Recently I was re-reading a paper on predictive soil mapping (Park et al, 2001), and considered testing one of their proposed terrain attributes in GRASS. The attribute, originally described by Blaszczynski (1997), is the distance-weighted mean difference in elevation applied to an n-by-n window of cells:

Equation 4 from (Park et al, 2001)

 
where n is the number of cells within an (odd-number dimension) square window excluding the central cell, z is the elevation at the central cell, z_{i} is the elevation at one of the surrounding cells i, d_{i} is the horizontal distance between the central cell and surrounding cell i. I wasn't able to get a quick answer using r.neighbors or r.mfilter, so I cooked up a simple R function to produce a solution using r.mapcalc. The results are compared with the source DEM below; concave regions are blue-ish, convex regions are red-ish. The magnitude and range are almost identical to mean curvature derived from v.surf.rst, with a Pearson's correlation coefficient of 0.99. I think that it would be of general interest to add functionality to r.neighbors so that it could perform distance-weighted versions of commonly used focal functions.

Elevation surface (left) and resulting mean curvature estimate (right)

The rnorm() function in R is a convenient way to simulate values from the normal distribution, characterized by a given mean and standard deviation. I hadn't previously used the associated commands dnorm() (normal density function), pnorm() (cumulative distribution function), and qnorm() (quantile function) before-- so I made a simple demo. The *norm functions generate results based on a well-behaved normal distribution, while the corresponding functions density(), ecdf(), and quantile() compute empirical values. The following example could be extended to graphically describe departures from normality (or some other distribution-- see rt(), runif(), rcauchy() etc.) in a data set.

Major updates to the SoilWeb iPhone Application.

I was recently asked to review a soon to be published book on PostGIS, a spatial extension to the very popular Postgresql relational database. I was very excited about receiving an early copy of this book, as the authors have provided countless tips, fixes, and clever query examples on the PostGIS mailing list over the years. After spending a couple weeks looking through the book, I have to say that I am very impressed with the quality and completeness. Indeed, this is the book that I wish would have been available when I was starting out with PostGIS. The authors do an excellent job of promoting the idea that a relational database and SQL are well suited for spatial data modeling and analysis.

Western Fresno Soil Hierarchy: partial view of the hierarchy within the US Soil Taxonomic system

Maybe this is just craziness, but wouldn't be neat to have an XML formatted version of the Keys to Soil Taxonomy? The format might look something like the following code snippet, although there may be more efficient uses of XML... The only problem I can see is that it would take a hell of a long time to type in the entire 300+ page document. A complete document of this nature would support all kinds of new and creative uses for the 'keys-- electronic look-up, automated generation of a PDA-ready version, an awesome teaching tool, or just something that could be used to generate cool figures. Anyone know of a quick way to get this put together, or of any similar document that has already been published? Anyone want to help type-in the data?

 
Parent material data is stored within the copm and copmgrp tables. The copm table can be linked to the copmgrp table via the 'copmgrpkey' field, and the copmgrp table can be linked to the component table via the 'cokey' field. The following queries illustrate these table relationships, and show one possible strategy for extracting the parent material information associated with the largest component of each map unit.

 
Several of the example queries are based on this map unit:

Finally made it out to Folsom Lake for a fine day of sailing and GPS track collecting. Once I was back in the lab, I downloaded the track data with gpsbabel, and was ready to import the data into GRASS.

I was particularly interested in how fast we were able to get the boat going, however my GPS does not keep track of its speed in the track log. Also, I had forgotten to set GPS up for a constant time interval between track points. Dang. In order to compute a velocity between sequential points from the track log I would need to first do two things: 1) convert the geographic coordinates into projected coordinates, and 2) compute the associated time interval between points.

bwplot annotation example

Sometimes you want to add a little text to box and whisker plots produced by the lattice function bwplot(). Here is one approach. Could be optimized a bit more to reduce manual specification of some elements. Suggestions welcomed.

Just read about a new R package called alphahull (paper) that sounds like it might be a good candidate for addressing this request regarding concave hulls. Below are some notes on computing alpha-shapes and alpha-hulls from spatial data and converting the results returned by ashape() and ahull() into SP-class objects. Note that the functions are attached at the bottom of the page. Be sure to read the license for the alphahull package if you plan to use it in your work.

Alpha-Shape Example

 
Missing Data
Soil scientists routinely sample, characterize, and summarize patterns in soil properties in space, with depth, and through time. Invariably, some samples will be lost or sufficient funds required for complete characterization can run out. In these cases the scientist is left with a data table that contains holes (so to speak) in the rows/columns that are missing data. If the data are used within a regression, missing values in any of the predictor or the response variable result in row-wise deletion-- even if 9/10 variables are present. Furthermore, common multivariate methods (PCA, RDA, dissimilarity metrics, etc.) cannot effectively deal with missing data. The scientist is left with a couple options: 1) row-wise deletion of cases missing any variable, 2) re-sampling or re-characterizing the missing samples, or 3) estimating the missing values from other variables in the dataset. This last option is called missing data imputation. This is a broad topic with countless books and scientific papers written about it. Here is a fairly simple introduction to the topic of imputation. Fortunately for us non-experts, there is an excellent function (aregImpute()) in the Hmisc package for R.