D.6 GEOSTATISTICAL TEMPORAL-SPATIAL OPTIMIZATION SOFTWARE (GTS)

Approximate Cost: Free

Source: https://itrcweb.org/team/GTS-Optimization-Software

Current Version: v1.0

Operating System Needs: Windows XP (has been successfully installed and used on Windows 7, though not formally supported)

Input Structure: ASCII (text) flat file with fixed column header names; one data row per measurement; tab-delimited preferred but not required

Overview

Geostatistical Temporal-Spatial software (GTS) is a statistical and geostatistical decision-logic groundwater monitoring optimization software that is publicly available as open-source freeware. GTS is a quantitative calculation tool that includes options to customize its use. It was developed for the Air Force Civil Engineer Center (AFCEC), known previously as Air Force Center for Engineering and the Environment (AFCEE). Given an existing long-term monitoring (LTM) network, GTS uses a combination of statistical techniques to answer two questions:

1. What is the optimum number and placement of wells in that network?
2. What is the optimal sampling frequency for wells in the network?

GTS has five modular components linked together in a user-friendly interface: Prepare, Explore, Baseline, Optimize, and Predict. The Prepare and Explore modules allow the user to import and manage analytical and water-level data, identify outliers, explore basic statistical features of the data (including simple trends), and also to rank contaminants in terms of optimization potential. The Baseline module creates nonlinear trends and trend maps, and constructs base maps to quantify and visualize plume extent. Baseline also allows you to create potentiometric surface maps. The Optimize component runs two distinct types of temporal optimization—iterative thinning and temporal variograms—as well as spatial optimization involving both a search for statistical redundancy and an assessment as to whether and where new wells should be added. The software is designed so that you may chose only to perform the temporal optimization as a stand-alone module. However, the spatial analysis depends on the temporal analysis being performed first in sequence to obtain the spatial results. Finally, the Predict module focuses on flagging newly imported data that are inconsistent with projected trends and maps.

Add-Ins Available

None

Ease of Use and Data Import

GTSGeostatistical Temporal-Spatial optimization software is officially limited to the Windows XP platform, though some users have successfully installed and used it on Windows 7. The wizard interface offers a number of defaults and makes it easy to click through a basic analysis. Basic defaults can be configured and set with many preferences to allow a highly customized optimization. Process flow is logical from top to bottom and left to right when navigating the sequence of operations in the dialog window on each page. Interpreting the results properly requires an intermediate backgroundNatural or baseline groundwater quality at a site that can be characterized by upgradient, historical, or sometimes cross-gradient water quality (Unified Guidance). or training in statistics and geostatisticsA branch of statistics that focuses on the analysis of spatial or spatiotemporal data, such as groundwater data (Gilbert 1987).. GTS includes simple plots, exploratory tools, and trend analyses, as well as sophisticated statistical techniques and optimization algorithms written in the open-source statistical computing environment R (www.r-project.org).

GTSGeostatistical Temporal-Spatial optimization software requires input of a structured ASCII (text) flat file. The fields can be delimited in a variety of ways, such as tab-delimited or comma-separated values (CSV), but must have specific field names, generally corresponding to the format of AFCEC’sAir Force Civil Engineer Center'sEnvironmental Resource Program Information Management System (ERPIMS) database. The order or sequencing of data fields is not critical. Shape files of facility boundaries, sites, roads, and other infrastructure can be imported. Each groundwater measurement must occupy one record of the input text file. Fields required for a GTS analysis are listed within the GTS Users Guide. Data files in Excel or spreadsheet format must be exported to text format prior to GTS input. A data filtering tool allows analysis of selected records.

Types of Distributions

GTSGeostatistical Temporal-Spatial optimization software accepts data of any distributional type. Although you cannot apply data transformations within GTS, most of the procedures within GTS are quasi-nonparametricStatistical test that does not depend on knowledge of the distribution of the sampled population (Unified Guidance). and do not require explicit fitting of parametricA statistical test that depends upon or assumes observations from a particular probability distribution or distributions (Unified Guidance). models or distributional testing.

Visualization

GTSGeostatistical Temporal-Spatial optimization software includes sophisticated built-in graphics for data visualization, including contour mapping, complex nonlinear trends, post-plots, and shape file annotation. GTS provides automated batch processing of graphics in order to sequentially plot multiple wells, contaminants, aquifer zones, and time periods. Graphics are designed to be final pictures for reports, however, the program cannot batch print graphics. In addition, users cannot tweak or alter the graphics formatting. On the other hand, some interactive widgets are provided, for instance, zooming and scaling tools, and pointers for identifying specific locations on plan-view maps. Individual graphs are best exported using the Windows Snipping tool or an equivalent screen capturing application.

