Methodology of Köppen-Geiger-Photovoltaic climate classification and implications to worldwide mapping of PV system performance

A Photovoltaic (PV) already proves but even more promises to be massively deployed worldwide. To evaluate the performance of PV systems globally and assess risk due to different climate conditions, we propose a methodology for the global Köppen-Geiger-Photovoltaic (KGPV) climate classification that divides the globe into 12 zones with regard to the temperature, precipitation and irradiation, and standardizes the evaluations of the performance in regions with similar climatic conditions. Additionally, we present implications of KGPV to si- mulated PV performance using monthly data, for current and future operation of PV systems worldwide including climate change scenarios. A set of electrical and thermal performance indicators of crystalline silicon PV modules in different KGPV zones is analyzed and their evolution over time due to climate changes caused by high greenhouse gas emissions discussed. Results show that the KGPV scheme proves to be a convenient methodology to relate the KGPV climate zones with PV performance.


Introduction
Nowadays, solar photovoltaic (PV) systems are a viable and competitive option for the transition to sustainable energy systems. This technology can be used in any location and at any scale (in residential, commercial and also industrial sectors), offering us countless opportunities to produce renewable electrical energy for our needs. Trend studies present an exponential growth in total installed capacity together with further increase of efficiency and further decrease of the system prices, which boost the interest of broader communities across the globe and attracts new stakeholders to the variety of markets (ETIP-PV, 2017;IEA PVPS, 2018). The global installed capacity of PV systems reached 480 GW in 2018 (IRENA, 2019) and is projected to surpass the 1 TW by 2021-2022(SolarPower Europe, 2018. To support and expand the worldwide deployment of solar PV energy, the PV community raised interest in the understanding and improvement of the long-term performance and reliability of PV modules and systems under different climate conditions (e.g. SOLARTRAIN (Weiß et al., 2017), Pearl-PV (Reinders et al., 2018)) and as mentioned in (Haegel et al., 2019), adopted PV module designs related to specific climate stresses will be needed.
In this regard, a simplification of the problem can be addressed by categorizing the regions with similar climate conditions. Examples of categorization of certain climate zones for single locations or single countries have been presented in the literature: Jordan et al. have defined the Desert, Hot & Humid, Moderate and Snow climate zones for the study of the PV modules degradation (Jordan et al., 2016); Kohl et al. classified ground measurement stations into Maritime, Moderate, Arid, Alpine and Tropical (Koehl et al., 2018); Eder et al. have correlated indoor and outdoor conditions defining the climate zones as Tropical, Arid, Moderate and Alpine (Eder et al., 2018); and Dash et al. categorized regions across India using temperature and irradiation (Dash et al., 2017).
However, according to our knowledge, no global climate classification scheme for PV existed, which could help stakeholders to easily analyze the regions of their interest. The PV community has been mostly using the Köppen-Geiger (KG) climate classification scheme (Kottek et al., 2007), which is composed of three letters for each location: (1) main climate, (2) seasonal precipitation type and (3) level of heat. This scheme is widely used because of the easy understanding of categories but unfortunately includes only temperature and precipitation, while letters (2) and (3) provide not significant information for PV.
Furthermore, the lack of standardization of climate zones brings confusion in the community since different terms are used to name the climate of specific regions.
Those issues motivated us to develop a new climate classification scheme combining temperature, precipitation and irradiation, called as the Köppen-Geiger-Photovoltaic (KGPV) climate classification. The KGPV scheme is based on the KG climate classification but extended by the solar irradiation since this variable is the primary one for solar PV energy production. Accordingly, it includes two classification levels: the Temperature-Precipitation (TP-) zones and the Irradiation (I-) zones. Preliminary results for the US and Chile were presented in (Ascencio-Vasquez et al., 2018). Hereby, we present an update of the scheme and also the worldwide KGPV map composed by 12 climate zones.
The usefulness of the KGPV scheme can be proved by comparing the climate zones with relevant PV indicators, such as the energy yield, performance ratio or module operating temperature. According to it, the need for a tool for PV performance modelling is needed. Different alternatives of PV performance assessment over large geographical regions have been already proposed by several research groups (Bora et al., 2018;Dubey et al., 2013;Huld et al., 2010;Kirn et al., 2015;Louwen et al., 2017;Wild et al., 2015). Some of them are already available to the customer as commercial (e.g. PVSyst (PVSyst, 2019)) or freeware tool (e.g. PVGIS (Huld et al., 2010;Huld and Gracia Amillo, 2015;Urraca et al., 2017), PVWatts (NREL andDobos, 2014) or external library PVLIB (Stein et al., 2016)). However, due to the complexity of combining available tools with a climate classification scheme, we combine existing models to create an in-house tool for the estimation of PV indicators. Fig. 1 illustrates the main components, climate variables and factors taken into account for the simulation of a PV system. The parameters affecting the PV performance can be grouped in solar irradiance (extraterrestrial irradiance, beam irradiance, diffuse irradiance, etc.), weather data (ambient temperature, wind speed and direction, precipitation, clouds, etc.), local conditions (soiling, shading, ground albedo, etc.) and PV module data (mounting, angular-reflection losses, spectral losses,

