Bayesian convolutional neural networks for predicting the terrestrial water storage anomalies during GRACE and GRACE-FO gap

Mo, Shaoxing; Zhong, Yulong; Forootan, Ehsan; Mehrnegar, Nooshin; Yin, Xiong; Wu, Jichun; Wang, Feng; Shi, Xinzheng

doi:10.1016/j.jhydrol.2021.127244

Cited by 54 publications

(54 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our results agree with the learning methods that incorporated hydroclimatic data to reconstruct the gap in terms of CC and NSE and outperform them in term of RMSE. Specifically, the results are more consistent with DNN and Bayesian CNN (BCNN) reconstructions (Mo et al, 2022;Sun et al, 2020), than SWARM driven, statistical learning driven (ANN, ARX, MLR) outputs from (Li et al, 2020). The training periods are similar in both studies (04/2002-01-06/2014) (Mo et al, 2022;Sun et al, 2020), but the testing period of performance is different: 04/2014 to 06/2017 (Sun et al, 2020), and 04/2014-06/2017, and 01/2018-09/2020 (Mo et al, 2022).…”

Section: Comparison With Other Reconstructed Datasupporting

confidence: 57%

“…We compared our reconstructed data with data from four other GRACE reconstructions (Li et al., 2020; Mo et al., 2022; Sun et al., 2020). These studies were concerned with filling the gap between the missions only.…”

Section: Resultsmentioning

confidence: 99%

“…For the missing solutions within the missions, linear or spline interpolations are performed to fill these gaps, without proper justification, or constraining the uncertainties in these solutions. For the gap between the two missions, different approaches have been used to fill the missing data and reconstruct the gap, and can be summarized as follows: For the ice sheets, the independent mass balance changes were incorporated with GRACE (‐FO) observations to infer the status of the ice loss between 2017 and 2018 (Velicogna et al., 2020); For the land hydrology, where the most of the reconstruction studies focused on, the total water storage (TWS), which represents the sum of all storage compartments (snow water equivalent, surface water, soil water, groundwater, and ice) was reconstructed using Global Positioning Systems (GPS) (Rietbroek et al., 2014), or informed using hydroclimate predictors within statistical learning framework for the first mission (Li et al., 2020; Li, Wang, et al., 2019; Mo et al., 2022; Sun et al., 2021, 2020; Wang et al., 2021; Yang et al., 2021). Assimilating GRACE into land surface modeling (e.g., CLSM‐F2.5) is another approach to fill the gap in GRACE (‐FO) over land (Li, Rodell, Kumar, et al., 2019).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Reconstruction of GRACE Mass Change Time Series Using a Bayesian Framework

Rateb

Sun

Scanlon

et al. 2022

Earth and Space Science

View full text Add to dashboard Cite

Since 2002, the Gravity Recovery And Climate Experiment (GRACE) and its Follow-On (FO) missions have resulted in a paradigm shift in our understanding of the changes in the Earth's gravity field and associated regional and global mass change and transport (Abich et al., 2019;Landerer et al., 2020;Tapley et al., 2004). GRACE (-FO) missions provide monthly observations of mass changes caused by land water storage (Rodell et al., 2018;Scanlon et al., 2019), ocean currents (Adhikari et al., 2019;Chambers, 2006), ice sheets, and glaciers (Velicogna et al., 2020) which enable us to constrain the dynamics of these systems and assess impacts of climate change and human interventions (

show abstract

Section: Comparison With Other Reconstructed Datasupporting

confidence: 57%

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Reconstruction of GRACE Mass Change Time Series Using a Bayesian Framework

Rateb

Sun

Scanlon

et al. 2022

Earth and Space Science

View full text Add to dashboard Cite

show abstract

“…Drought Condition Classification Based on the Water Storage Deficit Index C3. Comparison of the testing Nash-Sutcliffe efficiency coefficient (NSE) accuracy obtained in previous study (Mo et al, 2022) and this study for (a-c) the terrestrial water storage anomaly (TWSA) and (d-f) the water storage deficit (WSD) signals. The major differences between Mo et al (2022) and this study are as follows: (a) Mo et al (2022) used only four predictors (simulated/reanalyzed TWSA, cumulative water storage change, P, and T) derived from the ERA5 land data set without utilizing the data source selection strategy; (b) the Jet Propulsion Laboratory Gravity Recovery and Climate Experiment (GRACE) mascon product was used in Mo et al (2022), while a weighted average GRACE product is used here; and (c) The Bayesian Convolutional Neural Network method employed in this study has been improved (see Appendix A).…”

