An official website of the United States government
Abstract. Land surface models (LSMs) are critical components of Earth system models (ESMs), enabling the simulation of energy and water fluxes that are essential for understanding climate systems. Soil hydraulic parameters, derived using pedotransfer functions (PTFs), are crucial for modeling soil–plant–water interactions; they introduce uncertainties in soil moisture simulations. However, a key knowledge gap exists in understanding how specific soil hydraulic properties contribute to these uncertainties and in identifying the regions most affected by them. This study conducts an intra-model sensitivity analysis within the Community Land Model version 5 (CLM5), examining how alternative soil parameter settings influence soil moisture variability across the contiguous United States (CONUS) using empirical orthogonal function (EOF) analysis. The EOF analysis revealed dominant spatial and temporal patterns of soil moisture across multiple experimental configurations, highlighting the impact of soil parameter variability on hydrological processes. The results showed significant discrepancies in soil moisture simulations, particularly in the central Great Plains, which may be attributed to the combination of arid climatic conditions and limitations in modeling saturated hydraulic conductivity and soil water retention curves. Seasonal soil moisture dynamics showed broad similarity to ERA5-Land patterns, with differences in magnitude and phase, indicating the importance of refined parameterization, particularly in the representation of infiltration and drainage processes. Comparisons with ERA5-Land, used here solely as a model-based reference for pattern consistency, revealed stronger similarity in regions with consistent climatic gradients but persistent differences in hydrologically complex areas, particularly in areas with arid climates, such as the Great Plains, where hydrological processes remain difficult to represent. Because CLM5 is forced by GSWP3, whereas ERA5-Land is an offline HTESSEL replay forced by ERA5, differences reflect both forcing and structural contrasts in addition to parameter effects. This research demonstrates the necessity to refine soil parameter representations, utilize high-resolution datasets, and consider climatic variability to inform the model development of LSMs. Importantly, these findings also pave the way for future efforts that incorporate dynamic soil properties into LSMs. This work illustrates the influence of soil properties on simulated variability. While the analysis documents their importance, a future direction will be to develop approaches that allow these properties to vary dynamically within LSMs. This study contributes to ongoing efforts toward more integrated modeling frameworks that capture soil–hydrology–climate interactions.
more »
« less
Training machine learning with physics-based simulations to predict 2D soil moisture fields in a changing climate
The water content in the soil regulates exchanges between soil and atmosphere, impacts plant livelihood, and determines the antecedent condition for several natural hazards. Accurate soil moisture estimates are key to applications such as natural hazard prediction, agriculture, and water management. We explore how to best predict soil moisture at a high resolution in the context of a changing climate. Physics-based hydrological models are promising as they provide distributed soil moisture estimates and allow prediction outside the range of prior observations. This is particularly important considering that the climate is changing, and the available historical records are often too short to capture extreme events. Unfortunately, these models are extremely computationally expensive, which makes their use challenging, especially when dealing with strong uncertainties. These characteristics make them complementary to machine learning approaches, which rely on training data quality/quantity but are typically computationally efficient. We first demonstrate the ability of Convolutional Neural Networks (CNNs) to reproduce soil moisture fields simulated by the hydrological model ParFlow-CLM. Then, we show how these two approaches can be successfully combined to predict future droughts not seen in the historical timeseries. We do this by generating additional ParFlow-CLM simulations with altered forcing mimicking future drought scenarios. Comparing the performance of CNN models trained on historical forcing and CNN models trained also on simulations with altered forcing reveals the potential of combining these two approaches. The CNN can not only reproduce the moisture response to a given forcing but also learn and predict the impact of altered forcing. Given the uncertainties in projected climate change, we can create a limited number of representative ParFlow-CLM simulations (ca. 25 min/water year on 9 CPUs for our case study), train our CNNs, and use them to efficiently (seconds/water-year on 1 CPU) predict additional water years/scenarios and improve our understanding of future drought potential. This framework allows users to explore scenarios beyond past observation and tailor the training data to their application of interest (e.g., wet conditions for flooding, dry conditions for drought, etc…). With the trained ML model they can rely on high resolution soil moisture estimates and explore the impact of uncertainties.
