AGU 2024 poster

Transcript

1.           "& ') ( $'#%    ( !' "$'#%        &'      "& ') !$&"#      !$& !'#      !$!)#      "& ') ' &"#      ' & !'#      ' !)# Website: david0811.github.io Email: davidcl2@illinois.edu 1. MOTIVATION This work was supported by the U.S. Department of Energy, Office of Science, Biological and Environmental Research Program, Earth and Environmental Systems Modeling, MultiSector Dynamics under Cooperative Agreement DE-SC0022141. This work was partially supported by the Thayer School of Engineering at Dartmouth College. Any findings or recommendations expressed in this material are those of the author(s). Models involving soil moisture are crucial to understanding the evolution of many coupled natural-human systems under climate change. However, projecting long-term changes in soil moisture is difficult due to uncertainties surrounding soil dynamics and the representation of past and future climate. Fig. 1: Schematic representation of water fluxes within the simple soil moisture model. Additional details can be found in [1]. [1] Grogan, D.S., et. al., Water balance model (WBM) v.1.0.0: a scalable gridded global hydrologic model with water-tracking functionality, Geosci. Model Dev. (2022). https://doi.org/10.5194/gmd-15-7287-2022 David Lafferty & Ryan Sriver (UIUC) Danielle Grogan & Shan Zuidema (UNH) Atieh Alipour & Klaus Keller (Dartmouth) Iman Haqiqi (Purdue) Pre-calibrating a simple soil moisture model to facilitate uncertainty analysis We use a Python-implementation of the soil moisture module within the University of New Hampshire Water Balance Model [1]. Our model is a simple 1-dimensional representation of water fluxes within the active soil layer. Uncertain parameters include: alpha (efficiency of evapotranspiration) beta HBV (efficiency of runoff) wiltingp (wilting point) awCap (water capacity of the active layer) We pre-calibrate the simple model against 4 pseudo- observational products: VIC, MOSAIC, and NOAH model outputs from NLDAS-2, and the SMAP satellite L4 product. The pre-calibration uses Latin Hypercube sampling to draw 30,000 unique parameter samples for each pseudo- observational product. We then select the best 150 parameter combinations for each product based on the gridpoint-average RMSE. The simple model can reproduce each observational product over most of the domain and can generate an ensemble with reasonable coverage. Specific regions with large disagreements may require structural model changes. The pre-calibration constrains uncertain parameters to varying degrees. awCap and wiltingp are scalar factors applied to the corresponding fields from each observational product and are relatively well- identified. alpha and beta HBV are allowed to co-vary with soil content fractions (percentage sand, silt, clay) but are typically poorly identified. We plan to explore how the combined influence of hydrologic and climate uncertainty affects representations of past and future soil moisture, focusing on metrics relevant for agriculture in the central US. Better constraining these uncertainties can facilitate improved long-term decision- making regarding infrastructure investments and water management strategies in the agricultural sector. 4. PARAMETER IDENTIFICATION Fig. 2: (left) Spatial evaluation of the pre-calibration showing RMSE of the best parameter set, linear correlation coefficient of the best parameter set, and the percentage of observations falling within the simulation ensemble (top 150 parameter sets). (right) Daily timeseries for two locations showing observations in dark red, the best parameter set in dark gray, and the top 150 parameter sets in light gray. Associated evaluation metrics are also reported. Fig. 3: Parameter combinations of sufficient skill. The top 150 parameter combinations are shown for each observational product. The single best parameter combination for each product are shown as triangles on the abscissa. 