Step 2: Collect and Assess Observed Climate Data
You may need to use observed climate data to establish a baseline, to evaluate climate model data and/or for climate model bias correction. If you do not need observed climate data, proceed to step 3.
Possible sources of observed climate data include:
- local or national meteorological agencies
- local or national reports
- regional or global datasets.
Be mindful that observed data collection may not be straightforward for some countries. It could involve time and cost associated with this process. There are different sources of observed climate data.
After obtaining the data, it is recommended that you assess the adequacy of the data. Examples of the type of questions you need to ask to assess the data follow.
Does the data have metadata?
Good data is generally accompanied by good metadata (i.e. information about the data and how it was generated). For station data, this metadata may include the code or number of the station, the latitude and longitude of the location, the name of the institution which conducted the observation, whether the data is quality checked or not, code for missing data, data format, and contact person. For gridded data, this may include how the data was constructed and by who, a data version number, and any caveats.
If the collected data does not have metadata, you could trace back by contacting the person/institution/website from which you obtained the data, or reading the literature from which you were referred to the data in the first place. If, after going through this process, the metadata is still not found, you could choose to disregard the data if possible and try another resource. If there is no other data source available, you may reassess the type of data you need (i.e. go back to step 1). If, again, this is not possible and you have to use these data, do so with caution and report the ‘caveat’ when publishing your results (step 10).
Are there gaps in the observed data? Do you need to fill the missing data?
Observed climate data may be incomplete or missing for several reasons, including instrument problems or a station having been closed.
How missing data is dealt with will vary from one case to another and depends upon the purpose of using the observed data and/or the subsequent impact modelling. In some cases, the overall methodology may ‘allow’ the use of data with missing values of less than a given threshold. Other cases may require a complete dataset and hence need a process for filling the missing data.
There are a number of accepted methods to account for missing data. These methods often involve the use of simultaneous values at nearby stations to calculate an estimated value for that missing data time. For rainfall and temperature, some popular methods include the normal ratio method and simple inverse distance weighting. For some variables, gaps in the data might be filled based on other variables; for instance, humidity can be calculated from the minimum temperature and solar radiation can be calculated from extraterrestrial radiation which is a function of geographic location and the Julian day.
Where observational data from stations is not available, alternative data resources like those from remote sensing-based observation could be considered. There are many satellite-derived global precipitation products with a spatial resolution of at least 0.25° and temporal resolution of at least three hours. However, they are subject to biases and systematic errors caused by various reasons such as sampling frequency. It is recommended to check and evaluate, and/or bias correct these datasets before application, depending on your context.
CASE STUDY EXAMPLES
 See, for example, Eischeid JK, Pasteris PA, Diaz HF. 2000. Creating a serially complete, national daily time series of temperature and precipitation for the Western United States, Journal of Applied Meteorology, 39, 1580-1591.
 Ebert EE, Janowiak JE, Kidd C. 2007. Comparison of near-real-time precipitation estimates from satellite observations and numerical models. Bulletin of American Meteorological Society, 88, 47-64.