Computational Algorithm
The GRACE Groundwater Subsetting Tool (GSST) Web Application relies on the Earth Observation data collected by NASA through satellites which map the gravitational field of the Earth. Changes in gravity are driven by changes in water storage, offering a rare opportunity to monitor groundwater level through satellites coupled with estimated surface water.
The GRACE mission was launched in March 2002. It consists of a pair of satellites that are 400km above the Earth and are separated by 200km. As the satellites pass over different regions of the Earth, the front and rear satellites are pulled slightly forward and backward in response to subtle changes in the Earth’s graviatational field caused by changes in surficial mass. This causes the distance between the satellites to vary, and the changes are recorded by a k-band microwave whose accuracy is within 10 microns. The GRACE satellites follow a varying path that covers the entire Earth about once per month. This data is then processed by NASA to produce a map of the Earth’s gravitational field. Each month a new map is generated and the differences are calculated to produce a gravity anomaly map. The changes in mass are assumed to be primarily caused by the change in water storage. Each month NASA generates a gridded map of total water storage anomaly at 3-degree resolution. This map is then down-scaled using a mass conservation algorithm to 0.5-degree resolution and made available for download in netCDF multidimensional raster format.
Derivation of Groundwater Dataset
The groundwater component of the GRACE raw data can be separated using a mass balance approach, with NASA’s Global Land Data Assimilation System (GLDAS) models to compute the surface water component of the data. To compute total surface water storage, we sum the components of the GLDAS models that represent surface water storage and subtract this total from the GRACE dataset to estimate a groundwater storage anomaly dataset.
This GSST application uses four sets of data:
The GRACE TWSa dataset
The GLDAS canopy storage dataset (CAN)
The GLDAS snow water equivalent (SWE)
The GLDAS soil moisture (SM)
To compute the groundwater storage anomaly (GWa), we use three components of the GLDAS models: CAN, SWE, and SM. We convert each GLDAS component to an anomaly format by subtracting the mean centered on values from 2004 to 2009 and then average across the three GLDAS models to produce a component anomaly dataset: CANa, SWEa and SMa. We use the standard deviation from the three GLDAS models to help estimate uncertainty.
We download GLDAS files, format them as netCDFs and store them locally. Normally the data is acquired in a gridded format with a 1-degree latitude by 1-degree longitude resolution, which we then convert to a 0.5-degree resolution. This conversion is performed by an area-weighted average of the four GRACE grid cells coincident with each GLDAS grid cell.
The converted files are then used to compute the groundwater anomaly using a mass balance approach. The groundwater anomaly is the difference between the TWSa and the sum of the surface water component anomalies.
The result of this computation is the ground water storage anomaly, a tested and approved method to predict long-term changes in groundwater storage.
Grid Subsetting
For regional subsetting, the user provides a shapefile that defines the boundary of the region of interest. We then select the cells that have cell centers within the defined boundary and calculate the average storage anomaly for each of the components: TWSa, SWEa, CANa, and SMa resulting in a time series from 2002 to the present for each component on a monthly time step. The figure below shows the Chad Basin in Niger subsetted and displayed with the region shapefile. For water storage, the average of each component is multiplied by the area of the region, resulting in volume anomalies.
Uncertainty Estimates
It is critical to understand that the results of these predictions have uncertainties and limitations.
To compute the uncertainty of the groundwater storage component, we combine the uncertainty estimates from both the GRACE and GLDAS by computing the square root of the sum of the squares of the uncertainty of the individual components as measured by their standard deviations.
The resulting estimates of groundwater data are not suitable for highly precise or localized applications, such as the placement of wells; rather, these data serve as an estimate of general trends in groundwater storage.
Storage Depletion Curve
The GGST offers an option of viewing time series data in the format of a storage depletion curve, which is the time-integral of the storage anomaly.
The storage depletion curve presents cumulative changes in water component storage relative to levels when the GRACE missions began distributing data in April 2002. The storage depletion curve is used in groundwater management since it offers a simple visualization of how much storage aquifers have gained or lost since a given point in time.
To compute the depletion, we sum the GWSa over time to determine changes in groundwater storage volume over time for the region. These data show if a region is depleting storage in the region, or if groundwater is recharging in the region thereby providing valuable information relative to groundwater sustainability.
Here is an illustration of Northern Africa and the Arabian Peninsula from 2002 - 2021. It shows that the groundwater in that region has been depleting since early 2009 and onward.
Limitations
GRACE data come with limitations that users need to know and understand. The data are provided at a relatively low resolution (1-degree latitude by 1-degree longitude) representing a 100 km x 100 km square, approximately. At such a low resolution, basing decisions on a single cell comes with high and unknown uncertainties. Raw GRACE data is at an even coarser resolution (3-degrees latitude by 3-degrees longitude) which is then processed to higher resolutions TWSa data.
Even with these limitations, GRACE data provide valuable insights into aquifers such as regions that are depleting and recharging, hence allowing managers to sustainably use their groundwater resources. The best use of the GGST is to draw general trends in aquifers rather than selecting a placement of a well.
It is also recommended that, whenever possible, these data be validated with local data. GGST displays the uncertainties in the data calculations as error bands on time series, providing context on regions and different time periods.
Software Availability
The GGST web application was created using Tethys Platform, developed in the BYU Hydroinformatics Laboratory. It can be accessed on a Tethys portal associated with the NOAA GeoGLOWS project by browsing to this link and selecting the Grace Groundwater Subsetting Tool application.