Improve VR-Seed by Subdividing SSURGO data by Slope

Farmers often use SSURGO soil type data to assign seed rates.  Their main assumption relates to water holding capacity and organic matter as these macro soil attributes greatly influence ability to support higher populations.  However there are some limitations to this public data set.  They blame various things, mostly related to resolution (aka "the darn lines are in the wrong place.")

When frustrated, farmers sometimes ask for normalized historic yield maps to improve or replace SSURGO data.  However, when we’ve done this we often encounter resistance when they finally get to the field.   Why?  Because it ends up assigning low rates to “good soil” that they have tiled or otherwise improved.  Often times the yield data has problems, like switching varieties, missing data in middle of a field or only a few years of “good data.”  We can over exaggerate underlying soils rates but all this adds up to conceptually liking the idea but practically avoiding it given lots of poor quality yield data.  Can you say, tail wagging the dog?!

A long time goal of ours was to better describe predicted soil moisture using slope.  Despite the ease at which one can look and perceive slope from an elevation map it is quite challenging to mathematically capture it in most existing GIS software.  Well, we finally accomplished just that… both max-grade and position on hill… top or bottom portion of the hill itself.

Above is a corn yield map from drought stricken 2012 field.  As you can clearly see, the yields don’t change exactly with the specific soil type zones change.  However, it's decent, albeit, not perfect.  Where "off", is that because they are wrong or because a B slope soil type is simply too great a generalization for this application?   Said another way, B slope simply means gently rolling, but doesn’t say “this is a hill top.”  Sometimes it is, but often it has hill tops and valleys both, which yield differently.   Given the tabular data quality of the SSURGO maps (and their ease of retrieval,) we’re not convinced we should abandon them just yet.  Instead, how about we improve it?

Below is a VR seed map for corn with original SSURGO map overlaid.  You can see how the slope extends the normal zone rates across various soil types.  To accomplish this we started with a VR-Seed map based on soil type and then used some logarithmic math to distribute based on a proprietary slope algorithm.  The lower position in the hill (at same slope) gets more seed than does the same slope toward the hill top.  In fact, the math effectively describes the concave or convex proprieties of all “hills and valleys” at the given 100 ft square resolution.

Additionally, this VR-Seed file means that the farmer won’t have those hard polygon transitions while traversing the field.  In fact, the rates will gradually shift higher or lower as he transitions between soils and varying slopes.  Same soil type but in a valley?  No problem… we’ll drop more seed.  Same soil but on the top of the hill?  No problem, we are dialing back.  This elevation influenced data is able to be accomplished even with WAAS data because of the relative nature of slope position and 100 ft horizontal resolution.    The ultimate result is a refined seed rate without a crop consultant even visiting the field.

The final map is our proposed VR-Seed map for soybeans.  As you might notice we invert the rates to lower population to prevent lodging in the appropriate positions in valley of high OM soil types.  At this point, we believe a good farmer could “trump” these maps with his knowledge about drainage, fertility or other oddities not represented in these two data sets that make his soils different than the underlying SSURGO properties.  (or, subdivide his SSURGO using smart zone soil sampling first and then apply this slope logic afterward!) 

We look forward to bringing this to more customers as we streamline the algorithms and make it easy for more people to deploy inside Optmzr