A framework for analysis and benchmarking of heterogeneous datasets: A grassland soil example

Abstract:

Land-atmosphere exchanges are extremely heterogeneous in time and space, yet measurements and models are limited in spatiotemporal resolution. As land-atmosphere models become more complex, computers more powerful, and field observations more numerous, there is an opportunity and a need to develop an experimental design for heterogeneous field data collection and to build scalable networks for analysis and benchmarking of such datasets. We are developing a standardized system to measure, analyze, and benchmark heterogeneous land-atmosphere exchanges at various spatiotemporal resolutions. We present results from our first testbed where we measure soil moisture, soil temperature, and soil respiration within a 100 x 100 m grassland area at 64 locations in a grid system. Soil moisture and temperature are measured once every 30 minutes at all locations and soil respiration at this same temporal resolution at 4 locations within this grid and at 14 days resolution in all locations in the grid. We developed a web-based interactive data visualization for observation of the rapid changes that occur in time and space, that would allow to test the performance of different models to reproduce observed temporal and spatial patterns (e.g. https://simondanisch.github.io/WGLDemos/soil/, using the Julia language and the Makie.jl package). The objective of this testbed and data analysis framework is to be compatible across spatiotemporal scales and land-atmosphere pools, exchanges, and properties measurements. We intend for the framework to be available publicly and improve over time (e.g. adding models and benchmark metrics), as we include routes for adding more data streams from soil and atmospheric instruments.