Spatial database is the cornerstone of any GIS. It is designed for data with spatial attributes. With the tremendous improvement in several technologies for instance: sensors and internet of things, the size of spatial data in general and environmental data in particular is growing quickly. The extraction of relevant information via spatial queries requires searching through this huge volume of data, which is time consuming. Therefore, there is a growing need for better performance. Unfortunately, existing algorithms and methods are based on traditional computing framework (uniprocessors) which makes them incapable to deal with large scale data and as a result we get poor performance. In this work, we report our designs, implementation and experiments of large scale spatial raster selection query based on the modern high performance GPUs. Our aim is to speed up the underlying query execution using parallel primitives and pure CUDA. Our test results show that our methods are faster and good performance is achieved even with large scale rasters and large scale dataset.