Enter Zoom Meeting

HS3.3

Spatio-temporal and/or (geo)statistical analysis of hydrological events, floods, extremes, and related hazards

Many environmental and hydrological problems are spatial or temporal, or both in nature. Spatio-temporal analysis allows identifying and explaining large-scale anomalies which are useful for understanding hydrological characteristics and subsequently predicting hydrological events. Temporal information is sometimes limited; spatial information, on the other hand has increased in recent years due technological advances including the availability of remote sensing data. This development has motivated new research efforts to include data in model representation and analysis.

Statistics are in wide use in hydrology for example to estimate design events, forecast the risk and hazard of flood events, detect spatial or temporal clusters, model non-stationarity and changes and many more. Statistics are useful in the case when only few data are available but information for very rare events (extremes) or long time periods are needed. They are also helpful to detect changes and inconsistencies in the data and give a reliable statement on the significance. Moreover, temporal and spatial changes often lead to the violation of stationarity, a key assumption of many standard statistical approaches. This makes hydrological statistics interesting and challenging for so many researchers.

Geostatistics is the discipline that investigates the statistics of spatially extended variables. Spatio-temporal analysis is at the forefront of geostatistical research these days, and its impact is expected to increase in the future. This trend will be driven by increasing needs to advance risk assessment and management strategies for extreme events such as floods and droughts, and to support both short and long-term water management planning. Current trends and variability of hydrological extremes call for spatio-temporal and/or geostatistical analysis to assess, predict, and manage water related and/or interlinked hazards.

The aim of this session is to provide a platform and an opportunity to demonstrate and discuss innovative applications and methodologies of spatio-temporal analysis in a hydrological (hydrometeorological) context. The session is targeted at both hydrologists and statisticians interested in the spatial and temporal analysis of hydrological events, extremes, and related hazards, and it aims to provide a forum for researchers from a variety of fields to effectively communicate their research.

Convener: Yunqing Xuan | Co-conveners: Emmanouil Varouchakis, Gerald A Corzo P, Vitali DiazECSECS, Francisco Munoz-Arriola, Adrian Almoradie
Presentations
| Wed, 25 May, 13:20–15:55 (CEST)
 
Room 2.15

Wed, 25 May, 13:20–14:50

Chairpersons: Yunqing Xuan, Emmanouil Varouchakis, Vitali Diaz

13:20–13:30
|
EGU22-13051
|
ECS
|
solicited
|
Virtual presentation
Panayiotis Dimitriadis et al.

Long-range dependence (LRD) estimators are traditionally applied in the lag domain (e.g., through the autocovariance or variogram) or in the frequency domain (e.g., through the power-spectrum), but not as often in the scale domain (e.g., through variance vs. scale). It has been contended that the latter case introduces large estimation bias and thus, corresponds to "bad estimators" of the LRD. However, this reflects a misrepresentation or misuse of the concept of variance vs. scale. Specifically, it is shown that if the LRD estimator of variance vs. scale is properly defined and assessed (see literature studies for the so-called climacogram estimator), then the stochastic analysis of variance in the scale domain can be proven to be a robust means to identify and model any LRD process ranging from small scales (fractal behavior) to large scales (LRD, else known as the Hurst-Kolmogorov dynamics) for any marginal distribution. Here, we show how the above definitions can be applied both in spatial and temporal scales, with various applications in geophysical processes, key hydrological-cycle processes, and related natural hazards.

How to cite: Dimitriadis, P., Iliopoulou, T., Sargentis, G.-F., and Koutsoyiannis, D.: Spatial and temporal long-range dependence in the scale domain, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-13051, https://doi.org/10.5194/egusphere-egu22-13051, 2022.

13:30–13:37
|
EGU22-4640
|
ECS
|
Virtual presentation
Gábor Szatmári et al.

