Home About Research Team Data Widget





About National Soil Moisture Network

Objective

This project utilizes in situ measurements of soil moisture and NRCS SSURGO soil characteristics and PRISM data to develop high-resolution gridded soil moisture products. These data will be disseminated through this web portal.

Products

Near Real-Time In-Situ Interpolated Soil Moisture Percentiles
Each day, soil moisture VWC at 7 am from networks are extracted. Due to each network report data with different latencies, data is retrieved with a 1-day latency. All the stations that have been used in the interpolation are shown as hollow circles in the maps. The number of available stations varies each day, depends on the in situ networks. Then, soil moisture VWC is converted to percentile based on the empirical CDF of that station. The CDF is computed from the start of the station until 12/31/2018, and a two-sided 15 days moving window is applied to generate an empirical CDF. Three interpolation methods: Inverse Distance Weighting (IDW), Ordinary Kriging (OK) and Regression Kriging (RK) have been used to interpolate soil moisture percentile data into 4 km grids. For the RK product, based on previous studies, precipitation from PRISM and soil texture from gSSURO have been selected as auxiliary variables.

Near Real-Time NLDAS Noah Soil Moisture Percentiles
Each day, the 6 AM hourly NLDAS-2 Noah model output is downloaded from the NASA GES DISC portal (https://hydro1.gesdisc.eosdis.nasa.gov/). Due to the delay in NLDAS-2 model output, data is retrieved at a five day latency. The NETCDF file is used to first replace any masked values with no data. Next, the Liquid SM content (non-frozen) and Total SM content variables are retrieved from the NETCDF. If a frozen soil filter is applied, soil moisture for the whole column is removed if the total and liquid soil moisture are not equal in any layer. Next, the native units of NLDAS soil moisture (kg/m2) are converted to volumetric water content (VWC) by dividing the moisture content by the soil layer thickness (mm). VWC anomalies are then generated by subtracting the mean daily value (1998-2017) from the daily VWC. Next, VWC percentiles are calculated according to each location's empirical CDF. The CDF is computed from the 1998-2017 VWC anomaly time series and a two-sided five day moving window is applied to generate an empirical CDF (n=100). Finally, the NLDAS VWC percentile data are clipped to the US and rescaled to the PRISM (4 km) grid using a cubic spline one dimensional interpolation.

Near Real-Time SMAP Soil Moisture Percentiles
Each day, the NSIDC subscription service provides a link to the most recently processed SMAP L3 enhanced data. This latency can vary but is limited to 50 hours. To match the NLDAS time step, SMAP data is processed at a five day latency. The lat, lon, SM, SM_Error, and Flags variables are extracted from the hdf5 file. It should be noted that the SMAP data is originally stored on NASA's EASE grid. SMAP quality control flags are then applied to the data and only SM data with recommended quality are stored (https://nsidc.org/data/smap/spl3smp_e/data-fields). Next, soil moisture retrievals from the AM and PM passes are merged using an averaging approach if there are overlapping retrievals. VWC is then converted to anomalies by subtracting the mean daily value (March 31, 2015-September 1, 2018) from the daily VWC. Next, VWC percentiles are calculated according to each location's empirical CDF. The CDF is computed from the March 31, 2015-September 1, 2018 VWC anomaly time series and a two-sided 30 day moving window is applied to generate an empirical CDF (n is variable). Finally, the SMAP VWC percentile data are clipped to the US and rescaled to the PRISM (4 km) grid using centroid matching. SMAP soil moisture data are also provided as a 3-day average to provide better coverage of satellite measurements.

Near Real-Time Blended Products
To match the NLDAS time step, the simple blend product is generated at a five day latency. The simple blend is computed using a simple average to combine the NLDAS Noah 0-10 cm VWC percentiles, the SMAP L3 enhanced VWC percentiles (~5 cm), and the 5 cm regression kriging (RK) in situ VWC percentiles. Blends of NLDAS and regression kriging soil moisture are also produced for the depths of 5, 20, and 50 cm. The uncertainty of the blended products is expressed as the relative difference between soil moisture measurements from different products.