Az
Azimuth J. Ascencio-Vásquez, et al. Solar Energy 191 (2019) 672-685 etc.). At system level, the PV modules are connected in strings to an inverter that converts the electricity from the direct (DC) to alternating current (AC) and regulates the voltage to the one determined by the electrical grid to dispatch the energy to the power system. Having both, the climate classification and the PV performance, bring us the possibility to standardize the evaluations per climate zone. The validity of the model is made by comparing ground measurements of irradiance and temperature from the Baseline Surface Radiation Network (BSRN) stations (Driemel et al., 2018).
Finally, to prove the applicability of our methodology, the climate zones and PV indicators are projected towards 2100, using the dataset "SSP5-8.5" developed in the frame of the Scenario Model Intercomparison Project (ScenarioMIP) (O'Neill et al., 2016). This paper is divided into six sections: first, the datasets are described as well as the methodology to handle the global gridded data; then, the Köppen-Geiger-Photovoltaic climate classification is defined, and followed by the methodology to calculate the performance of PV systems; consequently, the mapping of worldwide PV performance is shown and analysed in view of the new KGPV zones; finally, we evaluate the evolution of PV indicators towards 2100 under the published climate change projections.

General approach and datasets used
The main steps to achieve the global climate classification and global PV performance modelling are illustrated in Fig. 2, where temporal-and spatial-aggregations need to be carried out to combine global gridded datasets for ambient temperature, precipitation and irradiation.