Eutrophication, water quality and environmental status of lakes is a global issue that depends not only on the quality and quantity of nutrients stored in lake sediments but also on their relative content. On the example of Lake Balaton (Hungary), we jointly modelled the spatial distribution of the nutrients nitrogen and phosphorus, and their ratio (i.e. nitrogen to phosphorus ratio) in the sediments of the lake and then provided spatial predictions at different supports (i.e. point, basin and entire lake) with the associated prediction uncertainty. The objective of our study was to illustrate the merits of applying multivariate geostatistics when spatial modelling of more than one variable is targeted at various scales in water ecosystems. Exploratory variography confirmed that there is a spatial interdependency between the nutrients and therefore it is better to jointly model their spatial distribution. The results revealed that by the application of multivariate geostatistics the spatial interdependency existing between the nutrients under study can be explicitly taken into account and exploited in the course of spatial modelling to provide coherent and more accurate spatial predictions that could support the complex assessment of the water quality and environmental status of Lake Balaton. Besides, stochastic realizations reproducing the joint spatial variability of the two nutrients can be generated that allow to compute stochastic realizations of their ratio, furthermore, to provide spatially aggregated predictions for larger supports (e.g. basins or entire lake) with the associated prediction uncertainty, which may be better fit to the end-users' demands on spatially explicit information about sediment nutrients. Our study highlighted that it is worthy of applying multivariate geostatistics in case spatial modelling of two or more variables, which jointly vary in space, is targeted in water ecosystems.

 

Acknowledgements: Our research was funded by the National Research Development and Innovation Office (NKFIH), grant number K-131820. The work of Gábor Szatmári was supported by the Premium Postdoctoral Scholarship of the Hungarian Academy of Sciences (PREMIUM-2019-390).

How to cite: Szatmári, G., Kocsis, M., Makó, A., Pásztor, L., and Bakacsi, Z.: Joint spatial modelling of sediment nutrients and their ratio in Lake Balaton (Hungary) using multivariate geostatistics, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4640, https://doi.org/10.5194/egusphere-egu22-4640, 2022.

13:37–13:44
|
EGU22-6244
|
Virtual presentation
Xiaoyi Wang

The prediction skill of the S2S period (usually referred to as two weeks but less than a season) always suffers from the integrated influence between weather and climate. Although S2S prediction has attracted more attention than before and multiple efforts have prompted related studies, the accuracy of surface soil moisture prediction in the dry condition is generally inferior to that of other variables in the S2S period. This study proposes a framework combing an ensemble empirical mode decomposition (EEMD) and Multilayer perceptron (MLP) to predict the surface soil moisture (0-7 cm) in two drought-prone regions. The proposed method has been verified and optimized in the Netherlands and Spain by using hindcasts driven by ERA5 reanalysis which can be regarded as a proxy for the real weather. The concrete practice consists of 1) calculating intrinsic mode functions (IMFs) collection and their residual components of selected ERA5-Land variables that are sensitive to surface soil moisture after data pre-processing, 2) similar IMFs curves classifications, 3) further feature selection according to classified sets, and 4) performing IMFs-driven MLP daily predictors and integrating the predicted IMFs and residual components to obtain the predicted surface soil moisture. The positive results show that this framework can be served as a regional S2S forecasting approach based on ERA5-Land reanalysis data, and with an expected daily ERA5-Land update of 5 days behind real time in 2022 instead of the current 3-month latency, the employed hybrid model is anticipated to explore realistic hydrological and agricultural applications.

How to cite: Wang, X.: Surface soil moisture in sub-seasonal to seasonal (S2S) prediction driven by a hybrid model over drought-prone regions, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6244, https://doi.org/10.5194/egusphere-egu22-6244, 2022.

13:44–13:51
|
EGU22-14
|
ECS
|
Virtual presentation
Behnam Sadeghi et al.

