Spatial statistics has made a noteworthy impact on applied work in the 21st century. Increasingly, researchers incorporate spatial referencing in analyzing data. However, with the substantial rush to play with geostatistical models, Markov random field models, and spatial point pattern models, the low hanging fruit is gone. For the future of the field, new challenges will emerge from theory, methodology, computation, and application. I will focus on application. I will first emphasize critical issues that distinguish such application from the world of big data and machine learning. Next, I will consider an illustrative challenge concerning the inter-relationships among data fusion, preferential sampling, and marked point patterns, with care needed to put these pieces together. Then, I will describe three demanding applications which currently stretch available tools. These involve spatial joint species distribution modeling over large numbers of species and locations, space-time behavior of insurgent and intervention events over several years in Iraq, and dynamic functional activity (from f-MRI data) in terms of blood-oxygen level across brain regions over time.