2. SOIL MOISTURE MODEL 3. PRE-CALIBRATION 5. FUTURE WORK Unavailable water Active soil layer Precipitation & snowfall Canopy Evapotranspiration Available water capacity Wilting point Snowpack Runoff david0811.github.io Preprint coming soon! dcl257@cornell.edu 1) Climate-driven changes to soils are highly uncertain This work was supported by the U.S. Department of Energy, Office of Science, Biological and Environmental Research Program, Earth and Environmental Systems Modeling, MultiSector Dynamics under Cooperative Agreement DE-SC0022141. This work was partially supported by the Thayer School of Engineering at Dartmouth College. Any findings or recommendations expressed in this material are those of the author(s). • Soil moisture is crucial for understanding many climate risks, but long-term projections are uncertain due to both climate and hydrologic modeling factors. • Which uncertainties determine how key soil moisture metrics change in future? Fig. 1: Schematic representation of major water flux and storage components within pyWBM. [1] Grogan et al. (2022), Water balance model (WBM) v.1.0.0: a scalable gridded global hydrologic model with water-tracking functionality, Geosci. Model Dev. [2] Xia et al. (2014). Evaluation of multi-model simulated soil moisture in NLDAS-2, J. Hydrology. [3] Reichle et al. (2019), Version 4 of the SMAP Level-4 Soil Moisture Algorithm and Data Product. J. Adv. Model. Earth Sys. [4] Pierce et al. (2023), Future Increases in North American Extreme Precipitation in CMIP6 Downscaled with LOCA. J. Hydrometeor. David Lafferty1*, Danielle Grogan2, Shan Zuidema2, Atieh Alipour3†, Iman Haqiqi4, Ryan Sriver1, Klaus Keller3 1University of Illinois, 2University of New Hampshire, 3Dartmouth College, 4Purdue University, *now at Cornell University, †now at NOAA Combined climate and hydrologic uncertainties shape projections of future soil moisture extremes • pyWBM is a Python implementation of the soil moisture submodule of the University of New Hampshire Water Balance Model [1]. • We encode pyWBM in JAX to provide JIT- compilation and automatic differentiability. • We use a simple 1d representation of water fluxes within the active soil layer. Uncertain parameters include: 𝛼 (controls evapotranspiration) 𝛽R (controls runoff) wiltingp (wilting point) awCap (water capacity of the active layer) • We leverage the differentiability of pyWBM to calibrate using mini-batch gradient descent. • We calibrate against NLDAS-2 outputs [2] (VIC, MOSAIC, NOAH) and the SMAP L4 dataset [3]. We also calibrate using 13 different loss functions (RMSE, MSE, KGE, etc). • pyWBM can reproduce each product over most of the domain and generates an ensemble with reasonable coverage. • Sampling relevant climate and hydrologic uncertainties is crucial for understanding future hydroclimate risks. • Soil moisture is often an intermediate variable to downstream models of more societally-relevant outcomes––how do these uncertainties propagate to (e.g.) flood risk or crop yield impacts? • More sophisticated Bayesian or ML-based calibration approaches may lead to a better fit and/or improved uncertainty characterization. Fig. 3: Spatial distribution of trends and associated delta sensitivity indices. We analyze three soil moisture metrics, delineated across rows. The left subfigure (a) shows the 2.5th, median, and 97.5th percentile of trends across our ensemble; the right subfigure (b) shows the associated delta sensitivity indices. Fig. 2: (top) CDFs constructed over all gridpoints in the domain. Solid black line represents the best fit for each product, while the gray curves show calibration results with other loss functions. Colored curves show pairwise comparisons among products. (bottom) Example hindcast for a cropland location in Illinois, where the gray shading indicates our ensemble range. • We convolve our calibrated parameter ensemble with a set of climate projections from LOCA2 [4] to generate a 2340-member ensemble of future soil moisture projections. • We use Delta sensitivity analysis to measure which uncertain factors affect the trends of key soil metrics, including average soil moisture and the intensity of wet and dry extremes. • Our ensemble projects drying across most of the region; we expect dry extremes to exacerbate while wet extremes are more uncertain. Both climate and hydrologic uncertainties play a key role. 2) We develop pyWBM: a simple, fast, differentiable soil moisture model 3) We calibrate pyWBM against existing soil products, accounting for parameter uncertainty        4) We generate a large ensemble of future soil projections and use sensitivity analysis to reveal key uncertainties 5) Conclusions & future work ( % ( ( % ( ( % ( ( ( %. ) ( % ( ( % (        Coverage (anomaly space): 75.45% a) Projected trends (2016-2100) b) Sensitivity analysis