Streamflow Prediction in the Mekong River Basin Using Deep Neural Networks

Nguyen, Thi-Thu-Ha; Vu, Duc-Quang; T., Son; Dang, Thanh Duc

doi:10.1109/access.2023.3301153

Cited by 2 publications

(24 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, we introduce a novel hybrid TNN model for indirectly estimating RZSM, with the objective of enhancing PI decision-making. To evaluate the proposed model's performance, we will conduct a comparative analysis against hybrid-GRU [27], LSTM-Technique [32], and Multi-head TNN [25] within an irrigation framework. Both hybrid-GRU [27] and LSTM-Technique [32] predict RZSM using a sequential DL model, incorporating multiple auxiliary-RZSM variables as input to make informed irrigation decisions.…”

Section: Methodsmentioning

confidence: 99%

“…Consequently, the more advanced DL-based techniques for detailed SSM information, such as perform downscaling (increasing spatial/temporal details) by correlating the SSM with auxiliary SSM information available at target location [20] - [23], can be extended to RZSM. Stream flow, another significant environmental parameter, exhibits remarkable similarities with moisture distribution within the root zone [24], [25]. Both parameters offer insights into the time-series movement of water, albeit within different mediums.…”

Section: Introductionmentioning

confidence: 99%

“…Both parameters offer insights into the time-series movement of water, albeit within different mediums. Just as auxiliary information-based DL models have proven effective in the SSM domain, we can apply similar knowledge and methodologies from existing DL research in stream flow to enhance our comprehension of RZSM [25].…”

Section: Introductionmentioning

confidence: 99%

“…The transformer neural network (TNN) is the state-of-theart DL model to process sequential data [25], [28]. It is a new cognitive model eschewing recurrence implementations in the previous benchmark sequential DL models, such as Long short-term memory (LSTM) [29] and Gated recurrent unit (GRU) [30].…”

Section: Introductionmentioning

confidence: 99%

“…Based on the hybrid TNN RZSM estimates, we simulated PI schedules for the test locations for further evaluation. The proposed hybrid TNN model is compared with TNN [25], LSTM [27] and GRU [32] based related works for performance evaluation. A comparison of irrigation scheduling with these three models using quantitative measurements is conducted for the amount of water saved.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Hybrid Transformer Network for Soil Moisture Estimation in Precision Irrigation

Madhukumar,

Wang,

Everingham

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Accurate root zone soil moisture (RZSM) estimation is essential for precision irrigation (PI) systems that seek to optimize water use efficiency. Large-scale in-situ sensors for direct measurement are costly, while existing satellites lack depth resolution for direct RZSM data. Hence, in-direct RZSM estimation methods are required. Literature illustrates that RZSM at a location is related to changing soilwater-plant characteristics. Therefore, these characteristics can provide auxiliary information on RZSM changes. By leveraging auxiliary information derived from changing soil-water-plant characteristics, this paper enables indirect RZSM estimation at non-sensor locations, effectively addressing the limitations inherent in direct RZSM measurement techniques initially discussed. Compared to existing methods, deep learning (DL) is most suitable for such data associations as they are auto-tuned to extract relative relationships from diverse big data. Among DL models, sequential models are apt for finding these relationships as all these variables are time-series sequences. The transformer neural network (TNN) is the state-of-the-art DL model for analyzing sequences. However, for in-direct RZSM estimation, the data associations within a location and multiple sensor sites with the target need to be found. Conventional TNN cannot incorporate such simultaneous multi-associations, hence, we develop a new TNN model called the hybrid TNN model, which is able to facilitate the capturing of complicated dependencies through thoughtful feature selection and engineering. First, sensor locations exhibiting analogous downscaled 1-km satellite soil moisture (SM) are identified. Next, a dynamic multilayer perceptron (D-MLP) network discerns highly correlated auxiliary-RZSM data, utilizing both ground-based and downscaled satellite data. Following this, the dual attention module identifies essential multi-associations, leveraging selected sensor and target region information. Finally, the Bayesian layer averages multi-location RZSM using the conditional probability generated based on relative relationships to yield the target location RZSM estimate. Our proposed model shows 13.066% better RZSM estimation compared to popular sequential models. The hybrid TNN RZSM estimates are used to monitor root water depletion tolerance levels for optimal PI schedules, which shows 10.846% water and 10.339% cost-saving on our selected sites. Overall the proposed model effectively demonstrates that a more accurate PI predictive algorithm saves water, improves resource conservation, and reduces irrigation costs.INDEX TERMS Agriculture, deep learning, hybrid transformer neural network, precision irrigation, root zone soil moisture.

show abstract

Section: Methodsmentioning

confidence: 99%