Primary Uses for Groundwater Data Analysis

GTSGeostatistical Temporal-Spatial optimization software can be used at various stages in the life cycle of groundwater monitoring, but is best for optimizing long-term monitoring networks, once characterization has been completed and remedies are in place. Although the exploratory tools can be used during any stage of a facility’s life cycle, GTS generally assumes that a given site has been adequately characterized, is undergoing long-term monitoring, and that enough well locations exist and sampling data collected so that statistical redundancy in locations and sampling events might exist.

Benefits

• Applicable to site-specific plumes or site-wide studies (for example, entire facilities or installations) involving multiple source areas, plumes, and monitoring conditions.
• Does not require plume-specific configuration data, fate-and-transport models, or other hydrogeologic modeling information

• Stand-alone spatial and temporal optimization modules that can be used independently

• Exploratory statistical tools for assessing data characteristics, ranking contaminants for optimization potential, and analyzing multiple aquifer horizons

• Fitting of nonlinear and seasonal time series data

• Semi-nonparametric surface map estimates made using quantile local regression, a smoothing technique not bound by the constraints of krigingA weighted moving-average technique to interpolate the data distribution by calculating an area mean at nodes of a grid (Gilbert 1987).

• Empirical, data-driven assessment of redundancy (reduced-network is optimal if it can accurately reproduce base maps).

• Automated redundancy searches, both during temporal and spatial optimization

• Use of multiple cost-accuracy tradeoff curves to gauge points of optimality
• No limitations on the number of monitoring wells or sampling events
• Spatial analysis uses quasi-genetic algorithm to determine essential and redundant wells
• Imports multiple shape files for boundaries and infrastructure
• Temporal analysis proposes optimal sampling intervals specific to the number of quarters
• Database filtering tool helps select records for "what if" analysis

Limitations and Data Requirements

• Preparing the data set can be challenging with potentially a large number of data fields
• Effective spatial optimization in GTSGeostatistical Temporal-Spatial optimization software requires a minimum of 15-20 wells and at least two sampling events per well; temporal optimization requires at least one well and 6-8 distinct sampling events per location.

• Quantile local regression, the GTS spatial mapping engine, by design is a ‘smoother’ rather than an interpolator (thus may not replicate or ‘honor’ observed measurements when creating map estimates, unlike, for instance, kriging)

• Does not offer sophisticated handling of radiochemical data, particularly measurements recorded with non-positive values (zeros or negatives); must first convert these data to positive values, unless they represent nondetectsLaboratory analytical result known only to be below the method detection limit (MDL), or reporting limit (RL); see "censored data" (Unified Guidance). with a known, positive detection or reporting limit

• Does not track changes in contaminant or plume mass or allow users to specify contaminant mass as an optimization criterionGeneral term used in this document to identify a groundwater concentration that is relevant to a project; used instead of designations such as Groundwater Protection Standard, clean-up standard, or clean-up level.

• May not give valid spatial results in subsurface environments that are highly fractured and discontinuous with poor hydraulic connection.
• Note: Spatial mapping techniques in general (not just those in GTS) inherently assume that concentration patterns at known wells can be extended (interpolated, smoothed) to unsampled locations. This may be problematic at sites with large contrasts in hydraulic conductivity (preferential pathways).

References

Cameron, K., P. Hunter, and R. Stewart. 2011. Demonstration and validation of GTS long-term monitoring optimization software at military and government sites. ESTCP Project ER-200714. www.serdp.org.

Cameron, K. 2004. “Better optimization of LTM networks.” Bioremediation Journal8 (03-04): 89-108.

Cameron, K., and P. Hunter. 2004. Optimizing LTM networks with GTS: three new case studies. Conference on Accelerating Site Closeout, Improving Performance, & Reducing Costs Through Optimization, Dallas.

Cameron, K. M., and P. Hunter. 2003. “Optimization of LTM networks at AF Plant 6 using GTS”. In V.S. Magar & M.E. Kelley (Eds.), In Situ and On-Site Bioremediation – 2003. Proceedings of the Seventh International In Situ and On-Site Bioremediation Symposium (Orlando, FL; June 2003), Columbus, OH: Battelle Press.

Cameron, K., and P. Hunter. 2002. “Using spatial models and kriging techniques to optimize long-term ground-water monitoring networks: a case study”.Environmetrics13: 629-656.

Cameron, K., and P. Hunter. 2000. "Optimization of LTM networks: statistical approaches to spatial and temporal redundancy." Spring Natl. Meeting of American Institute of Chemical Engineers, Atlanta.

 

 

 

Publication Date: December 2013