Climate Data Types

Weather radar image from West Africa. From SAT 24.

The generation of climate data at local and national levels is typically the responsibility of a country’s national meteorological service (NMS). NMS are mandated to generate weather and climate data and issue forecasts and warnings. Generating the data requires a functioning, well-maintained and well-distributed network of stations as well as the capacity to store and analyze the data and use it to model future conditions. Through strong climate data and information and analytical capacity, a country can see improvements in its adaptive capacity as well as its capacity to manage water resources, food security and disaster risks.. More and more data can also be accessed on line and it is important to understand the advantages and limitations of each type of data, and select data and information that is the best compromise between intended use and ease of access.

Historical and Current Data


Station data capture local conditions best as they are in situ (“in position”) measurements of meteorological variables. Station data, or information

Weather Station

based on such data, can be obtained from the National Meteorological Services (NMS) but is not always free of charge. In addition, in most of African countries spatial (and often temporal) coverage of such data may not be sufficient to meaningfully capture local details important for adaptation projects. In a lot of countries a systematic drop in the number of operating and reporting stations has been observed since the 1980s (Figure 3.1). The quality of the data may also vary and quality checks are always necessary.  Additional data may have been collected by businesses, farmers’ associations, agricultural extension services and academic institutions but may not be integrated in the main database. A subset of station data is available on line through the GTS system but the completeness of that data and their spatial and temporal coverage are usually less extensive than the data archived at the NMS.

Figure 3.1 Number of rainfall stations operating in Mali since early 20th Century. Source: Samake (AGRHYMET)


Since the early 1980s satellites have been recording some meteorological variables at high spatial and temporal resolution and these archives are now long enough for climatic analyses. Satellite-based estimates of rainfall and temperature can be accessed online in different data repositories and have continuous spatial coverage in the form of a grid of values. They are usually global in nature. However, those data rely on indirect measurements (proxies) of rainfall quantities and temperature, and thus can  present a biased vision of the conditions on the ground. For example, satellite rainfall estimates are based on the temperature at the top of the clouds, which is then correlated to temperature and rainfall records from stations on the ground. In the tropics, this relationship is strongly affected by the availability of in-situ data, while in the extratropics estimates are closer to reality.
Rainfall and temperature estimated from satellite recordings are available from the following datasets:

  • CPC Morphing Technique (CMORPH) – a high spatial and temporal resolution dataset in real time. Only available since 2002.
  • Tropical Rainfall Measurement Mission (TRMM) focuses on rainfall estimates in the Tropics and is available since 1998. There is however a delay of about a month in data availability and monthly aggregates are more accurate than higher temporal resolution products.
  • Land Surface Temperature (LST) provides land-surface estimates for Africa and Latin America and is available since July 2002 for Africa.

In addition to meteorological variables, satellites also capture useful indicators such as vegetation cover with the Global Normalized Difference Vegetation Index (NDVI), available for 1981-2006 period and the TERRA-MODIS NDVI and Enhanced Vegetation Index (EVI) since 2000. More recently, soil moisture data have also become available.  When using satellite rainfall data it is important to remember that precipitation estimates have different validity in different regions, and thus need to be validated with in-situ observations for local applications.


Another class of data providing continuous spatial coverage are so called ‘gridded data.’ From a user’s perspective they are similar to satellite data in that data are available on a regular grid and for a given period of time. Most of these datasets are based on in-situ (station) information interpolated to a regular grid. Some datasets merge in-situ observations with satellite records to compensate for biases in satellite data. The quality of these merged products depends on the quantity of the in-situ data incorporated. Gridded datasets are usually global in nature and their spatial resolution is variable. Among the gridded datasets are:

Another class of gridded data is called ‘reanalysis’ data, which uses observations from sources at the surface and  higher in the atmosphere (planes, radio soundings, etc.) to interpolate meteorological variables using physical models,  ensuring stronger consistency between variables and values. It provides a wider array of meteorological variables, including atmospheric measurements above the surface.  Reanalysis products are usually high-resolution and date further back in time than other data types. However, this method relies strongly on dynamical models and some variable types are more reliable than others; for example, wind and pressure are fairly accurately captured through reanalysis while rainfall data are often biased. It is most useful for researchers studying the causes of climatic phenomena rather than for direct applications in adaptation.

Forecasts and Projections

Weather forecasts are the most widely known class of future information and are usually provided by the NMS. They are used in EWS and other short term (a few days) outlooks. In the last two decades additional forecast systems have been implemented by the WMO to make predictions over  longer time scales. The most advanced systems currently available provides seasonal climate forecasts in the tropics.


