Speaker: Jarad Niemi, Iowa State University, "Advancing scientific practice through agricultural statistics"
Abstract: This talk will be broken into two parts: 1) my applied work and its impact in agriculture policy and 2) development of statistical emulators for agricultural computer models. In my role as the Data Team lead for the Prairie STRIPS project, I led the statistical analysis of more than a decade worth of data from dozens of PIs on the effect of prairie strips planted within row-crop agriculture. This work led to a PNAS manuscript that provided the foundation for inclusion of prairie strips within row-crop agriculture as a conservation practice in the Conservation Reserve Program within the United States Department of Agriculture. Currently I am turning my attention to the development of renewable natural gas from manure and herbaceous biomass through anaerobic digestion.
One huge externality of current agricultural practice is the loss of soil and nutrients into waterways. The water erosion prediction project (WEPP) is a computer model that utilizes slope, soil type, weather, and land management to predict soil loss. We are developing Gaussian Process (GP) emulators of the WEPP computer model in order to expand daily, watershed-based, soil loss estimates from Iowa to the world. We have developed a one-at-a-time knot selection procedure to overcome the analytical intractability of GPs when you have more than a few thousand observations. We are also actively developing dynamic automatic relevance determination methodology for functional inputs. Combining these and other developments into a WEPP emulator will allow us to design efficient computational experiments and generate daily estimates of worldwide soil loss.