Section: Table C2mentioning

confidence: 91%

Hydrological Droughts of 2017–2018 Explained by the Bayesian Reconstruction of GRACE(‐FO) Fields

Zhong

Forootan

et al. 2022

Water Resources Research

Self Cite

View full text Add to dashboard Cite

Drought is an event of prolonged water shortages and usually associated with long-standing (months or even years) and devastating impacts on ecosystem, agriculture, and society (Hao et al., 2018;X. He, Estes, et al., 2019;Van Loon, 2015). To better manage these events and alleviate their impacts, it is extremely essential to understand the characteristics of historical drought events well, including their duration, intensity, affected areas, severity, and recovery process. This information is useful for developing proper water resources management policies to tackle challenges that likely occur in future (

show abstract

“…In contrast, mathematical and statistical approaches, although containing no physical information, were also proven to reasonably fill the GRACE and GRACE-FO gaps [13]. To include physical information into TWS reconstruction, mathematical algorithms were either modified with meteorological parameters [14,15] or trained with SLR data [16], with only GRACE data [17], with GRACE and Swarm data [18], or with a combination of satellite data with meteorological/hydrological parameters [19,20], and were demonstrated to correlate well with GRACE [21,22]. The main advantage of mathematical approaches is that they are able to predict climate signals such as droughts [23] or floods [24] to reconstruct signals of the El Niño oscillation [22] or human-induced TWS changes [25] at any spatial resolution.…”

Section: Introductionmentioning

confidence: 99%

Autoregressive Reconstruction of Total Water Storage within GRACE and GRACE Follow-On Gap Period

et al. 2022

View full text Add to dashboard Cite

For 15 years, the Gravity Recovery and Climate Experiment (GRACE) mission have monitored total water storage (TWS) changes. The GRACE mission ended in October 2017, and 11 months later, the GRACE Follow-On (GRACE-FO) mission was launched in May 2018. Bridging the gap between both missions is essential to obtain continuous mass changes. To fill the gap, we propose a new approach based on a remove–restore technique combined with an autoregressive (AR) prediction. We first make use of the Global Land Data Assimilation System (GLDAS) hydrological model to remove climatology from GRACE/GRACE-FO data. Since the GLDAS mis-models real TWS changes for many regions around the world, we further use least-squares estimation (LSE) to remove remaining residual trends and annual and semi-annual oscillations. The missing 11 months of TWS values are then predicted forward and backward with an AR model. For the forward approach, we use the GRACE TWS values before the gap; for the backward approach, we use the GRACE-FO TWS values after the gap. The efficiency of forward–backward AR prediction is examined for the artificial gap of 11 months that we create in the GRACE TWS changes for the July 2008 to May 2009 period. We obtain average differences between predicted and observed GRACE values of at maximum 5 cm for 80% of areas, with the extreme values observed for the Amazon, Alaska, and South and Northern Asia. We demonstrate that forward–backward AR prediction is better than the standalone GLDAS hydrological model for more than 75% of continental areas. For the natural gap (July 2017–May 2018), the misclosures in backward–forward prediction estimated between forward- and backward-predicted values are equal to 10 cm. This represents an amount of 10–20% of the total TWS signal for 60% of areas. The regional analysis shows that the presented method is able to capture the occurrence of droughts or floods, but does not reflect their magnitudes. Results indicate that the presented remove–restore technique combined with AR prediction can be utilized to reliably predict TWS changes for regional analysis, but the removed climatology must be properly matched to the selected region.

show abstract

Bayesian convolutional neural networks for predicting the terrestrial water storage anomalies during GRACE and GRACE-FO gap

Cited by 54 publications

References 58 publications

Reconstruction of GRACE Mass Change Time Series Using a Bayesian Framework

Reconstruction of GRACE Mass Change Time Series Using a Bayesian Framework

Hydrological Droughts of 2017–2018 Explained by the Bayesian Reconstruction of GRACE(‐FO) Fields

Autoregressive Reconstruction of Total Water Storage within GRACE and GRACE Follow-On Gap Period

Contact Info

Product

Resources

About