Contributing Networks

National Soil Moisture Network does not collect soil moisture data. We are reliant on data contributions from state and federal networks to produce daily gridded soil moisture products. Users should visit each network's website to obtain original soil moisture data. Links can be found below:
Networks Measurement depths(cm)
Soil Climate Analysis Network (SCAN)
5, 10, 20, 50, 100
Climate Reference Network (CRN)
5, 10, 20, 50, 100
SNOwpack TELemetry Network (SNOTEL)
5, 20, 50
Oklahoma Mesonet (OKM)
5, 25, 60, 75
West Texas Mesonet (WTM)
5, 20, 60, 75
Delaware Environmental Observing System (DEOS)
5
Missouri Agricultural Weather Database (MHAWD)
5
University of Georgia Weather Network (UGA)
30
Kansas Mesonet (KSM)
5, 10, 20, 50
North Carolina Econet (NCEconet)
20
New York State Mesonet (NYM)
5, 25, 50
Roaring Fork Observation Network (iRON)
5, 20, 50
Illinois Climate Network (ICN)
5, 10, 20, 50
SoilSCAPE
Varying Depths
Iowa Environmental Mesonet (IEM)
30, 60, 125


Background

Spatial interpolation of in situ soil moisture is challenging because there are many factors that influence how soil moisture varies at regional scales including soil properties, topography, vegetation/land cover/land use and climate (Crow et al., 2012). Heterogeneity of soil properties influences all components of soil water balance and therefore has a significant impact on soil moisture. Rodriguez-Iturbe et al. (1995) found soil properties affect soil moisture distribution through commanding infiltration of moisture. Famiglietti et al. (1998) and Vereecken et al. (2007) concluded that soil hydraulic conductivities and porosity jointly influence the variability in surface soil moisture content. However, to date, no studies have developed gridded soil moisture products that account for the spatial variations in soil properties. The NRCS SSURGO data provides a unique gridded database of soil properties (gSSURGO) that is ideally suited for this project.

Soil moisture is also influenced by other factors. For example, routing processes that are affected by topography influence near-surface soil moisture. Numerous studies have demonstrated that soil moisture is generally lower at locations that are further upslope (Champagne et al., 2010; de Rosnay et al., 2009; Mohanty et al., 2000b). Land cover also plays an important role in determining soil moisture through modifying infiltration and evapotranspiration. Mohanty et al. (2000a) found that vegetation dynamics had a significant impact on intra-seasonal spatial patterns of soil moisture, which is consistent with other studies (Jacobs et al., 2004; Vinnikov et al., 1996). Isham et al. (2005) used a stochastic rainfall model and to demonstrate that the spatial variability of soil moisture is not only controlled by the spatial variability of vegetation, but also by rainfall. Precipitation is the dominant meteorological control of soil moisture (Seneviratne et al., 2010). The impact that all of these factors have on soil moisture varies depending on the spatial scale. Therefore, accurately interpolating soil moisture based on in situ (i.e., point) measurements is challenging and requires consideration of many different factors.

Methods

This work builds on the North American Soil Moisture Database (NASMD) (Quiring et al., 2016). This project collects daily soil moisture observations from the NASMD stations that provide data in near real-time. These data are collected, harmonized and quality controlled using the North American Soil Moisture Database QAQC algorithm (Quiring et al., 2016). Our central hypothesis is that NRCS SSURGO data, along with PRISM climate data and other ancillary data (e.g., land use/land cover and elevation) can be used to develop more accurate gridded soil moisture surfaces. This approach has previously been applied to produce more accurate estimates of soil carbon. A similar smart-interpolation approach is also used by PRISM to develop gridded climate data from station-based measurements. The procedures of data processing is shown in the following figure. More information about QAQC could be found here.


Data processing

Cooperators, Partners, and Subject Matter Experts

USGS-CIDA Center for Integrated Data Analytics
USDA-NRCS National Soil Survey Center
USDA-NRCS National Water and Climate Center
The National Integrated Drought Information System (NIDIS)
USDA Southern Plains Climate Hub
U.S. Climate Reference Network


OSU Logo NIDIS Logo NOAA Logo USDA Logo USGS Logo TAMU Logo
Questions and Comments? quiring.10@osu.edu