An official website of the United States government
This study explores machine learning for estimating soil moisture at multiple depths (0–5 cm, 0–10 cm, 0–20 cm, 0–50 cm, and 0–100 cm) across the coterminous United States. A framework is developed that integrates soil moisture from Soil Moisture Active Passive (SMAP), precipitation from the Global Precipitation Measurement (GPM), evapotranspiration from the Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS), vegetation data from the Moderate Resolution Imaging Spectroradiometer (MODIS), soil properties from gridded National Soil Survey Geographic (gNATSGO), and land cover information from the National Land Cover Database (NLCD). Five machine learning algorithms are evaluated including the feed-forward artificial neural network, random forest, extreme gradient boosting (XGBoost), Categorical Boosting, and Light Gradient Boosting Machine. The methods are tested by comparing to in situ soil moisture observations from several national and regional networks. XGBoost exhibits the best performance for estimating soil moisture, achieving higher correlation coefficients (ranging from 0.76 at 0–5 cm depth to 0.86 at 0–100 cm depth), lower root mean squared errors (from 0.024 cm3/cm3 at 0–100 cm depth to 0.039 cm3/cm3 at 0–5 cm depth), higher Nash–Sutcliffe Efficiencies (from 0.551 at 0–5 cm depth to 0.694 at 0–100 cm depth), and higher Kling–Gupta Efficiencies (0.511 at 0–5 cm depth to 0.696 at 0–100 cm depth). Additionally, XGBoost outperforms the SMAP Level 4 product in representing the time series of soil moisture for the networks. Key factors influencing the soil moisture estimation are elevation, clay content, aridity index, and antecedent soil moisture derived from SMAP.
more »
« less
Responses of soil moisture to climate variability and livestock grazing in a semiarid Eurasian steppe
Soil water is vital for sustaining semiarid ecosystems. However, data on soil moisture have unlikely been continuously collected for a long time (e.g., >50 years), let alone under various combinations of climates and livestock grazing intensities. The objective of this study was to formulate and parameterize an ecohydrological model for predicting long-termvariability of soil moisture, taking a typical Eurasian grassland located in northeast China as the testbed. The parameters were determined by extensive literature review, field reconnaissance, laboratory analyses of soil and grass samples, and model calibration using daily soil temperatures and soil moistures measured at four depths from 2014 to 2017. The model, driven by the daily climate data from 1955 to 2017, performed well in reproducing the measurements. Across the assessment years of 1960 to 2017, the daily soil moistures were predicted to vary from 0.02 to 0.38. Overall, the soil moistures at a shallower depth were smaller but had a wider range than those at a deeper depth, with a largest mean and a widest range around the 30 cm depth. Regardless of the depths, the soil moistures pulsed in beginning March and plateaued from May to September. Livestock grazing was precited to reduce top 1.5-cm soil moistures but increase moistures of the beneath soils. The optimal grazing intensity was determined to be around 3.0 cattle ha−1, above which wind erosion would become a concern. The grazing impacts on soil moisture were found to monophonically decrease with increase of evapotranspiration or annual precipitation of larger than 220mm. For the years with an annual precipitation of less than 220mm, such grazing impacts either increased or decreased with increase of precipitation, depending on the relative magnitude of evapotranspiration. Climate change will diminish soil moisture pulses in early spring, likely intensifying soil erosion by wind.