more »
« less
- Award ID(s):
- 2134892
- PAR ID:
- 10468497
- Publisher / Repository:
- Frontiers in Water
- Date Published:
- Journal Name:
- Frontiers in Water
- Volume:
- 4
- ISSN:
- 2624-9375
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Vadose zone soil moisture is often considered a pivotal intermediary water reservoir between surface and groundwater in semi-arid regions. Understanding its dynamics in response to changes in meteorologic forcing patterns is essential to enhance the climate resiliency of our ecological and agricultural system. However, the inability to observe high-resolution vadose zone soil moisture dynamics over large spatiotemporal scales hinders quantitative characterization. Here, utilizing pre-existing fiber-optic cables as seismic sensors, we demonstrate a fiber-optic seismic sensing principle to robustly capture vadose zone soil moisture dynamics. Our observations in Ridgecrest, California reveal sub-seasonal precipitation replenishments and a prolonged drought in the vadose zone, consistent with a zero-dimensional hydrological model. Our results suggest a significant water loss of 0.25 m/year through evapotranspiration at our field side, validated by nearby eddy-covariance based measurements. Yet, detailed discrepancies between our observations and modeling highlight the necessity for complementary in-situ validations. Given the escalated regional drought risk under climate change, our findings underscore the promise of fiber-optic seismic sensing to facilitate water resource management in semi-arid regions.more » « less
-
Abstract Integrated hydrologic models can simulate coupled surface and subsurface processes but are computationally expensive to run at high resolutions over large domains. Here we develop a novel deep learning model to emulate subsurface flows simulated by the integrated ParFlow‐CLM model across the contiguous US. We compare convolutional neural networks like ResNet and UNet run autoregressively against our novel architecture called the Forced SpatioTemporal RNN (FSTR). The FSTR model incorporates separate encoding of initial conditions, static parameters, and meteorological forcings, which are fused in a recurrent loop to produce spatiotemporal predictions of groundwater. We evaluate the model architectures on their ability to reproduce 4D pressure heads, water table depths, and surface soil moisture over the contiguous US at 1 km resolution and daily time steps over the course of a full water year. The FSTR model shows superior performance to the baseline models, producing stable simulations that capture both seasonal and event‐scale dynamics across a wide array of hydroclimatic regimes. The emulators provide over 1,000× speedup compared to the original physical model, which will enable new capabilities like uncertainty quantification and data assimilation for integrated hydrologic modeling that were not previously possible. Our results demonstrate the promise of using specialized deep learning architectures like FSTR for emulating complex process‐based models without sacrificing fidelity.more » « less
-
While machine learning approaches are rapidly being applied to hydrologic problems, physics-informed approaches are still relatively rare. Many successful deep-learning applications have focused on point estimates of streamflow trained on stream gauge observations over time. While these approaches show promise for some applications, there is a need for distributed approaches that can produce accurate two-dimensional results of model states, such as ponded water depth. Here, we demonstrate a 2D emulator of the Tilted V catchment benchmark problem with solutions provided by the integrated hydrology model ParFlow. This emulator model can use 2D Convolution Neural Network (CNN), 3D CNN, and U-Net machine learning architectures and produces time-dependent spatial maps of ponded water depth from which hydrographs and other hydrologic quantities of interest may be derived. A comparison of different deep learning architectures and hyperparameters is presented with particular focus on approaches such as 3D CNN (that have a time-dependent learning component) and 2D CNN and U-Net approaches (that use only the current model state to predict the next state in time). In addition to testing model performance, we also use a simplified simulation based inference approach to evaluate the ability to calibrate the emulator to randomly selected simulations and the match between ML calibrated input parameters and underlying physics-based simulation.more » « less
-
Abstract More frequent and severe droughts are driving increased forest mortality around the globe. We urgently need to describe and predict how drought affects forest carbon cycling and identify thresholds of environmental stress that trigger ecosystem collapse. Quantifying the effects of drought at an ecosystem level is complex because dynamic climate–plant relationships can cause rapid and/or prolonged shifts in carbon balance. We employ the CARbon DAta MOdel fraMework (CARDAMOM) to investigate legacy effects of drought on forest carbon pools and fluxes. Our Bayesian model‐data fusion approach uses tower observed meteorological forcing and carbon fluxes to determine the response and sensitivity of aboveground and belowground ecological processes associated with the 2012–2015 California drought. Our study area is a mid‐montane mixed conifer forest in the Southern Sierras. CARDAMOM constrained with gross primary productivity (GPP) estimates covering 2011–2017 show a ~75% reduction in GPP, compared to negligible GPP change when constrained with 2011 only. Precipitation across 2012–2015 was 45% (474 mm) lower than the historical average and drove a cascading depletion in soil moisture and carbon pools (foliar, labile, roots, and litter). Adding 157 mm during an especially stressful year (2014, annual rainfall = 293 mm) led to a smaller depletion of water and carbon pools, steering the ecosystem away from a state of GPP tipping‐point collapse to recovery. We present novel process‐driven insights that demonstrate the sensitivity of GPP collapse to ecosystem foliar carbon and soil moisture states—showing that the full extent of GPP response takes several years to arise. Thus, long‐term changes in soil moisture and carbon pools can provide a mechanistic link between drought and forest mortality. Our study provides an example for how key precipitation threshold ranges can influence forest productivity, making them useful for monitoring and predicting forest mortality events.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript
- Journal Article:
- https://doi.org/10.3389/frwa.2022.927113
