarro, A. A., & Marchante, R. (2009). Estimation of daily Linke turbidity factor by using global irradiance measurements at solar noon. Solar Energy, 83, 1177-1185. Rabl, A. (1981). Yearly average performance of the principal solar collector types. Solar Energy, 27(3), 215-233. Rigollier, C., Lefèvre, M., & Wald, L. (2004). The method Heliosat-2 for deriving shortwave solar radiation from satellite images. Solar Energy, 77, 159-169. Rigollier, C., & Wald, L. (1998). Using Meteosat images to map the solar radiation: improvements of the Heliosat method. Paper presented at the Proceedings of the 9th Conference on Satellite Meteorology and Oceanography, Darmstadt, Germany. Ruiz-Arias, J. A., Alsamamra, H., Tovar-Pescador, J., & Pozo-Vázquez, D. (2010). Proposal of a regressive model for the hourly diffuse solar radiation under all sky conditions. Energy Conversion and Management, 51(5), 881-893. Schillings, C., Meyer, R., & Mannstein, H. (2004). Validation of a method for deriving high resolution direct normal irradiance from satellite data and application for the Arabian Peninsula. Solar Energy, 76, 485-497. Suri, M., Huld, T., & Dunlop, E. (2005). PV-GIS: a web-based solar radiation database for the calculation of PV potential in Europe. International Journal of Sustainable Energy, 24, 55-67. Suri, M., Huld, T., Dunlop, E., Albuisson, M., & Wald, L. (2006). Online data and tools for estimation of solar electricity in Africa: the PVGIS approach. Paper presented at the 21st European photovoltaic solar energy conference and exhibition. Zarzalejo, L. F., Polo, J., Martín, L., Ramírez, L., & Espinar, B. (2009). A new statistical approach for deriving global solar radiation from satellite images. Solar Energy, 83, 480-484. Zelenka, A., Perez, R., Seals, R., & Renne, D. (1999). Effective accuracy of satellite-derived hourly irradiances. Theoretical and Applied Climatology, 62, 199-207. Page 38 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report ANNEX 1. MONTHLY GLOBAL HORIZONTAL IRRADIANCE: Page 39 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Page 40 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report ANNEX 2. MONTHLY DIRECT NORMAL IRRADIANCE: Page 41 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Page 42 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Page 43 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report ANNEX 3. MONTHLY DIFFUSE HORIZONTAL IRRADIANCE: Page 44 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Page 45 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report Page 46 Renewable Energy Resource Mapping: Tanzania. Interim Solar Modelling Report | February 2015