A common problem in geochemical exploration projects is the limited number of collected samples due to budgetary, time, and other constraints. Therefore, to study spatial mineralisation patterns using available samples in both sampled and unsampled areas, the interpolation of the available data is essential to assign estimates to unsampled areas. Because interpolation estimates are based on the data available only within the search window, in continuous field geochemical modelling such interpolations using any single method are often the main source of uncertainty. Error propagation analysis needs to be considered to evaluate interpolation errors’ effects in geochemical anomaly detection. One method for analysing the propagation of errors in models and evaluating their stability is Monte Carlo Simulation (MCSIM). In this method, the P50 (median) value (called ‘return’) and the uncertainty value (called ‘risk’) are calculated. Here the uncertainty is calculated as 1/(P90-P10) for which P10 (lower decile) and P90 (upper decile) are the average 10th and 90th percentiles of the multiple simulated values, correspond to each element. We have applied this method to Swedish till data, collected throughout the country by the Geological Survey of Sweden. The main concern is whether to evaluate if the samples are sufficient and representative of the target elements concentrations for geochemical studies. To address this concern, the sampling uncertainty in a statistical sense (not geochemical) per element was studied using the return-risk matrix. This matrix was applied to volcanogenic massive sulfide (VMS) target elements, then subsequently to the samples per bedrock. Therefore, a large number of simulated values (e.g., 5,000, which is higher than the number of the samples, i.e., 2,578) was generated using MCSIM. Where the quantified return is low or negative, and the quantified uncertainty is high, particularly higher than its relevant return, additional sampling is required to achieve the minimum required spatial continuity in the data or the stability of the later applied classification models. This affects the certainty of the models generated in the study area. In the Sweden data, all the elements assessed have relatively high returns and low uncertainty, demonstrating the stability of the parameters. The process was subsequently applied to samples separated into the main lithological categories or geological domains to determine if the stability in the patterns is affected by rock type (and associate natural variability in the background). In Swedish till samples, the statistical sampling quality is acceptable in the bedrocks of Exotic Terranes, Archean, Baltoscandian, and Idefjorden. However, it is not acceptable in the Palaeoproterozoic units and the Eastern Segment, due to the risks being higher than the returns, which may increase the error propagation effect on the interpolated map and efficiency of the classes obtained by different classification models.

How to cite: Sadeghi, B., Cohen, D., and Müller, D.: Improved decision-making in geochemical sampling based on both frequency and Bayesian frameworks, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-14, https://doi.org/10.5194/egusphere-egu22-14, 2022.

13:51–13:58
|
EGU22-5381
|
ECS
|
Highlight
|
|
On-site presentation
Sergio Redondo et al.

Floods are a major threat to the security of populations worldwide. Their impact is dependent on the flood extension and water levels,  topographic factors, and many other variables. Meteorological factors such as precipitation, snow melt and base flow also influence the magnitude of flood events. Factors like river morphology, slopes, presence of flood plains, vegetation and soil types determine the response of the river or lake to meteorological conditions and hydrological events. Meteorological factors tend to be cyclical while topography is generally considered to remain constant over extended periods of time. However, river bathymetry is subject to changes over time due to sediment transport. For example, sediment re-positioning and accumulation can modify the bathymetry of water bodies.  Furthermore, lakes or reservoirs receive and accumulate sediments transported from upstream which could influence flood levels. In this study we use a two-dimensional hydraulic model (Telemac2D) to simulate different flood scenarios coupled with several different bathymetries of Trois-Lacs Lake in the province of Quebec, Canada. Four bathymetries were obtained between November 2020 and August 2021 and 3 historical bathymetries were also provided (yrs, 1974, 2004, and 2019).  To compare the bathymetries, total ‘available water volume’ is calculated, taking a common and constant reference water surface elevation. Streamflows entering the lake system were estimated using Hydrotel, a physics-based semi distributed hydrological model. These streamflows are used to calibrate the two-dimensional hydraulic model with measured water levels.  The project may help to establish a direct relation between sediment shifting and deposition and water distribution for extreme flood events, while also allowing the local community to improve measures for civil security and land-use planning at a regional scale.

How to cite: Redondo, S., Boucher, M.-A., Lacey, J., and Parent, J.: Quantifying the impact of bathymetry changes on flood  events for the Trois-Lacs Basin, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5381, https://doi.org/10.5194/egusphere-egu22-5381, 2022.

