A seven-terabyte dataset that enables micro-level climate change predictions has been validated. A review describing and testing the dataset was published in Scientific Data, an open access publication of the journal Nature to describe scientifically significant datasets. You can view the forecasts based on the dataset and request your own by following the link.
The formation of the dataset was started in 2013 by the International Center for Tropical Agriculture (CIAT) and scientists from other associations, so that it was possible to foresee how this or that climate change will affect local agriculture and prepare the necessary organizational solutions. In tropical regions, there can be a huge difference between two neighboring areas, and forecasts for a nearby town will not help prepare for what awaits your village.
After the first testing, it became clear that this project can help not only small farmers. To date, the data from this project, which has undergone many additions and debugs, has been used in 350 scientific papers and is being extended to an even larger area of the Earth.
Examples of forecasts of precipitation (top) and average temperature (bottom) based on dataset / © Ramires-Villegas, Jarvis et al., Nature, 2020.
Climate models are typically built for scales of 70-400 kilometers. However, if you are concerned not only with global forecasts and want to understand what will happen to a specific point on the map, literally on your street, you will need corrected data. Global and regional models analyze climatic conditions more roughly and simplify natural processes. Their results can be very different from realistic scenarios.
The authors of the project not only scaled the forecast maps, but also adjusted the bias for clarification. Climatic probabilities are calculated for 436 probable scenarios. For the forecast, each scenario takes into account about two dozen different variables: for example, the average and maximum average monthly temperature and the amount of precipitation.
An example of scaling rainfall data. a) Initial data; b) Forecast for the next 30 years; c) Forecast for the occurrence of climatic anomalies; d) Centered forecast for climatic anomalies; e) Data using interpolation; f) Prediction data projected onto the surface with a spatial resolution of 30 arc seconds / © Ramires-Villegas, Jarvis et al., Nature, 2020.
First of all, the data set is needed for drawing up plans in agriculture. Clarification helps to avoid failing strategies on the ground. “Climate models show the earth system as a whole, but they are not perfect. These errors can affect our farming models. Because these models help decision-making, the consequences can be dire,”explains Julian Ramirez-Villegas, principal investigator of the project.
But the benefits are not limited to agriculture alone. The dataset has been used in a wide variety of areas: to map the potential global spread of the Zika virus, plan investment strategies for international development, and predict reductions in the number of outdoor skiing days in Canada. Forecasts of such accuracy can be useful in architecture and planning, in the protection of monuments of human history.