Methodology
This methodology explains how we combined climate data with real-world coffee farming data from the International Multilocation Variety trial (IMLVT) to help you identify varieties that are well adapted to your local climate conditions. Additionally, by combining knowledge on where a variety performs well today and projections on a location’s future climate, the tool also supports you in deciding how future-proof your current choice of varieties is likely to be, based on locations whose current climate resemble your projected future climate.
Climate data: understanding your Environment
The climate data used in CafeClima is powered by ACLIMATAR, a dedicated climate data engine with crop-specific climate projections (tailored for coffee, but also adaptable to cocoa and tea). ACLIMATAR is designed as a reusable infrastructure and can support similar applications and analytical use cases beyond CafeClima, subject to collaboration and access agreements. For questions and potential collaborations, please reach out to: Alliance-ACLIMATAR@cgiar.org
1.1 Climate analysis
The climate analysis combines historical climate observations, near-term trends (‘nearcasts’), and future projections to describe past, present, and potential future climatic conditions at high spatial resolution (5 × 5 km). This resolution allows interpretation at municipality or district level.
The analysis is based on globally gridded climate datasets (downscaled and biased corrected) for precipitation, temperature, and bioclimatic variables, complemented by climate model projections. Historical and near-term climate conditions were derived from observed datasets, while future conditions were estimated using downscaled and bias-corrected Global Climate Model outputs. Solar radiation and potential evapotranspiration (PET) were included to identify wet and dry seasons.
Observed precipitation and temperature patterns were derived from CHIRPS and CHIRTS. These datasets were used both to describe past climate conditions and to calculate PET-based seasonality indicators. Climate risk factors relevant to coffee and human health were analyzed using AdaptationAtlas data by comparing a historical reference period with a mid-century future period and then recalculated based on known coffee distribution in the local country, to derive coffee-tailored risk thresholds.
Bioclimatic variables from CHIRTS, CHIRPS and WorldClim also provided the reference framework for defining agro-climatic suitability under current conditions. Future climate projections were primarily based on CMIP6 (high-emissions scenarios). Note that different datasets use different future time horizons, reflecting the structure of the underlying models - this is why we commonly refer to ‘past’, ‘current’ and ‘future’ time horizons in the platform. In this context, ‘future’ can be seen as to be equivalent as mid-century (2050s).
Climate data sources and timeframes (overview)
| Data source | Variable / used | Historical period | Future / projection period | Climate / emissions scenario |
|---|---|---|---|---|
| CHIRPS | Precipitation | 1990–2022 | 2026–2055 | CMIP6 |
| CHIRTS | Temperature | 1995–2014 | 2026–2055 | CMIP6 |
| WorldClim | Bioclimatic variable | 1970–2000 | 2020–2049 / 2040–2069 | CMIP5 / CMIP6 |
| AdaptationAtlas | Climate risk indicators | 1995–2014 | 2026–2055 | High emissions |
| Global Aridity & PET Database | Solar radiation, PET | Climatology | — | Observed |
| GCMs (CMIP5/6) | Climate projection | — | Mid-21st century | RCP 6.0 / SSP 3–7.0 / SSP 5–8.5 |
1.2 Agroclimatic zones and adaptation zones
For coffee, typical Arabica and Robusta growing climates (agro-climatic zones (ACZs)) were distinguished based on a global analysis of the known (global) distribution of coffee growing regions and the bioclimatic conditions observed at those locations. Each ACZ represents a set of climatic conditions that are considered suitable for cultivation. While all ACZs are suitable, they differ in their temperature and precipitation characteristics (bioclimatic variables).
In selected countries (El Salvador, Guatemala, Honduras, Nicaragua, and Peru), finer-grained and locally validated ACZs are available. These classifications, however, are designed for within-country comparison only; for example, a zone described as “hot and dry” reflects conditions relative to other coffee-growing regions in that country.
By projecting these ACZs into the future and comparing current and future spatial distributions, the magnitude of climatic change was assessed and translated into an Impact gradient. Three Impact gradient categories are distinguished, reflecting increasing levels of change:
Incremental adaptation zones
Where agro-climatic conditions are projected to remain stable and suitability is maintained.
Systemic adaptation zones
Where shifts between different suitable ACZs are expected, but conditions projected to remain within the known climatic range for coffee farming. Such a change means adjusting to a new growing climate but with overall suitability maintained. Diversifying may help to reduce risk.
Transformational adaptation zones
Where conditions are projected to move beyond climates currently known as suitable for coffee farming, indicating a potentially high risk of productivity loss. Here, more substantial adaptation efforts will be required, with diversification strongly required and maybe even shifting to alternative crops.
Important Note: The data shown is climate, not weather data. What does this mean?
Climate data is the mean climate calculated over a time period of 30 years, whilst weather is the current state of the atmosphere. Think of it like climate is what you have grown used to since you were a child, and weather is what you see when looking at the window right now.
We are using interpolated global climate data which is the best data publicy available for global coverage, and which combines weather data from weather stations on the ground and in the regions with satellite observations. Weather stations allow for relatively high precision, however, they are oftentimes placed kilometers apart, and especially historic records may be hard to obtain or non-existent, when stations have only been established in recent decades. Thus, data between these observation points needs to be calculated from mean values and using elaborated formulas that take into account e.g. local relief. This data, however, will never be as precise as the weather and climate data measured by your local weather station directly on the field.
Regarding the ‘future data’:
As is the case of any future outlook, the climate models have a considerable degree of uncertainty and should be considered projections, not predictions.
It is also important to note that the Global Climate Models currently available show a high level of agreement on an increase in temperature, but disagreement about the regional and seasonal distribution of precipitation. The resulting consensus model of the independent projections is therefore, to a large degree, influenced by the temperature increase, while precipitation disagreement is masked.