13:58–14:05
|
EGU22-6992
|
ECS
|
Virtual presentation
Ruili Fu et al.

An extreme wave always evolves from a wave group in random wave trains. Therefore, better insights into extreme wave groups are crucial for the safety design of marine structures. In the present work, the marginal and bivariate distributions of extreme wave group energy and duration are investigated based on the field datasets from Norway's North Sea. The most probable extreme wave group energy and duration can be obtained based on the distributions, then evolutions of wave shapes of extreme wave groups are investigated and compared with the present extreme wave group theories. It is found that the wave shapes are asymmetry with time-spatial evolution.

How to cite: Fu, R., Zheng, J., Tao, A., and Wang, G.: Statistical properties of extreme wave groups based on field data, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6992, https://doi.org/10.5194/egusphere-egu22-6992, 2022.

14:05–14:12
|
EGU22-11000
|
ECS
|
Virtual presentation
Monique Rodrigues da Silva Andrade Maia et al.

The element phosphorus total (Pt) is considered a basic element for life on earth. It controls key processes of CO2 absorption from tropical forests to food production. For the Amazon region, estimates of Pt in the soil are scarce. In this study, we developed models through equations of pedotransfer function (PTF's) using data collected in the field (RAINFOR data). Were generated 16 regression models based on the Akaike information criteria (AIC) with R² above 65% was validated with independent RAINFOR data. The results Pt distribution were spatialized using interpolations by geostatistical method of inverse distance weights (IDW) and shown through maps.

How to cite: Maia, M. R. D. S. A., Anderson, L. O., and Quesada, C. A. N.: Estimates of total phosphorus for Amazonia based on an expanded harmonized soil database, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11000, https://doi.org/10.5194/egusphere-egu22-11000, 2022.

14:12–14:19
|
EGU22-11644
|
On-site presentation
Emmanouil Varouchakis et al.

A geostatistical analysis based on a machine learning method was conducted to generate reliable spatial maps of groundwater level variability and to identify groundwater level patterns over the island of Crete, Greece. Geostatistics plays an important role in model-related data analysis and preparation, but has specific limitations when the aquifer system is inhomogeneous. Self-Organizing Maps can be applied to identify locally similar input data and then by means of Ordinary Kriging to estimate the spatial distribution of groundwater level. The proposed methodology was tested on a large dataset of groundwater level data in a complex hydrogeological district, and the results were significantly improved if compared to the use of classical geostatistical approaches.

This work was developed under the scope of the InTheMED project. InTheMED is part of the PRIMA programme supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 1923.

How to cite: Varouchakis, E., Trichakis, I., and Karatzas, G.: Space-time groundwater level distribution estimation in a complex system of aquifers , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11644, https://doi.org/10.5194/egusphere-egu22-11644, 2022.

14:19–14:26
|
EGU22-12784
|
Virtual presentation
Andres Julian Ruiz Gomez et al.

Initial works on forecasting focused their efforts on one-dimensional precipitation time-series analysis. However, rainfall phenomena are sometimes quite heterogeneous and spatially variable in space and time, especially in extreme events. To address this issue, an integrated approach might be needed, where not only the spatio-temporal variability of rainfall is considered, but also the uncertainty that is present in forecasting. The objective of this research is to analyse the relationship between spatio-temporal rainfall objects estimated from numerical weather prediction models and their hydrological response in a river basin. It is assumed that a better understanding of this relation could help to characterize and forecast extreme phenomena. For this study, the Dapoling-Wangjiba catchment is evaluated, where observed precipitation and discharge data from 2006 to 2009 were available. The analysis is based on four main components:  first, the rainfall perturbed members data are obtained through the TIGGE dataset from ECMWF. Second, the object-based methodology ST-CORA is used in order to characterize the possible rainfall events via its spatio-temporal characteristics such as centroid, spatial coverage and duration. Third, a fully distributed and calibrated HAPI model is used for obtaining a simulated discharge in the catchment outlet and finally, an analysis between the statistics of the object’s characteristics and the hydrological response is carried out. The results of this research are expected to be used in future improvements on how forecasting and early warning and nowadays emitted and understood.