Seasonal forecasts take advantage of the fact that Sea Surface Temperatures (SSTs) influence moisture and atmospheric circulation in the tropics, but tend to evolve more slowly than the atmosphere by several months.As a result, SSTs can be used to predict the overall state of the atmosphere in a given region/climatic system in the coming months. A number of regions have been running seasonal forecasts over past decades and, in Africa, most notable seasonal forecasts are elaborated during the Climate Outlook Fora organized by ACMAD and ICPAC. During such fora the general tendencies of the season are predicted and presented in probabilistic format (Figure 3.2). These predictions can be helpful in planning water resource and on-farm management at national, subnational and individual levels.

Figure 3.2 Probabilistic seasonal forecast for the June-August 2017 season in West Africa issued by ACMAD.


A wide range of regional and international initiatives allow access to climate change projection data. Some of the portals giving access to projections are listed below:
However, to make best use of the available data and information, the following needs to be considered:

  • Scope of the study – how much climate information is really needed. E.g., general tendencies vs. precise high resolution data for further impact modeling
  • Accuracy of climate models over the region of interest – at local scale GCMs can present strong biases (e.g., precipitation can be strongly underestimated and using direct raw data may lead to biased conclusions with respect to future conditions); a thorough literature review is needed to assess such biases and eventually correct them
  • Uncertainty – is usually assessed through the use of a variety of models and assessment of minimum vs. maximum changes projected
  • Documentation about data source – sufficient explanations with respect to how the data were processed; this is particularly important when working with downscaled data.


Although GCMs are valuable predictive tools, they cannot account for fine-scale heterogeneity of climate variability and change due to their coarse

Figure 3.3 Many of the processes that control local climate, e.g. topography, vegetation, and hydrology, are not included in coarse-resolution GCMs. The development of statistical relationships between the local and large scales may include some of these processes implicitly. Source: Viner, 2012

resolution. Numerous landscape features such as mountains, water bodies, infrastructure, land-cover characteristics, and components of the climate system such as convective clouds and coastal breezes, have scales that are much finer than 100–500 kilometers. Such heterogeneities are important for decision makers who require information on potential impacts on crop production, hydrology, species distribution, etc. at scales of 10–50 kilometers.
Various methods have been developed to bridge the gap between what GCMs can deliver and what stakeholders require for decision-making. The derivation of fine-scale climate information is based on the assumption that the local climate is conditioned by interactions between large-scale atmospheric characteristics (circulation, temperature, moisture, etc.) and local features (water bodies, mountain ranges, land surface properties, etc.). It is possible to model these interactions and establish relationships between present-day local climate and atmospheric conditions through the downscaling process. It is important to understand that the downscaling process adds information to coarse GCM outputs so that information is more realistic at a finer scale, capturing sub-grid scale contrasts and inhomogeneities. Figure 3.3 presents a visual representation of the concept of downscaling.

Downscaling can be used to increase spatial or temporal resolution. There are two principal downscaling methods:

  • Dynamical: Similar to GCMs, this method involves incorporating additional data and physical processes in regional- or local-scale models, at a much higher resolution than is seen in GCMs. This method has numerous advantages but is computationally intensive and requires large volumes of data and a high level of expertise to implement and interpret. The resources required place this method beyond the capacities of most institutions in developing countries.
  • Statistical: This method involves establishing statistical relationships between large-scale GCM-modeled climate features and local climate characteristics. In contrast to the dynamical method, statistical methods are easy to implement and interpret. They require minimal computing resources but rely heavily on historical climate observations and the assumption that currently observed relationships will carry into the future. However, high quality historical records often are not available in developing countries.

The diversity of existing downscaling methods reflects the diversity of goals of and resources for each downscaling exercise. Thus, there is no single best downscaling approach, and downscaling methods will depend on the desired spatial and temporal resolution as well as the impacts being assessed. In most cases, a sequence of different methods is needed to obtain results at the desired resolution.
When working with downscaled information the following needs to be kept in mind:

  • Information on downscaling and the limitations of the results need to be understood and taken into account. Any results at resolutions higher than the native resolution of the GCMs has undergone downscaling and may mislead the user that the results are automatically more accurate because of their finer scale.
  • Downscaling processes usually adds uncertainty to projections. Additional uncertainty should be factored into the interpretation of the results or indeed into the decision of whether downscaling is appropriate.
  • Downscaling methods should be validated first on historical data to assess their potential biases and uncertainties.

In summary, downscaling is not a trivial process and requires some expertise. It may require significant investment in time, tools and capacity. The user should not believe that the results are automatically more true or valid because they are presented at higher resolution. The need for downscaling and the tradeoff between higher-resolution and increased uncertainty need to be carefully assessed.

« Sectoral Information Needs
Climate Variables »