Title: Soil property estimation and design for agroecosystem management using hierarchical geospatial functional data models Authors
|Wikle, Christopher -|
|Holan, Scott -|
|Myers, D. Brenton -|
Submitted to: Journal of the Indian Society of Agricultural Statistics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 16, 2014
Publication Date: N/A
Interpretive Summary: Scientists in many fields are often confronted with analyzing and interpreting datasets containing a large number of related variables. An example is in proximal soil sensing, where spectrometers may be used to collect soil reflectance data at hundreds or even thousands of wavelengths. Relating those large datasets to parameters of interest requires advanced multivariate statistical techniques. A further complication is that the data are often collected at multiple locations or over time, leading to the need to account for spatial and/or temporal dependence in the analysis. In this research a new approach called multi-dimensional spatial functional models was developed to deal with these large spatial datasets. This approach provides for (1) estimating the error present in the model, (2) defining what wavelengths and wavelength ranges are important, thus reducing the complexity of the instrumentation required, and (3) describing an optimal design that reduces the amount of work required for future samplings. The new approach has potential for improved interpretation of large datasets such as those obtained in proximal soil sensing. This could enhance the utilization of these data by scientists and practitioners in the fields of precision agriculture and digital soil mapping.
Technical Abstract: Sustainable agriculture requires a site-specific approach to address crop management problems and environmental degradation processes that are spatially and temporally variable. These issues lead to production losses (water stress, low fertility, pest problems), soil degradation (erosion, soil organic carbon losses, compaction), and water quality degradation (sediment, nutrients, agrochemicals) - often at the sub-field scale. Management solutions must be implemented at the resolution of the problems; however, changes require information on the magnitude and extent of the issue. Unfortunately, landscape processes and properties can change at a finer spatial resolution than can be practically analyzed with lab methods due to time and cost of sampling and analysis. Thus, it is increasingly important to augment lab methods with field-sensor methods that can accurately characterize within-field variability at a more reasonable cost and with reliability and timeliness. These instruments can produce large data profiles and require calibration and prediction methods that can accommodate big data sets. We consider a functional spatial approach to perform calibration, spatial prediction, and design in this big data context. Specifically, using hierarchical Bayesian methodology we develop a signal/feature extraction approach for characterizing the visible and near-infrared (VNIR) spectroscopic wavelengths that are important predictors of cation exchange capacity (CEC) over space. This methodology is also used to develop optimal sampling locations to minimize the mean squared prediction error corresponding to a predicted spatial surface of this CEC response variable.