The following is a summary of the paper by Brendan Malone which is nominated for best paper in Pedometrics 2013.
A remarkable growth in the application of Digital Soil Mapping (DSM) is currently being experienced around the world. It is being used to address important environmental issues over a range of spatial extents — fields and farms, landscapes and regions, countries and continents, and importantly, globally. There is potentially a significant amount of comprehensive spatial soil information throughout the world. Yet, what we have recognised, and what could potentially be an operational hurdle further into the future, is the inequality between the ‘scale’ of the digital soil information which is available, and that which is required to address a particular issue or question. For example, a soil organic carbon map produced at the national extent (perhaps for a national soil carbon accounting purpose), may be inappropriate at the field extent. Alternatively, existing digital soil information may be available for points, but is required over small areas i.e. each prediction represents a integral of the soil attribute of interest across the spatial dimensions of the area. The problems may be addressed through implementation of spatial scaling methods. Our paper examines this issue and provides a broad overview of spatial scale concepts and spatial scaling procedures that are specifically relevant for DSM.
First, we clarify some fundamental concepts of scale. Essentially, digital soil maps have three spatial scale entities: extent, resolution, and support. Map extent is the areal expanse or coverage of a mapping domain. Resolution is the grid-cell spacing or pixel size of the raster. While support is likened to a volume or area.
Secondly we set about describing a suite of pedometric techniques that could be used for spatial scaling. One may think of these as upscaling or downscaling methods, where upscaling may involve an increase in extent, support, or grid cell resolution size, which could either be modified conjunctively or just focusing on one or two spatial entities only. Downscaling is essentially the opposite process of upscaling. However, we describe spatial scaling for DSM with due reference to the scale entities (extent, resolution, support). Such that fine-gridding, deconvolution, and disseveration are different variants of downscaling operations. While coarse-gridding, convolution, and conflation are variants of upscaling operations. We provide some of the theory of each of these operations, and provide examples of their usage either from the literature or from our own data.
One important discovery from this paper is that we can efficiently, without changing the grid cell resolution; obtain areal or block averages from point support maps. This is likened to going from a source digital soil map with 20m point support to a destination map where the predictions represent averages on supports of 20m × 20m. This can be achieved with block kriging. Another useful application of spatial scaling for DSM is the downscaling from coarsely resolved to finely resolved maps.
We feel we have thrown light on the issue, provided some new solutions, and are very optimistic about future developments which inevitably will help to solve pressing soil and environmental issues around the world.
Malone, B.P., McBratney, A.B., Minasny, B., 2013. Spatial Scaling for Digital Soil Mapping. Soil Science Society of America Journal 77, 890-902. (The article is OPEN ACCESS until end of 2014)