Global gridded climate datasets
The precipitation data is taken from the dataset called "GPCCv2018" and generated by the Global Precipitation Climatology Centre (GPCC) of the German Weather Service (Schamm et al., 2016), which integrates the largest number of precipitation stations worldwide (Duan et al., 2017). This dataset is given with a high resolution (0.1°× 0.1°) and is based on measurements by national meteorological and hydrological services, regional and global data collections as well as stations of World Meteorological Organisation (WMO).
The ambient temperature data was taken from the "CRU TS4.01" dataset of the Climatic Research Unit (CRU), University of East Anglia (Harris et al., 2014), based mainly on the stations network of the WMO, as well as the GPCCv2018. From this dataset, we have extracted the daily average ambient temperature (T ambavg ) for each month to be used in the climate classification scheme, and the monthly maximal and minimal temperature (Tamb max andTamb min ) to model the daylight ambient temperature as a function of time (T amb ) for the PV performance modelling.
Regarding the irradiation data, several data source comparisons and evaluations have been carried out, since the inaccuracy of the estimations can be high (Jones et al., 2017;Urraca et al., 2018cUrraca et al., , 2018aUrraca et al., , 2018bUrraca et al., , 2017. While for temperature and precipitation, thousands of ground-based measurements are available and integrated into large datasets, for solar irradiation no global dataset based on ground-measurements exist, because the ground stations are not equally spread around the world. For example, the Baseline Surface Radiation Network (BSRN) includes only 76 stations worldwide including pyranometers, pyrheliometers, etc. mainly located in the northern hemisphere (Driemel et al., 2018;WRMC-BSRN, 2019).
Due to the lack of the irradiation data, two alternatives to derive the solar data over large regions or globally have been developed: (i) satellite-based models, which estimate the irradiation from the satellite data combined with the ground measurements (Amillo et al., 2018); and (ii) reanalysis models, which are based on large computational coupled models (Atmospheric and Ocean-Wave models) where irradiation is calculated from millions of climate measurements from different sources, such as, ground stations, satellites or aircrafts (Zhang et al., 2016). In this work, we have taken the global horizontal irradiation (GHI ) from the reanalysis-based dataset ERA-Interim (Uppala et al., 2011) provided by the European Centre for Medium-Range Weather Forecasts (ECMWF).
The validation of the synthetic T amb and synthetic GHI is made by comparison with 22 stations BSRN installed around the World (see Table A1 in Appendix). All measurements when GHI is lower than 30 W/m 2 and higher than 1300 W/m 2 are neglected.
The climate variables projected towards the year 2100 are derived from the "SSP5-8.5" climate change scenario, created by the Institute Pierre Simon Laplace (IPSL) (Boucher et al., 2018). The details of this world scenario are presented in (Kriegler et al., 2017;O'Neill et al., 2016;Riahi et al., 2011). The simulations are applied for the time period of 2015 to 2100 on monthly basis and a spatial resolution of 250 km × 250 km.

Temporal-and spatial-aggregation
Since we obtain datasets from different providers and also in different formats, temporal and spatial adjustments are needed to aggregate all the data. We aim a long-term evaluation, thus, it is relevant Fig. 2. General concept used in our approach.
to define a representative time period from which the historical data can be extrapolated and used to estimate the near-future. To choose the time range of historical data, we have evaluated (and relabelled) six climate zones defined by the KG scheme (Kottek et al., 2007): A-Tropical, B-Desert, C-Steppe, D-Temperate, E-Cold and F-Polar climates. The average ambient temperature per each climate zone has been calculated on annual basis. As already reported in climate change reports (IPCC, 2018), rising temperature trends are evident from the middle of the 20th century in all climate zones and they are shown in Fig. 3. Precipitation and irradiation show no clear trend nor pattern.
The evolution of temperature across the full period can be divided into two linear segments: the data before and after 1990 (see Fig. 3). Accordingly, we selected the time period for the climate classification from 1990 to 2016, where the data is averaged on monthly basis. The temporal aggregation is followed by spatial aggregation. The datasets are resampled to a generic global grid resolution, ranging from 180°E to 180°W in Longitude and 90°S to 90°N in Latitude, with a resolution of 0.5°× 0.5°after applying a bilinear interpolation.

The Köppen-Geiger-Photovoltaic climate classification scheme
The aggregated datasets generate the map of the KGPV climate classification scheme presented in Fig. 4. The developed climate classification divides the Globe according to two classification levels: first, the TP-zones are defined following the threshold criteria proposed by (Kottek et al., 2007) and relabelled as A-Tropical, B-Desert, C-Steppe, D-Temperate, E-Cold and F-Polar climates; and second, the I-zones are defined as L-Low, M-Medium, H-High and K-Very High irradiation zones. The 30th, 50th and 80th percentiles (1130, 1560 and 2070 kWh/ m 2 , respectively) of the global values (excluding Antarctica) for annual global horizontal irradiation (H ann ) are chosen as the thresholds for Izones. Those values were empirically defined for Europe based on the weather conditions on the continent. They also present a reasonable categorization in the rest of the World. Both threshold criteria and order of classification are presented in Fig. A1 in Appendix.
Combining six TP-zones and four I-zones, we get 24 possible KGPV climate zones designated with two letters. However, less important zones can be neglected due to two main reasons: ▪ Land-Surface ratio (LS r ): this indicator is defined as the portion of land per climate zone over the global land surface and given in percentage (see Table 1), ▪ Population Density: since large PV systems are usually installed in or next to populated areas (Bollinger and Gillingham, 2012;Dharshing, 2017;Graziano and Gillingham, 2015), we use this indicator to identify the unpopulated zones, where is less probable to install PV systems. The global population density data has been taken from (Center for International Earth Science Information Network -CIESIN -Columbia University, 2018), and correlated with the climate zones to get the average population density per climate zone (see Table 2).
To reduce the 24 zones, we define a surface-population density indicator by multiplying the land-surface ratio by the population density per climate zone (which represent roughly the number of people per climate zone) and neglect small areas with low population to get 12 climate zones (represented as the highest 11 values in Table 3, plus one zone for Polar climates). The neglected zones are relabelled to the closest neighbouring zone in view of the irradiation criterion as shown in Table 4. As an exception, the climate zone "FL" (Polar climate with Low Irradiation) is included in the map as a representative of all the Polar zones.

