“Geovariances successfully helped us in building high quality hydrostratigraphic surface maps using local geostatistics”

Saint Johns River Water Management District
Jeffrey B. Davis
P.G, Division of Groundwater Programs

Saint Johns River Water Management District required Geovariances to review their project of mapping hydrostratigraphic surfaces in Peninsular Florida. Local geostatistics has revealed to be the solution to answer their issue.

 

Quick facts

The St. Johns River Water Management District, located in northeast Florida USA, needed to create boundary surfaces of hydrostratigraphic units for input into groundwater models. Previous work typically focused on specific regions of the state and issues arose in how researchers identified specific units and from mapping across political boundaries.

A project was initiated to assimilate hydrogeologic data from multiple sources and develop consistent regional hydrogeologic unit surface estimations for the Florida peninsula.

The challenges

  • Florida geology is complicated by karst surfaces developed on many of the geologic units. Structural features include broad uplifted regions and erosional basins.
  • The number of boreholes that penetrated the various units varied from over 5,000 for the Upper Floridan aquifer to 151 for the deepest unit mapped, the Lower Floridan aquifer. Spatial distribution of boreholes also varied with some areas having clustered data points and others with scarcely sampled data. The point distribution is also a concern when one unit has thousands of points and a lower unit only has a few, sometimes resulting in a surface unrealistically crossing an overlying surface.
  • New data are constantly being added and a simple method to update the surfaces was required.
  • A critical goal was to have the surface estimations honor the data points and capture both the short-range features (such as boreholes drilled into buried sinkholes) and the long-range regional structure. The resulting surface should also be realistic without erratic high or low values where there were no data.
  • A peer review of the techniques was desired since there may be legal challenges during the District regulatory process.
  • Staff intended to structure this work as an educational opportunity for future Isatis projects.

The solution

An initial Isatis study was developed by SJRWMD staff and exported to Geovariances for review. The Geovariances team reviewed the data, techniques used, and results. Remote meetings were held to discuss their findings and to determine what would be the best approach based on these initial findings. Tests using ordinary kriging for each unit were done and problems were identified such as lower surfaces crossing upper surfaces in the area where the lower units were scarcely sampled.

The results

Based on these and other tests it was determined that a Local Geostatistics (a variant of kriging which makes use of local parameters) approach would be best. An example of the local range map for the top of the Floridan Aquifer System is shown below.

The varying range for the residuals can account for the increased variability in areas with a smaller range in the central area of the District versus the northern and southern regions where the behavior of the variable is smoother. This helps solve the problems related to mapping the highly variable karst surface in the central regions compared to other regions where the surface is not as affected by karst or structural features.

Local range map for the top of the Floridan Aquifer System
Local range map for the top of the Floridan Aquifer System

Validation was performed by dividing the dataset into two parts and using one half of the points to estimate the value at the other half of the points. To resolve the problem of inconsistencies related to an upper, highly sampled surface compared to a deeper, scarcely sampled surface an approach based on thickness was utilized for the deeper units. Consistent with the regional geologic structure, it was determined that the tops of the deeper horizons were well correlated with the upper more densely sampled horizons. The upper unit grids were smoothed to minimize the impact of local variations. Thickness was calculated where data was available. A comparison of the three different method results demonstrate the effectiveness of kriging the  thickness based on a filtered upper unit.

The 3rd figure below shows the deeper surfaces, captures the trend of the overlying filtered surface and produces a realistic surface based on the limited data.

Cross-sections of boundaries surfaces
Cross-sections of boundaries surfaces

As new data are obtained the surfaces can be updated by using Isatis journal files geveloped by Geovariances. The new data is imported into Isatis and the journal files are run to generate the new surfaces. As new data are added over, the parameters will be reviewed and the journal files may need updating.

Maps of surfaces of boundaries between hydrogeologic units of expected quality with Isatis advanced geostatistics

Top of the Floridan Aquifer System in Peninsular Florida
Top of the Floridan Aquifer System in Peninsular Florida

Mapping surfaces of boundaries between hydrogeologic units presented challenges due to the short range variations from karst development and disparity between the number of data points available for upper and lower hydrostratigraphic surfaces. Ordinary kriging created problems with surfaces crossing in areas with scarcely sampled data. The Moving Geostatistics approach provided varying parameters for different areas which worked well for mapping at this scale and with a variety of geological provinces. Once a correlation between the upper and lower units was established the thickness filtering helped resolve issues related to mapping a lower surface with 151 points compared to an upper surface with over 5,500 points for an area the size of peninsular Florida.