How to cite: Ruiz Gomez, A. J., Corzo Perez, G., Van Andel, S. J., and Santos Granados, G. R.: Analysis of spatiotemporal rainfall objects in hydrological ensemble forecast predictions , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12784, https://doi.org/10.5194/egusphere-egu22-12784, 2022.

14:26–14:33
|
EGU22-12878
|
ECS
Cindy Beltran Mora

Drought events are more common nowadays than it used to be in the past in different areas around the world. Some of its consequences are the reduction of cropping areas, lower rates of percolation for recharge of aquifers and scarcity of drinkable water. To analyse and identify droughts location, dates of occurrence and severity, it is necessary to collect meteorological data. However, based on the location of the study region, some places do not have measurement stations, hence, spatiotemporal databases are the best alternative. Present paper shows the analysis of the past and future scenarios of hydrological droughts due to climate change in the Jiet river basin in Romania. Spatiotemporal data from remote sensing with different resolutions is analysed and processed. Data from year.1990. to year 2020. At a resolution of 0.1̊ is compared to projection scenario RCp 8.5 during 2030 to 2060.

An assessment of hydrological droughts for past scenarios is made by defining a statistical threshold (85 percentile) from historical data. Further, an analysis of characteristics of hydrological drought events is performed for present and future scenarios. Drought is measured through soil moisture analysis, using results from HEC – HMS v4.9 BETA, hydrological modelling tool. This study presents the results of the first stage of the process were a spatiotemporal analysis of the calibration performance of the of the hydrological model and the droughts found are characterized. 

How to cite: Beltran Mora, C.: Climate change analysis of hydrological droughts in Jiet catchment, Romania., EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12878, https://doi.org/10.5194/egusphere-egu22-12878, 2022.

14:33–14:40
|
EGU22-12919
|
ECS
|
Highlight
|
Virtual presentation
Extreme Weather Impacts on Precipitation-Runoff Processes in Heavily Managed Landscapes
(withdrawn)
Noel Aloysius
14:40–14:47
|
EGU22-12857
|
ECS
|
|
Virtual presentation
Karel Aldrin Sanchez Hernandez and Gerald Augusto Corzo Perez

Many regions in the world are threaten by Climate change, and there is a large global concern on the dependency of water contributions from neighbouring countries. In order to understand more how the water contributions from other region affect a river basin, global spatiotemporal information could be used to obtain budgets balance. This study proposes a methodology to analyse the atmospheric moisture balance around hydrological units (watersheds) using ERA 5 reanalysis data sets, allowing the evaluation of the role of spatiotemporal patterns associated with the transport of regional moisture fluxes and understanding how these components modulate regional water heterogeneity, sources and sinks. This study consists of 3 phases: 1) collection and validation of the required hydrometeorological sets and variables and two-dimensional discretization of the hydrographic domain or unit establishing the boundaries for computational analysis; then, estimation and evaluation of the contribution patterns of transported moisture fluxes based on the Eulerian model developed by Brubaker,1993. Finally, for each region, we proceed to estimate the spatiotemporal variations of the atmospheric water balance by establishing the calculation of the precipitation recycling rate as well as the fractions of horizontal moisture flux contributions from each direction or boundary, as well as their seasonality and interannual variability, magnitudes and concentration rates associated with flux divergence. As a case study, the Pamplonita river basin in Colombia was selected. Here we present these results, that have provided valuable information related to the identification of biases in the estimation of atmospheric water supplies, monitoring strategies and hydrological balance.

How to cite: Sanchez Hernandez, K. A. and Corzo Perez, G. A.: Bidimensional Spatiotemporal analysis of local atmospherically fluxes and regional moisture budgets in river basins, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12857, https://doi.org/10.5194/egusphere-egu22-12857, 2022.