Please Visit https://adaptation.aclimatar.org/ for more detailed information on our approach to crop-specific climate data analysis, the full methodology and climate adaptation advice. You might also want to take our learning open-access learning course: Decision-making in a changing climate: From analysis to guided adaptation action in cocoa, coffee and tea farming systems.
Variety Trial Data: The IMLVT Approach
To understand how different coffee varieties behave, we use data from the International Multi-Location Variety Trial (IMLVT). This global effort, started in 2015, tests 31 different Arabica coffee varieties across 29 sites in 18 countries, including extreme environments with high heat and long droughts. Of these 31 coffee varieties, we present data for 26 Arabica varieties on the platform (AB3, Batian, BLP10, Catigua MG2, Catuai, MundoMex (EC15), Mundo Maya (EC16), Geisha, Centroamericano (H1), IPR 103, IPR 107, K7, Kartika 1, Lempira, Marsellesa, Mundo Novo, Oro Azteca, Pacamara, Parainema, Paraiso, Ruiru 11, S795, S4808, SL28, sln.5B, sln.6).
IMLVT used a randomly blocked fisher design to evaluate varieties across the globe. Please visit our dedicated section on the International Multilocation Variety Trial (IMLVT) or the official WCR IMLVT page for more detailed information on the trial design and participating institutions.
Overall, data from 20 trial sites in 13 different countries can be accessed through the platform.
2.1 Variety performance analysis
Field trial data was analyzed using two-stage analysis, that first consists of single trial site analysis, followed by a join analysis of all trial sites. Single trial site analysis allows to locally characterize how varieties respond to a specific location’s conditions such as soil, local management (e.g., P&D control) during the trial years. By comparing varieties side-by-side in one place, we can identify which ones yield the most or resist diseases like coffee leaf rust under those specific conditions.
Multi environment trial (MET) analysis was then performed, feeding from the results of the single-site analysis. We used linear mixed models (LMM) with a Factor Analysis approximation of the variance-covariance matrix. The MET analysis allows to jointly analyze all trial sites, while ensuring that the noise level of each trial is factored in the analysis, for objective performance evaluation of the varieties.
The output of the MET analysis is two-fold: a general estimate of the performance and stability of each variety (how consistently a genotype performed across diverse environments), and adjusted estimate of each variety performance for each trial site.
These results allow to characterize with minimum noise the:
- Stable varieties: Those that give a reliable yield across many different climates, such as Oro Azteca or Lempira.
- Specific adaptations: Varieties that might fail in some places but perform exceptionally well in others, such as certain hybrids that thrive in stable temperatures but struggle in areas with high seasonality.
2.2 Analog Identification and Similarity Calculation
The goal of “analog identification” is to find locations that are “climate twins” to our trial sites. If a variety performs well at a trial site, we expect its response to the climatic component to be similar for its climate analogs. This helps us to support variety decision making under current climate, but also future projections regarding what varieties are expected to also perform well under projected future climate at a given location.
A Matching Analogs to Current Climate
First, using bioclimatic indicators (i.e. relevant climate measures for coffee growing) we created a climate profile for each trial site, based on the weather observed during the actual trial period and the overall climate for that location. We then searched the globe for other locations that under current climate have a matching profile, giving extra weight to temperature because of its high impact on Arabica coffee.
We are only showing analogs with at least 99% similarity, i.e. analogs whose climatic characteristics are within the top 1% of climatically most similar location to the location of interest. Climate analog sites are ranked from highest to lowest similarity to your location. A similarity degree greater than 99.5% means very high, greater than 99% means acceptable.
For each of the analogs, users can then access the sites’ trial data, showing the relative performance of each variety. The relative view makes it easier to compare variety performance at sites that are climatically similar to a users’ location, supporting them in making trade-off decisions about which variety to plant. Absolute yield values (e.g., cherries per hectare) are not given because these are highly variable and dependent on management practices.
Please note: We consistently use the best available, fit-for-purpose climate data for each analysis. Therefore, climate diagrams might slightly differ between sections. ACLIMATAR climate diagrams rely on CHIRPS/CHIRTS for precipitation and temperature, while the analog matching uses agro-climatic variables from WorldClim v2.1. Both datasets are widely used, scientifically robust, and represent best practice for global-scale climate analysis.
B Matching Analogs to Future Climate (“Space for Time”)
Trial sites that are similar to climate zones of a location in current times will likely not be the same sites projected to be similar in the future. Since we can’t time travel and to help with forward-looking planning we apply space-for-time substitution. This means, we identify those trial sites that are most similar to the future climate of a user’s location.
And by then showing the variety performance results of these future climate twins, we make use of the real-world performance data that we have right now (from the trial sites) to show high performing varieties. By finding these future twins, we can give farmers an early indication of which varieties are expected to be more resilient as their local climate changes.
Important note: While climate analogs are a powerful tool, they focus on long-term averages. These, as any future climate projection, carry inherent uncertainties (e.g. regarding future emission pathways). Additionally, other factors like local soil properties, new pests, or extreme weather events also play a major role in a variety's success and should be considered alongside this data.
The analog matching in CafeClima is based purely on climatic variables. While the tool also displays information on soil types, these are provided for illustrative purposes only and are not used in the calculation of climate analogs. As such, the platform is intended to show potential variety performance under similar climatic conditions, rather than to provide definitive predictions.
Users should confirm any insights with on-site trials and take into account additional local factors, including soil properties, microclimate variations, production objectives, and the availability of seed at the national or regional level. The tool is best used as an exploratory resource and a starting point for discussion, helping users identify promising varieties for further evaluation rather than as the sole basis for planting decisions.
Please visit our ‘Next steps’ section to find additional information on other factors to consider when choosing a variety and planning renovation or planting.