Inter-annual variability

Given the short term variation of the solar resource due to climate and environmental and atmospheric phenomena (tropical storms, changing jet streams, volcano eruptions, etc), it is necessary to collect several years of data in order to develop a long term trend of solar resource variability. Several methods have been developed to correct long term modeled data using short-term measured data and to reduce the uncertainty to +/-5 percent. Inter­annual variability of DNI is much higher (twice as much or more) than that in GHI.

Typical meteorological year (TMY) datasets are in principle, reflecting the long­term trend and thus, absorbing the inter-annual variability in the hourly values. How­ever, it’s very important to keep in mind that solar projects are financed based on a revenue model with forecasted energy production targets. These energy production targets use TMY data as the input to photovoltaic models. At the end of the year, the actual energy produced by the plant is compared to the target and the revenue gener­ated. Therefore, it is important to understand how much less energy than the original target could be expected at the end of a year with poor solar resource. It is desired to narrow this uncertainty as much as possible in order to develop a more precise revenue model while at the same time, having the revenue model include enough toler­ance to accommodate years with insolation lower than typical. A deeper knowledge of the expected inter-annual variability at the project location will reduce risk and can help to provide better financial estimates and contractual terms.

Finding the solar irradiance data applicable to the location of the project is one of the main challenges in the energy output calculations of any solar system. This is because of the generally limited or absolute lack of measured solar data at any given location around the world. Satellite technology, image processing methods, meteorological science and interpolation methods have enabled indirect measurements of solar irradiance on the earth’s surface. These techniques have also enabled the

Table 9.1 Characteristics of solar resource data.

Data Type

• Ground measurements

• Satellite data

• Modeled data (synthetic time series)

Data Quality

• Length of time series

• Data completeness

• Accuracy of instruments

• Data sampling rate (minute, hourly, daily, monthly, yearly)

• Errors and biases


• Long term patterns

• Typical year

• Inter-annual variability

• Uncertainty

generation of synthetic time series based on best available data for a specific location. Table 9.1 provides an overview of solar resource data characteristics.

Ideally, ground measurements would provide the most accurate representation of the solar resource patterns at that specific location. The main challenge with ground measurements is the limited number of stations available and potential data quality issues when the stations exist. Sometimes, a meteorological station is installed a pri­ori at the location of a future PV power plant project but even in this case, the short length of the time series (often less than a year) severely limits the quality of the data set and it should not be considered as the sole source of solar resource data.

A typical meteorological year is a data set comprising hourly values of solar radi­ation and meteorological variables covering a full year, or 8760 hours. These datasets are assembled with long time series of measured data through a method that selects and concatenates this data into one single year. The goal is to generate a representa­tive year that filters the extreme variations from year to year but captures the long term trend. For this reason, a TMY dataset is not necessarily a good indicator of the conditions expected for this year or any particular year. The TMY datasets are highly valuable for computer simulations.

The designer has to develop certain criteria to select the solar resource data set that is thought to best represent the location of the project. To date, there is no stand­ard method to guide the selection of the solar resource data.

In North America, it is worth noting the work done by the National Renewable Energy Laboratory (NREL), which has developed and maintained a comprehensive solar radiation database for the continental territory of the United States as well as Hawaii, Alaska and northern Mexico[13]. Canada is not as well-mapped, although the Canadian Weather Office (Environment Canada) maintains a database of meas­ured data from meteorological stations distributed on its territory[14]. Mexico and the remaining countries in the Americas tend to have only very localized records if at all

Table 9.2 Solar







This was NREL’s first release of a TMY dataset based on data from 1952 to 1975. Due to data quality issues, the use of this data set is not recommended.


The second release of a TMY dataset was completed in 1994 and is based on data collected from 1961-1990 by the National Solar Radiation Data Base (NSRDB). TMY2 data is available for 239 stations. Each station is classified based on the amount of measured data available to compile the TMY, Class A to Class C, with the former having the best quality data.


TMY3 dataset is the third release of NREL’s TMY and is based on the updated NSRDB data from 1991-2005. TMY3 data is based on ground and satellite measurements, including the time series used for the TMY2 dataset. This dataset contains 1,454 sites classified as Class I to Class III, with the former having the best quality data.


This dataset uses hourly radiance images from geostationary weather satellites, daily snow cover data and monthly averages of atmospheric water vapor, trace gases, and the amount of aerosols in the atmosphere to calculate the hourly total insolation incident on a horizontal surface. The methodology was developed by Dr. Richard Perez. The data is available on a 10 by 10 kilometer grid published on NREL’s interactive web page (Solar Prospector).


This dataset is based on satellite data and published by NASA through its Surface Meteorology and Solar Energy website.


Meteonorm is a computer program developed by Meteotest, a meteorology and environment company based in Switzerland. The TMY set generated by Meteonorm is based on ground measurements. This software generates a TMY for any location using an interpolation algorithm that relies on all measured data available in the vicinity of the location. This method calculates solar radiation, temperature and other meteorological parameters.


3Tier is a private company that provides TMY datasets for any location worldwide. The dataset is based on half-hourly, satellite images from 1998 to date. The information collected is processed with proprietary algorithms and peer-reviewed methods published in scientific literature.


This government agency does not generate TMY datasets. Historical and


current measured solar and meteorological data from ground stations is publicly available.

existent and there are often data quality issues. The use of synthetic time series is regu­larly the best option for selecting data in these regions, with notable exceptions.

For energy output projections of PV power plants in the United States, it is regu­lar practice to use NREL’s database as the primary source of datasets. Despite the comprehensiveness of this database, the designer should still verify the quality and applicability of the dataset to the project location. Other sources of solar resource datasets are indicated in Table 9.2.