Spatial–temporal modeling of root zone soil moisture dynamics in a vineyard using machine learning and remote sensing

Abstract

High-resolution spatial–temporal root zone soil moisture (RZSM) information collected at different scales is useful for a variety of agricultural, hydrologic, and climate applications. RZSM can be estimated using remote sensing, empirical equations, or process-based simulation models. Machine learning (ML) approaches for evaluating RZSM across numerous spatial–temporal scales are less generalizable than process-based models. However, data-driven ML approaches offer a unique opportunity to develop complex models of soil moisture without making assumptions about the processes governing soil water dynamics in a given study region. In this study, comparisons were made between two models, pySEBAL and EFSOIL, which were based on evaporation fraction (EF) and soil properties, and a data-driven model based on the Random Forest (RF) ensemble algorithm. These approaches were evaluated to demonstrate their capabilities for RZSM estimation. The EF obtained from Landsat images was used after validation with eddy covariance measurements as the major input to all three models, along with other meteorological and soil physical properties. The RF model was trained using in situ soil moisture data from Time Domain Reflectometry (TDR) sensors installed in a vineyard from 2018 to 2020. The predictor variables comprised of meteorological, soil properties, EF, and a vegetation index. The results reveal that there was a strong correlation between the in situ measured soil moisture and the RF predicted soil moisture at all sensor locations. Due to the complexity of the physical processes involved in soil water flow, the empirical models pySEBAL and EFSOIL were unable to reliably predict RZSM values at all monitored locations. The high RZSM predicted by pySEBAL demonstrated the presence of possible bias in the model’s algorithm used to estimate soil moisture. We also demonstrated that ML based on the RF algorithm may be used to predict spatially distributed RZSM when a few soil moisture ground measurements are combined with remote sensing to produce soil moisture maps.

Empirical analysis of machine learning-based moisture sensing platforms for agricultural applications: A statistical perspective

“…Based on this estimation, statistical evaluation of these models is tabulated in Based on this comparison, it can be observed that different models are useful for different application requirements. For instance, in terms of accuracy as observed from figure 5, DBN RBM [2], CRNS [3], SMAP RF DN [19], GOFCHS [27], TDR [28], and P Band & L Band [34] models outperform other models, thus, they can be used for highly accurate moisture detection applications. Similarly, cost of deployment & computational complexity is visualized from figure 6, wherein it is observed that HPCM [6], HF RFID TFS [9], PWM [10], PMMA [15], FFCSM [16], MHPS [21], ECT [24], PQCWC [25], and HSAAA [32] require lowest deployment cost, while HPCM [6], PHS [17], ECT [24], and PQCWC [25] have lower computational complexity when compared with other models.…”

“…While highly accurate sensing interfaces are costly, but can be used for high-speed moisture sensing applications. Specifically, DBN RBM, CRNS, SMAP LEWS LR [1] DBN RBM [2] LR RBM [2] BP RBM [2] CRNS [3] FBG [4] MWMS [5] HPCM [6] GPS [7] UHF RFID [8] HF RFID TFS [9] PWM [10] CM [11] PPMR [12] PLMR [12] RFID UHF [13] MSR [14] PMMA [15] FFCSM [16] PHS [17] SMAP [18] SMAP RF DN [19] FoS [20] MHPS [21] PRS [22] ECT [24] PQCWC [25] HDES [26] GOFCHS [27] TDR [28] SAR [29] SMI MODIS [30] CSMOS [31] HSAAA [32] SSMDI [33] P Band & L Band [34] FTO [35] eSMAP [36] PBG [38] MSNs [39] MSOCCML [40] SMAP TFC [41] CRNS [42] kCRNS [43] Computational RF DN, GOFCHS, TDR, and P Band & L Band models outperform other models; thus, they can be used for highly accurate moisture detection applications. While, HPCM, HF RFID TFS, PWM, PMMA, FFCSM, MHPS, ECT, PQCWC, ...…”

“…The model performance is limited in terms of accuracy, but can be used for approximating moisture levels for a wide variety of applications. To improve this performance, work in [26,27,28] proposes use of Internet of Trees (IoTr) with low cost & highly dispersive electromagnetic sensors (HDES), Graphene oxide for design of fibre core-offset humidity sensor (GOFCHS), and Time Domain Reflectometry (TDR) for low error, and high-accuracy moisture level estimations. These models utilize various machine learning techniques to improve quality of detection, but require high performance computational processors for effective implementation & control.…”

From vine to vineyard: the GRAPEX multi-scale remote sensing experiment for improving vineyard irrigation management

“…For root zone soil moisture (RZSM) estimation, the study by Kisekka et al (2022) compared two models: (1) a remote sensing-based approach, pySEBAL and EFSOIL; and (2) a data-driven model based on machine learning. The pySE-BAL and EFSOIL approach was unable to reliably predict RZSM at all monitored locations, while the machine learning model trained with in situ soil moisture data combined with meteorological, soil properties, EF (evaporative fraction), and a vegetation index has the potential to estimate spatially distributed RZSM combined with remote sensing information.…”

“…Notable are the papers by Kisekka et al 2022 and unpublished 20 that span the full spatial scale at the theoretical and practical end of the spectrum, respectively. This highlights the potential for combining remote sensing data with ground observations to bridge scales from plant to landscape and region.…”