Optimizing Resource Management Under Spatial Uncertainty
Investigators (most current known information)
The wise stewardship of natural resources is a complex task, requiring the coherent integration of data and associated uncertainty. This is especially true with the advent of agricultural precision technologies, and the associated data gathering and interpretation tasks. Any spatial map has data uncertainty, if this is neglected in ecological management decisions, a sub-optimal allocation of resources will occur. The purpose of this research was to develop methods to improve map accuracy as well as provide guidelines to assist managers in making informed decisions to reduce and manage spatial uncertainty. Several methods were employed to accomplish these objectives, including research to quantify, and reduce the impact of, sources of uncertainty with respect to classification schemes for remotely sensed imagery. Models were developed to quantify changes in accuracy associated with data aggregation, enabling assessment of alternative aggregation schemes, to correct topographic effects on vegetation classification for aerial photographs, and to identify spatial-temporal patterns of classification errors. Further work evaluated alternative interpolation schemes relative to data skew and data sparsity, and also developed and evaluated the economic benefit associated with incorporating spatial uncertainty assessments in variable rate soil fertility management. In this work, non-linear loss/profit functions were combined with local estimates of spatial uncertainty in P and K to identify optimal fertilizer rates. Over an 8-year horizon in a corn-soybean rotation, results indicate a net savings of approximately $14/ha for P, and $9/ha for K, using the uncertainty model as opposed to traditional spatial interpolation. In related efforts, "optimal" soil sampling strategies were developed that concurrently minimize the maximum kriging estimation variance as well as the spatial variability in kriging variance. For an arbitrary grid scale with isotropic spatial structure, a hexagonal compositing scheme with spacing congruent with the grid scale provides the "optimal" sampling strategy, in terms of variogram estimation and accuracy of spatial estimation.
Articles in Journals
Hedoin, E., R. H. Hornbaker and T. R. Ellsworth. 2005. "Optimizing P and K fertilizer management under models of spatial uncertainty." Journal of Precision Agriculture (in review).
Svoray, T. and Y. Carmel. 2005. "Empirical method for topographic correction in aerial photographs." (in preparation).
Carmel, Y. 2004. "Characterizing location and classification error patterns in time-series thematic maps." IEEE Geoscience and Remote Sensing Letters 1(1):11-14.
Carmel, Y. 2004. Controlling data uncertainty via aggregation in remotely sensed data. IEEE Geoscience and Remote Sensing Letters 1(2):39-41.
Ellsworth, T. R., E. M. El-Naggar, T. Smith and R. L. Mulvaney. 2004. "Optimal sampling strategies for soil amino sugar nitrogen." Soil Science Society of America Annual Meeting Abstracts No. 6113.
Ellsworth, T. R., P. M. Reed and R. J. Hudson. 2003. "An evaluation of interpolation methods for local estimation of solute concentration." In Scaling methods in soil physics, eds. Pachepsky et al, 143-162. CRC Press.