Worldwide PV performance modelling
Besides the climate classification, we have developed a comprehensive PV performance modelling using global gridded monthly data to calculate thermal and electrical indicators at any location. The highest uncertainty comes from ERA-Interim data due to the modelling of the cloudiness (Urraca et al., 2018b). However, also other models used for clear sky, decomposition and transposition modelling can introduce more uncertainties because they are validated only for specific locations around the world (Hofmann and Seckmeyer, 2017a). The selection of models is based on uncertainty studies (Lave et al., 2015; Skoplaki and Palyvos, 2009;Urraca et al., 2018b), privileging simplicity to reduce processing speed.
To get the long-term performance of PV systems, we simulate a typical day (from sunrise to sunset) for each month. Daily values are multiplied by the number of days in each month and summed up to the annual value. Performance modelling is done in three stages: the selection of input parameters, the modelling of temperature and irradiation, and the modelling of PV performance.     Table 4 Assumptions applied on KGPV labelling procedure merging to most relevant zones next to them indicated by the arrows. Different colours label TP-zones.

Input parameters
The first stage is presented in Fig. 5, where we define the models with initial parameters and input variables. From the timestamp and geographical data of the location (denoted as Gd, including the latitude, longitude and altitude), we calculate the sun position (SP), Sun Azimuth angle (SunAz) and Sun Zenith angle (SunZen), using the NREL SPA Algorithm (Reda and Andreas, 2008). The orientation (or azimuth angle, denoted as Az) and inclination angle (or tilt, denoted as ) of fixed-mounted PV modules is considered optimal once the PV module or system produces the maximal annual energy output (Palmer et al., 2018). Optimal position (OP), including Az and angles worldwide, are calculated by using a 3-order polynomial empirical model as reported in (Jacobson and Jadhav, 2018).
We define the following parameters for the PV system model:

Solar irradiance and ambient temperature modelling
Global Plane-of-Array Irradiance (G PoA ) for a typical day of each month is calculated from ERA-Interim irradiation data (GHI ERA ). First, we model the Clear Sky Global Horizontal Irradiance (GHI cs ) using the Simplified-Solis Model (Ineichen, 2008) from pre-processed data and formulas included in the PVLIB python package (Stein et al., 2016).
To get the GHI as a function of time (denoted asGHI t ), we calculate a Daily Clear Sky Index (k cdaily ) defined as the ratio of the GHI ERA and the daily sum ofGHI t ( ) cs for a typical day of each month (see Eq. (1)). The GHI t is calculated hourly from sunrise to sunset by multiplying GHI cs and k cdaily (see Eq. (2)).
Once the GHI t is calculated, we implement the Erbs decomposition model (Erbs et al., 1982), to obtain the diffuse horizontal irradiance (DHI t ) and direct normal irradiance (DNI t ), needed for the PV performance calculation (Kirn et al., 2015). Then, the transposition of irradiances has to be applied since the PV modules are tilted. The Angle of Incidence (AOI) for tilted modules is calculated and integrated into the Hay-Davies transposition model (Loutzenhiser et al., 2006)  The ambient temperature is modelled using two sinusoidal functions during the daylight time (Waichler and Wigmosta, 2003), where the minimum temperature is set at the sunrise and the maximal temperature is located at the middle between noon and sunset.

