The Exascale Geo-Statistics project (ExaGeoStat) is a parallel high performance unified framework for computational geostatistics on manycore systems. The software project aims at optimizing the likelihood function for a given spatial data in order to provide an efficient way for predicting missing observations in the context of climate/weather forecasting applications. The framework has three main components:

1. Generate synthetic datasets in the case of testing mode. The synthetic dataset includes 2D spatial locations and measurements vector. The user can enable testing mode by using (–test) input. Furthermore, real datasets can be used with ExaGeoStat, when passing input files names for both spatial locations and measurements.
2. An iterative procedure is performed over the given data (i.e., synthetic or real dataset) to estimate the parameter vector the maximize the likelihood function for the given data. The average number of iterations depends on the lower and upper bounds of the optimization algorithm given by the user. Three computation options are available here, Exact, Diagonal Super Tile (DST), and Tile Low-Rank (TLR)
3. The estimated parameter vector (i.e., variance, smoothness, and range) of the Matern matrix kernel can be used to predict measurements on missing spatial locations. The ExaGeoStat's prediction stage is able to predict the measurements from the MLE parameter vectors. This stage is also running in two different modes: test and real. For testing, unobserved spatial locations are randomly chosen from the original spatial locations and their measurements are marked as unknown. In this case, ExaGeoStat is able to compute the Mean Square Error (MSE) by comparing observed measurements with predicted measurements. For real cases, the user provides the program with the new spatial locations and the program predict the measurements for those new locations.