Wed, 25 May, 15:10–16:40

Chairpersons: Gerald A Corzo P, Francisco Munoz-Arriola, Adrian Almoradie

15:10–15:20
|
EGU22-4032
|
solicited
|
On-site presentation
Héctor Aguilera et al.

Hydraulic conductivity (Ks) is one of the most challenging, time-consuming, and expensive soil hydraulic properties to estimate. Pedotransfer functions (PTFs) of general use for Ks estimation are often site and sample-scale specific and perform poorly when extrapolated to different regions and extents. The present work develops a stepwise methodology for topsoil Ks mapping at a catchment scale based on easy, fast, and inexpensive measurements and auxiliary data (Aguilera et al., 2022). It includes a double-scale sampling of the Ks to account for small-scale variability in the spatial geostatistical interpolation. A supervised selection of variables through correlation analysis and hierarchical clustering of variables precedes the development of site-specific PTFs with machine learning (ML) techniques. The ML model is then used to generate new Ks point data predictions to extend the spatial coverage for mapping. Finally, the consistency of the final Ks map is assessed in terms of a geomorphological base map.

The variable selection process filtered out four predictor variables from the initial pool of fourteen predictors. An artificial neural network (ANN) provided the best Ks prediction model with one hidden layer and six input variables (latitude, longitude, silt, clay, medium sand, and land use). Latitude and longitude coordinates and land use are surrogates for other physical and environmental (e.g., anthropic) factors. The relative importance of input variables in the ANN was determined as the sum of the product of raw input-hidden, hidden-output connection weights across all hidden neurons using Olden's algorithm. Longitude, percentage of clay, and percentage of medium sand presented a stronger positive relationship with Ks, while irrigated and dry land uses together with the percent of silt were the variables with a more significant negative influence on Ks. The fact that Ks was positively related to clay content is surprising, and it appears to be related to soil plowing before sampling.

The ANN was used to estimate new Ks values from a subsequent sampling of model covariates, which doubled the input information for spatial interpolation using ordinary kriging. Overall, the spatial distribution of Ks was consistent with the lithological variability and other superimposed anthropic factors, as the method adequately considers the spatial variability of Ks added by anthropization to the already high natural heterogeneity. The produced maps will help in the hydraulic planning and flood risk management in the study area where high and low Ks, respectively, are clearly outlined.

Reference:

  • Aguilera, C. Guardiola-Albert, L. Moreno Merino, C. Baquedano, E. Díaz-Losada, P. Agustín Robledo Ardila, J.J. Durán Valsero, Building inexpensive topsoil saturated hydraulic conductivity maps for land planning based on machine learning and geostatistics, CATENA, Volume 208, 2022, 105788, https://doi.org/10.1016/j.catena.2021.105788.

This work is performed within the framework of the RESERVOIR project, part of the PRIMA Programme supported by the European Union. The PRIMA programme is supported under Horizon 2020 the European Union's Framework Programme for Research and Innovation. PRIMA RESERVOIR Grant Agreement number is 1924.

How to cite: Aguilera, H., Guardiola-Albert, C., Moreno Merino, L., Baquedano, C., Díaz-Losada, E., Robledo Ardila, P. A., de la Losa, A., and Durán Valsero, J. J.: Building saturated hydraulic conductivity maps with machine learning and geostatistics, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4032, https://doi.org/10.5194/egusphere-egu22-4032, 2022.

15:20–15:27
|
EGU22-8167
|
ECS
|
|
On-site presentation
Sebastian Müller et al.