PV performance modelling and performance indicators
Module temperature losses are estimated through the Module operating temperature (T mod ) calculated using the Ross model (Skoplaki and Palyvos, 2009) T mod strongly affects the power and energy production, since high operating temperatures generate thermal losses and decrease the efficiency of the PV module (Bogenrieder et al., 2018;Dash and Gupta, 2015;Hegedus, 2013;Schweiger et al., 2017). This phenomenon is expressed by the power-temperature coefficient ( ). The temperature loss factor ( f temp ) is calculated as: The PV energy output (E PV ) of the simulated PV system depends on climate variables, PV module characteristics and long-term degradation of the PV module materials and system components, and it can be calculated as: where PL t ( ) i is the total Performance Loss factor, defined as the reduction of the PV system performance over time.
Then, from the synthetic data generated, we calculate several performance indicators (yield, performance ratio, module operating temperature, etc.) that can help to compare and evaluate technologies depending on the location and climate zones for the PV deployment worldwide. Typically, long-term PV assessments are represented by the indicators expressed in the IEC standards (International Electrotechnical Commission, 1998), thus, we calculate the reference energy yield (Y r ), final energy yield (Y f ) and performance ratio (PR) using standard formulas. We also include the Unit Capacity Factor (UCF ) to characterize the availability of the annually produced energy, which can support the comparison among conventional and renewable power sources, or in terms of reliability, to identify zones with long operating stress for PV modules.
Additionally, four thermal indicators are calculated from annual basis: the average, maximal and minimal module temperature, as well as the daily cycling temperature. Those indicators are not related only to the decrease or increase of performance but also to the degradation processes of PV systems (Lindig et al., 2018) and PV materials (Eder et al., 2018). The calculation formulas are presented in Table A1. The flowchart for the PV performance modelling and calculation of indicators is procedure is presented in Fig. 7.

Worldwide PV performance mapping
Our methodology is used to simulate a PV system with crystalline silicon (c-Si) modules, since this technology has the highest market share (Philipps and Warnuth, 2019). The c-Si PV modules are typically characterized by a power-temperature coefficient ( ) equivalent to −0.45%/°C and the module temperature at Nominal Operating Conditions (T NOC ) equal to 44°C, deriving in a Ross coefficient (k Ross ) of 0.03°C m 2 /W. The balance-of-system efficiency ( BoS ) was selected to be 85% for all locations in the world. In this case study, shading, soiling and snow losses are excluded and the Performance Loss factor (PL) representing degradation is neglected. Spectral and angular-reflection losses are calculated for each location using the coefficients given in Ruiz, 2005, 1999). Fig. 8 shows the worldwide mapping of Performance Ratio (PR) for a typical PV system with c-Si PV modules. In Fig. A2(a and b), the Final Energy Yield (Y f ) and the maximal Module Operating Temperature (Tmod max ) are also plotted.
The synthetic data in hourly resolution calculated from ERA-Interim is correlated with ground measurements (GHI and T amb ) of 22 stations from the BSRN (more details in Table A1). The data from BSRN stations is aggregated to hourly basis for a typical day per month. Each station is compared to the closest simulated point in the spatial grid.
The scatter plots of both variables in Fig. 9a and b, justify the good agreement between simulated and measured data. The coefficient of determination (R 2 ), root-mean-squared-error (RMSE) and mean-biaserror (MBE) are indicated in the figures. The modelling in all locations present considerable low RMSE and MBE. Despite the relatively high R 2 , the uncertainty varies from location to location.  J. Ascencio-Vásquez, et al. Solar Energy 191 (2019) 672-685

