Stream-water temperature is a key variable controlling chemical, biological, and ecological processes in freshwater environments. Most models focus on a single river cross-section; however, temperature gradients along stretches and tributaries of a river network are crucial to assess ecohydrological features such as aquatic species suitability, growth and feeding rates, or disease transmission. We propose SESTET, a deterministic, spatially explicit stream temperature model for a whole river network, based on water and energy budgets at a reach scale and requiring only commonly available spatially distributed datasets, such as morphology and air temperature, as input. Heat exchange processes at the air–water interface are modelled via the widely used equilibrium temperature concept, whereas the effects of network structure are accounted for through advective heat fluxes. A case study was conducted on the prealpine Wigger river (Switzerland), where water temperatures have been measured in the period 2014–2018 at 11 spatially distributed locations. The results show the advantages of accounting for water and energy budgets at the reach scale for the entire river network, compared with simpler, lumped formulations. Because our approach fundamentally relies on spatially distributed air temperature fields, adequate spatial interpolation techniques that account for the effects of both elevation and thermal inversion in air temperature are key to a successful application of the model. SESTET allows the assessment of the magnitude of the various components of the heat budget at the reach scale and the derivation of reliable estimates of spatial gradients of mean daily stream temperatures for the whole catchment based on a limited number of conveniently located (viz., spanning the largest possible elevation range) measuring stations. Moreover, accounting for mixing processes and advective fluxes through the river network allows one to trust regionalized values of the parameters controlling the relationship between equilibrium and air temperature, a key feature to generalize the model to data-scarce catchments. 相似文献
We invert three-dimensional seismic data by a multiscale phase inversion scheme, a modified version of full waveform inversion, which applies higher order integrations to the input signal to produce low-boost signals. These low-boost signals are used as the input data for the early iterations, and lower order integrations are computed at the later iterations. The advantages of multiscale phase inversion are that it (1) is less dependent on the initial model compared to full waveform inversion, (2) is less sensitive to incorrectly modelled magnitudes and (3) employs a simple and natural frequency shaping filtering. For a layered model with a three-dimensional velocity anomaly, results with synthetic data show that multiscale phase inversion can sometimes provide a noticeably more accurate velocity profile than full waveform inversion. Results with the Society of Exploration Geophysicists/European Association of Geoscientists and Engineers overthrust model shows that multiscale phase inversion more clearly resolves meandering channels in the depth slices. However, the data and model misfit functions achieve about the same values after 50 iterations. The results with three-dimensional ocean-bottom cable data show that, compared to the full waveform inversion tomogram, the three-dimensional multiscale phase inversion tomogram provides a better match to the well log, and better flattens angle-domain common image gathers. The problem is that the tomograms at the well log provide an incomplete low-wavenumber estimate of the log's velocity profile. Therefore, a good low-wavenumber estimate of the velocity model is still needed for an accurate multiscale phase inversion tomogram. 相似文献
In downhole microseismic monitoring, accurate event location relies on the accuracy of the velocity model. The model can be estimated along with event locations. Anisotropic models are important to get accurate event locations. Taking anisotropy into account makes it possible to use additional data – two S-wave arrivals generated due to shear-wave splitting. However, anisotropic ray tracing requires iterative procedures for computing group velocities, which may become unstable around caustics. As a result, anisotropic kinematic inversion may become time consuming. In this paper, we explore the idea of using simplified ray tracing to locate events and estimate medium parameters. In the simplified ray-tracing algorithm, the group velocity is assumed to be equal to phase velocity in both magnitude and direction. This assumption makes the ray-tracing algorithm five times faster compared to ray tracing based on exact equations. We present a set of tests showing that given perforation-shot data, one can use inversion based on simplified ray-tracing even for moderate-to-strong anisotropic models. When there are no perforation shots, event-location errors may become too large for moderately anisotropic media. 相似文献
In this paper, we develop a general approach to integrating petrophysical models in three-dimensional seismic full-waveform inversion based on the Gramian constraints. In the framework of this approach, we present an example of the frequency-domain P-wave velocity inversion guided by an electrical conductivity model. In order to introduce a coupling between the two models, we minimize the corresponding Gramian functional, which is included in the Tikhonov parametric functional. We demonstrate that in the case of a single-physics inversion guided by a model of different physical type, the general expressions of the Gramian functional and its gradients become simple and easy to program. We also prove that the Gramian functional has a non-negative quadratic form, so it can be easily incorporated in a standard gradient-based minimization scheme. The developed new approach of seismic inversion guided by the known petrophysical model has been validated by three-dimensional inversion of synthetic seismic data generated for a realistic three-dimensional model of the subsurface. 相似文献
Markov chain Monte Carlo algorithms are commonly employed for accurate uncertainty appraisals in non-linear inverse problems. The downside of these algorithms is the considerable number of samples needed to achieve reliable posterior estimations, especially in high-dimensional model spaces. To overcome this issue, the Hamiltonian Monte Carlo algorithm has recently been introduced to solve geophysical inversions. Different from classical Markov chain Monte Carlo algorithms, this approach exploits the derivative information of the target posterior probability density to guide the sampling of the model space. However, its main downside is the computational cost for the derivative computation (i.e. the computation of the Jacobian matrix around each sampled model). Possible strategies to mitigate this issue are the reduction of the dimensionality of the model space and/or the use of efficient methods to compute the gradient of the target density. Here we focus the attention to the estimation of elastic properties (P-, S-wave velocities and density) from pre-stack data through a non-linear amplitude versus angle inversion in which the Hamiltonian Monte Carlo algorithm is used to sample the posterior probability. To decrease the computational cost of the inversion procedure, we employ the discrete cosine transform to reparametrize the model space, and we train a convolutional neural network to predict the Jacobian matrix around each sampled model. The training data set for the network is also parametrized in the discrete cosine transform space, thus allowing for a reduction of the number of parameters to be optimized during the learning phase. Once trained the network can be used to compute the Jacobian matrix associated with each sampled model in real time. The outcomes of the proposed approach are compared and validated with the predictions of Hamiltonian Monte Carlo inversions in which a quite computationally expensive, but accurate finite-difference scheme is used to compute the Jacobian matrix and with those obtained by replacing the Jacobian with a matrix operator derived from a linear approximation of the Zoeppritz equations. Synthetic and field inversion experiments demonstrate that the proposed approach dramatically reduces the cost of the Hamiltonian Monte Carlo inversion while preserving an accurate and efficient sampling of the posterior probability. 相似文献