The GeoStat Framework is a coherent ecosystem of Python packages for geostatistical applications and subsurface simulations. 
It provides an easily usable open source collection of software packages. They are well documented, including exhaustive hands-on guides and examples for helping non-programmers and non-domain-experts getting started. The main applications of the packages are:

  • GSTools & PyKrige - spatial random field generation, kriging, data normalization & transformation, and geostatistical analyses based on variogram methods
  • AnaFlow - (semi-)analytical solutions for specific groundwater-flow scenarios and the extended Generalized Radial Flow model
  • welltestpy - store, manipulate, and analyze well-based field testing campaigns with a focus on estimating parameters of subsurface heterogeneity from pumping test data.
  • ogs5py - pre-processing, operating, and post-processing of subsurface flow and transport simulations by providing a Python API for the FEM solver OpenGeoSys 5

With this collection of flexible toolboxes we aim to close the gap of missing software for real-world applications in the field of geostatistics. 
Especially GSTools is the first comprehensive Python-toolbox for covariance models, field generation, kriging, variogram estimation, data normalization and transformation. The formerly independed project PyKrige, developed by Benjamin Murphy, has been migrated to the GeoStat-Framework und we started to build a common Rust backend for the numerical heavy lifting called GSTools-Core.

We will show a set of complex workflow examples, like temperature trend analysis and pumping test ensemble simulations, and we will give an overview of the provided functionality and an outlook for the future. All workflows are made accessible as GeoStat-Examples repositories.

References

  • GSTools: Müller, S., et. al. https://doi.org/10.5194/gmd-2021-301, 2021 (in review)
  • AnaFlow: Müller, S., et. al. https://doi.org/10.1016/j.advwatres.2021.104027, 2021
  • welltestpy: Müller, S., et. al. https://doi.org/10.1111/gwat.13121, 2021
  • ogs5py: Müller, S., et. al. https://doi.org/10.1111/gwat.13017, 2020

Links

  • Website: https://geostat-framework.org/
  • GitHub: https://github.com/GeoStat-Framework
  • GeoStat-Examples: https://github.com/GeoStat-Examples

How to cite: Müller, S., Schüler, L., Zech, A., and Heße, F.: The GeoStat-Framework: Create your geostatistical model in Python!, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8167, https://doi.org/10.5194/egusphere-egu22-8167, 2022.

15:27–15:34
|
EGU22-6288
|
On-site presentation
Dionissios Hristopulos et al.

Gaussian processes are a flexible machine learning framework that can be used for spatial interpolation and space-time prediction as well. Gaussian process regression (GPR) is quite similar to the geostatistical kriging method.  It encompasses various types of kriging (e.g., simple, ordinary, universal and regression kriging).  In addition, it is formulated in an inherently Bayesian framework which allows taking into account a priori beliefs regarding the distribution of the model’s hyper-parameters. Thus, it also incorporates Bayesian versions of kriging [1].  GPR is based on the assumption that the stochastic component of the observations follows a Gaussian distribution.  However, this is not the case for various environmental variables (e.g., amount of precipitation, hydraulic conductivity, wind speed), which follow skewed probability distributions.  The skewness is handled within the geostatistical framework using nonlinear transforms such that the marginal distribution of the data in the latent space becomes normal.  This procedure is known as Gaussian anamorphosis in geostatistics.  In the context of GPR, the term warped Gaussian process is used to denote the nonlinear transformation of the observations [2].   Gaussian anamorphosis (warping) is usually implemented using explicit, monotonically increasing nonlinear functions.  A different approach involves generating the warping function with the help of the empirically estimated cumulative probability distribution of the data.  This approach provides flexibility because the transformation is data-driven (non-parametric) and is thus not constrained by specific functional forms.  Furthermore, the cumulative distribution function of the data can be accurately estimated using smoothing kernels [3].  We investigate warped Gaussian process regression using synthetic datasets and precipitation reanalysis data from the Mediterranean island of Crete. Cross validation analysis is used to establish the advantages of non-parametric warping for the interpolation of incomplete data. We demonstrate that warped GPR equipped with data-driven warping provides enhanced flexibility compared to "bare" GPR and can lead to improved predictive accuracy for non-Gaussian data.  