Implications of the KGPV zones to PV performance indicators
One location per KGPV zone was selected to analyse and compare the PV performance of a typical c-Si PV system (see the geographical distribution in Fig. A3). For comparison between deserts, we also included additionally the Atacama Desert-Chile and the Gibson Desert-Australia. In Table 5 we present the simulated PV indicators per climate zone (complementary explanation of indicators is presented in Table  A2. Regarding the final energy yield (Y f ), the highest production occurs as expected at locations with high irradiations (like deserts and steppes). However, due to higher temperatures, the Performance Ratio (PR) in desert areas is lower. Concerning the Unit Capacity Factor (UCF), which represents the percentage of time over one year when the PV system is under operation (Kirn et al., 2017), in places with high irradiation and large amount of sun hours during the year, the UCF can reach over 20%, though in cloudy places, such as the Cold or Temperate climates, it can hardly get over 14%. Tropical climate presents UCFs of 16-18% despite the high irradiation, which is mainly due to the tropical rain and cloud periods along the year and constant high ambient temperatures. In Minnesota-USA, we got an average UCF of 15%, but the temperature cycling ( T mod ) is extremely high which could accelerate the PV module degradation. Notably, Atacama Desert-Chile presents the highest Y f and UCF with relatively low temperatures, which is highly favourable for PV technologies, while Gibson Desert-Australia and Phoenix-USA present also high solar resources in terms of Y f and UCF, but high temperatures can be harmful for the long-term operation.

Application under SSP5-8.5 climate change scenario
Once defined the KGPV map and PV performance modelling structured, we evaluate the evolution of land-surface per climate zone and also the future operation of PV systems in view of the climate change scenarios.

KGPV zones towards 2100
First, the proposed climate classification is calculated from 2015 to  2100 using the SSP5-8.5 scenario (Kriegler et al., 2017;O'Neill et al., 2016;Riahi et al., 2011). To evaluate the trends of the percentage of global land-surface for each climate zone, indicated as Land-Surface ratio (LS r ), we calculate a 10-years rolling mean and relative values compared to the reference period ranging from 2015 to 2024. Fig. 10a and b show the relative changes in the climate zones (TP-zones and I-zones, respectively). Trigged by the rising temperature, the warmer TP-zones, such as A-Tropical, B-Desert and C-Steppes, will increase the LS r under the SSP5 scenario. Vice versa, F-Polar and E-Cold areas decrease their areas along the time.
On the other hand, the tendencies of I-zones do not follow a clear trend. This situation can be related with the work of (Wild et al., 2015), where differences in the evolution of cloud coverage and anthropogenic air pollution influenced by climate mitigation actions were detected at regional levels. Therefore, the land-surface ratio of I-zones evidence changes per period and it cannot be treated as the evolution of ambient temperature, since the changes are more locally dependent. Fig. 11(a and b) show the evolution of the ambient temperature (T amb ) and global horizontal irradiation (GHI) for locations indicated previously. Those climate variables affect directly the performance of PV systems. In all cases, T amb increases, but GHI behaves different depending on the location. The combination of both variables is expressed by the PV module operating temperature (T mod ) and the Performance Ratio (PR) (see Fig. 11(c and d)). For example, under SSP5 scenario, Ljubljana-Slovenia will increase both T amb andGHI , and for this reason, the rise of T mod will lead to decrease in thePR. Minnesota-USA is showing the fastest PR reduction due to the fast rising temperature. In Jos-Nigeria, GHI drops significantly, limiting the decrease of PR only to 1.8% in 2100.

PV performance towards 2100
In general, results indicate a really slight reduction of performance of PV systems due to the climate change, which can reach only 0.04% year, causing a decrease of 1.2% after 30 years lifetime operation for a PV system. Comparing with a typical 0.5%/year degradation rate, the climate change can be considered as minor issue for energy production from c-Si PV modules.