more »
« less
- Award ID(s):
- 1654957
- PAR ID:
- 10320071
- Date Published:
- Journal Name:
- Science of the total environment
- Volume:
- 781
- ISSN:
- 1879-1026
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Soil sampling pits across three hillslope positions - toeslope, backslope, and summit - were dug in 2020 in watershed N4D (burned every 4 years) and N1D (burned annually) to characterize the impacts of woody encroachment on subsurface soil physical, chemical, and biological properties. Pits were hand-dug to 120 cm in the toeslope position and to 60 cm deep at the backslope and summit positions. Soil pits in N4D were dug directly under dogwood shrubs (Cornus drumondii) while pits in N1B were dug under grasses and forbs. Soil pit faces were photographed to determine root fractions with depth, soil monoliths were take to charaterize soil macroporosity with depth while soil cores were taken in each horizon for water retention analysis. Soil sensors were also installed at four soil depths at the toeslope position and 3 soil depths at the backslope and summit positions to record half hourly soil moisture, soil temperature, soil water potential, soil electrical conductivity, and soil carbon dioxide, and soil oxygen. In addition, geophysical measurements were taken in N4D using time-lapse electrical resistivity.more » « less
-
Soil sampling pits across three hillslope positions - toeslope, backslope, and summit - were dug in 2020 in watershed N4D (burned every 4 years) and N1D (burned annually) to characterize the impacts of woody encroachment on subsurface soil physical, chemical, and biological properties. Pits were hand-dug to 120 cm in the toeslope position and to 60 cm deep at the backslope and summit positions. Soil pits in N4D were dug directly under dogwood shrubs (Cornus drumondii) while pits in N1B were dug under grasses and forbs. Soil pit faces were photographed to determine root fractions with depth, soil monoliths were take to charaterize soil macroporosity with depth while soil cores were taken in each horizon for water retention analysis. Soil sensors were also installed at four soil depths at the toeslope position and 3 soil depths at the backslope and summit positions to record half hourly soil moisture, soil temperature, soil water potential, soil electrical conductivity, and soil carbon dioxide, and soil oxygen. In addition, geophysical measurements were taken in N4D using time-lapse electrical resistivity in 2023.more » « less
-
Grazing can affect plant community composition and structure directly by foraging and indirectly by increasing wind erosion and dust storms and subsequently influence ecosystem functioning and ecological services. However, the combined effects of grazing, wind erosion, and dust deposition have not been explored. As part of a 7-year (2010–2016) field manipulative experiment, this study was conducted to examine the impacts of grazing and simulated aeolian processes (wind erosion and dust deposition) on plant community cover and species richness in a temperate steppe on the Mongolian Plateau, China. Grazing decreased total cover by 4.2%, particularly the cover of tall-stature plants (> 20 cm in height), but resulted in 9.1% greater species richness. Wind erosion also reduced total cover by 17.0% primarily via suppressing short-stature plants associated with soil nitrogen loss, but had no effect on species richness. Dust deposition enhanced total cover by 5.7%, but resulted in a 7.3% decrease in species richness by driving some of the short-stature plant species to extinction. Both wind erosion and dust deposition showed additive effects with grazing on vegetation cover and species richness, though no detectable interaction between aeolian processes and grazing could be detected due to our methodological constraints. The changes in gross ecosystem productivity, ecosystem respiration, and net ecosystem productivity under the wind erosion and dust deposition treatments were positively related to aeolian process-induced changes in vegetation cover and species richness, highlighting the important roles of plant community shifts in regulating ecosystem carbon cycling. Our findings suggest that plant traits (for example, canopy height) and soil nutrients may be the key for understanding plant community responses to grassland management and natural hazards.more » « less
-
An experiment was conducted from 2017-2023 at the University of St. Thomas research garden (Saint Paul, MN) to determine rates of nutrient recycling and loss from compost applied to urban gardens. Thirty-two 4 m2 study plots received one of six different soil amendment treatments, with four different crops growing on each plot. Meteorological data includes hourly measurements of rainfall, solar radiation, temperature and relative humidity, and wind speed and direction, from June 2017-October 2023. Hourly soil moisture measurements were recorded at depths of 10 cm, 20 cm, and 30 cm, from June-December 2021, June-October 2022, and June-October 2023. Annual crop harvest totals from each subplot are reported for 2017-2023. Leachate was collected from lysimeters installed in each of the 132 subplots weekly from June-October of each year (2017-2023), recording total volume. Leachate subsamples were analyzed for NO3-N, NH4-N, and PO4-P. Soil samples were collected at the beginning and end of the growing season in 2017, and every two weeks during the growing season from 2018-2023, and analyzed for pH, organic matter, Bray-1 extractable P, available K, nitrate, and ammonium, at the University of Minnesota Analytical Research Laboratory.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/https://doi.org/10.1016/j.scitotenv.2021.146705
