Assimilation of cosmic‐ray neutron counts for the estimation of soil ice content on the eastern Tibetan Plateau

Posted by Yijian Zeng in the category Case Studies

Accurate observations and simulations of soil moisture phasal forms are crucial in cold-region hydrological studies. In the seasonal frozen ground of Eastern Tibetan Plateau, water vapor, liquid and ice co-exist in the frost-susceptible silty-loam soil during winter. Quantification of soil ice content is thus vital in the investigation and understanding of the region’s freezing-thawing processes. This study focuses on the retrieval of soil ice content utilizing the in-situ soil moisture (i.e., the liquid phase) and cosmic-ray neutron measurements (i.e., the total water including both liquid and ice phases), with Observing System Simulation Experiments (OSSEs). To derive the total soil water content from neutron counts, different weighting methods (revised, conventional and uniform) for calibrating the cosmic-ray neutron probe (CRNP) were intercompared. The comparison showed that the conventional non-linear method performed the best. Furthermore, to assimilate neutron counts using Particle Filter, the STEMMUS-FT (Simultaneous Transfer of Energy, Mass and Momentum in Unsaturated Soil) model was used as the physically-based process model, and the COSMIC model (Cosmic Ray Soil Moisture Interaction Code) was used as the observation operator (i.e., forward neutron simulator). Other than background inputs from disturbed initializations in the STEMMUS-FT, model uncertainties were predefined to assimilate neutron counts. We observed that with enough spread of uncertainties, the updated states could mimic the CRNP observation. In all setups, assimilating CRNP measurements could enhance total soil water analyses, which consequently led to the improved detection of soil ice content and therefore the freezing-thawing process at the field scale.

Mwangi, S., Zeng, Y., Montzka, C., Yu, L., & Su, Z. (2020). Assimilation of cosmic‐ray neutron counts for the estimation of soil ice content on the eastern Tibetan Plateau. Journal of geophysical research : Atmospheres, 125(3), 1-23. [e2019JD031529].

Figure 1 Flow diagram of this study (the assimilation experiment approach is detailed in Figure 2.
Figure 2 Example illustrating the application of the a) ±0.05m^3 m^(-3) and e) ±0.1m^3 m^(-3)uncertainty range global perturbation together with histograms for prior ensembles generated for the b) 5 cm, c) 40 cm and d) 80 cm layers (last simulation timestep).
Figure 3 Data assimilation scheme (Particle Filter flow diagram) utilizing a pre-compiled look-up table
Figure 4a) Taylor diagram showing comparison of COSMIC derived soil water with those from equal, conventional and revised weighting approaches over the August to October 2016 period; b) cumulative weights over depth (COSMIC compared to uniform, conventional (linear and non-linear vertical), revised approaches) [Observed SWC-11·10·2016 23:45].
Figure 5 a) Neutron counts timeseries: simulated open loop, after the implementation of Sequential Importance Resampling-Particle Filter (with ±0.1m^3 m^(-3) and ±0.05m^3 m^(-3) uncertainty ranges) and observed; b) open loop; c) analysis ±0.05m^3 m^(-3); and d) analysis ±0.1m^3 m^(-3) scatter plots; e) SIC averages from the different scenarios, for qualitative comparison with the reference. SIC best estimate as derived in section 3.4 was also added for comparison.
Figure 6 : Posterior, prior and likelihood histograms for the last timestep (neutron counts and average TSWC, respectively): a) ±0.05m^3 m^(-3) and b) ±0.1m^3 m^(-3) uncertainty ranges.
Figure 7 a) Neutron counts timeseries for: analyses utilising LUT model generated particles, using perturbed and unperturbed likelihood, open loop and CRNP observations; b) Average TSWC (inferred from neutron counts analyses – using perturbed likelihood), USWC and SIC best estimates.


Leave a reply

Leave a Reply

Your email address will not be published.