Discussion
The versatile and adaptable structure of the combined KGPV scheme and PV performance modelling allows an easy integration of new climate variables, i.e. wind speed, relative humidity or ultra-violet  ▪ Irradiation models, such as Clear Sky, Decomposition and Transposition have to be improved and validated worldwide or at least in some specific climate zones. ▪ Our selection of the Simplified-Solis model (for estimation of Clear Sky Irradiance) and Erbs model (for transposition of irradiances) has been based on simplicity. New models (e.g. (Hofmann and Seckmeyer, 2017b;Ineichen, 2018)) can provide better results but require the inclusion of more climate variables, which can again, propagate more uncertainty or increase the computation resources needed. ▪ Transposition models need further validation for global applications and for different fixed mounting positions, as well as for tracking systems. ▪ The sun position algorithm is of minor importance due to really low calculation uncertainties (Hofmann and Seckmeyer, 2017a;Reda and Andreas, 2008), but horizon shadings due to mountains or even close obstacles can largely increase the accuracy of outcomes, especially in urban environments or highlands. ▪ For the energy output calculation, we kept the simplest model using the irradiance and ambient temperature. For future calculations, wind speed and non-linearity responses should be considered.  In addition, the climate change evaluation carried out has brought us new challenges regarding the PV performance over the course of the century. In the future, it will be important to include the GHG emissions and atmospheric composition to understand whyPR and T mod are changing in specific locations; is the changing because of the increase/ decrease in pollution? Are they directly correlated with climate mitigation actions?

Conclusion
We are proposing a new Köppen-Geiger-Photovoltaic classification scheme with 12 climate zones to standardize the evaluation of PV systems around the World. The proposed scheme is based on the widely used Köppen-Geiger classification scheme, upgraded with information of the solar irradiation, the most important parameter for photovoltaics.  The KGPV classification scheme divides the Earth, first, into Tropical (A), Desert (B), Steppe (C), Temperate (D), Cold (E) and Polar (F) climates, and then into Low (L), Medium (M), High (H) and Very High (K) irradiation zones. The 24 possible climate zones were reduced by the population density and land-surface ratio to only 12 climate zones of the KGPV scheme to keep the classification simple and easily understandable. With this approach we suggest to sacrifice less relevant zones (microclimates) to make the model friendlier to use.
The presented KGPV classification scheme is given in a resolution of 0.5°× 0.5°, which is enough to keep the computational time in reasonable time frames and evaluate the KGPV climate zones and PV performance globally through the combination of existing irradiance and PV performance models.
The proposed structure is modular and can be easily updated with new models and input data, as well as with new evaluation parameters. In the paper, we are presenting the analysis of performance indicators for the c-Si technology for each climate zone from an electrical and thermal point of view. The uncertainty of the model largely depends on the quality of the input data and can deviate largely from the acceptable values in specific locations like high mountains and coastal regions. However, the developed model together with the used irradiation database proves to be a reliable tool for PV performance modelling.
We have calculated several PV indicators regarding the electrical performance of PV modules and systems. Complementing the typical performance indicators, Performance Ratio (PR) and final Energy Yield (Y f ), we include the Unit Capacity Factor (UCF ) to compare the availability of PV power plants over a year. Furthermore, we introduced thermal indicators to compare locations according to the module operating temperatureT mod . Our study showed that the best location for PV is Atacama Desert-Chile area because of the very high irradiation and high numbers of sun hours due to the relative vicinity to the equator (for example, in comparison to the desert area of Phoenix). However, the highest PR values we obtained in Moscow (EL climate zone -Cold climate with low irradiation) but the annual energy production is lower due to the lower irradiation levels.
Moreover, we have applied our schemes (KGPV climate classification and PV performance modelling) to climate change scenarios. We considered the highest greenhouse gas emission scenario, the SSP5-8.5 under the frame of ScenarioMIP, to evaluate the performance of c-Si PV modules over the 21st century, concluding that cities or climate zones will have different climate evolution due to climate mitigation actions, but anyway due to the global rising temperature under this scenario, the Performance Ratio of c-Si PV modules can decrease in all locations, between 0.5% and 1% by 2050 and between 2% and 3% by 2100. Finally, KGPV classification together with the PV performance analyses can be a powerful tool to make long-term degradation and Levelized cost of energy (LCOE) analyses, and world classification mapping of PV performance including climate change scenarios.

Declaration of Competing Interest
The authors declare no conflicts of interest.