Keywords: Gaussian processes, Mediterranean island, non-Gaussian, warping, precipitation

Funding: This research is co-financed by Greece and the European Union (European Social Fund- ESF) through the Operational Programme «Human Resources Development, Education and Lifelong Learning 2014-2020»in the context of the project “Gaussian Anamorphosis with Kernel Estimators for Spatially Distributed Data and Time Series and Applications in the Analysis of Precipitation” (MIS 5052133).

References

[1] T. Hristopulos, 2020. Random Fields for Spatial Data Modeling. Springer Netherlands, http://dx.doi.org/10.1007/978-94-024-1918-4.

[2] Snelson, E., Rasmussen, C.E. and Ghahramani, Z., 2004. Warped Gaussian processes. Advances in neural information processing systems, 16, pp.337-344.

[3] Pavlides, A., Agou, V., and Hristopulos, D. T., 2021. Non-parametric Kernel-Based Estimation of Probability Distributions for Precipitation Modeling. arXiv preprint arXiv:2109.09961.

How to cite: Hristopulos, D., Agou, V., and Pavlides, A.: Data-driven Warping of Gaussian Processes for Spatial Interpolation of Skewed Data, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6288, https://doi.org/10.5194/egusphere-egu22-6288, 2022.

15:34–15:41
|
EGU22-11777
|
ECS
|
Highlight
|
|
Virtual presentation
Kabita Gautam et al.

Reducing the threat of severe spatiotemporal events like floods and droughts is raising concerns for water resource development and management. The severity of drought and floods increases more due to human interventions. Recent studies have focused on finding long-term solutions that mimic nature's process, while posing no environmental risks and targeting sustainability in traditional approaches. The terminology given in Europe for this natural solution is Nature-Based Solutions (NBS). Some examples of NBS are afforestation or reforestation, storage areas, vegetation buffers, and riparian forest. The main principle of NBS is that they slow down the rate of runoff by boosting interception, infiltration, or storage for flood water, hence mitigating the risk downstream. However, there is still not enough experimental nor theoretical experience on how they could be implemented to optimize their use. The way to represent NBS and the scale of implementation in models and real life is for now a process based on approximated propositions of the empirical knowledge of experts in the field. Although some experience have shown important contributions, this is not enough for an optimal implementation and a complete understanding of all possible outcomes. This is the problem expected to be addressed in this research. The main goal is to construct machine learning models to explore their use as an alternative (surrogate) that will aid in performing multiple scenario analyses of NBS, and quantifying their impacts. This approach will consider spatial and temporal data and create a link between several environmental variables and human actions without explicitly knowing the physical behavior of the system, yet clustering(grouping) behaviors or processes responses to structural properties of the hydrological model representation. The case study area for this research is the Bagmati River Basin of Nepal, covering a catchment area of 2822 sq. km, and the flow is dominated by spring and monsoon rainfall. Soil and Water Assessment Tool (SWAT) is used and the baseline scenario (without the implementation of NBS) is modeled. Different scenarios of afforestation, ponds, and conservation tillage will be intervened in the SWAT model and the changes made by those interventions will be replicated in Artificial Neural Network-Multi Layer Perceptron (ANN-MLP). Several unforeseen scenarios will also be tested in machine learning. Thus, the Spatio-temporal analysis will be done regarding the impact of NBS on the flows, and the machine learning model’s ability to replicate such complex systems will be evaluated.

 

The outcome presented here shows the construction of a SWAT model and the preliminary results of machine learning models capable of promptly predicting changes in flow induced by the adoption of various Nature-Based Solutions. It is anticipated to be a simple, effective, and time-saving way for studying the effectiveness of various Nature-Based Solutions for flood and drought mitigation. Thus, this study contributes to the experiences in interpreting and linking complex hydrological problems in machine learning systems.

 

Keywords: SWAT, Machine learning, Nature-Based Solutions, Hydrological extremes, Spatio-temporal analysis

How to cite: Gautam, K., Corzo, G., Maskey, S., and Solomatine, D.: Machine Learning Model to Reproduce Nature-Based Solutions for Flood and Drought Mitigation, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-11777, https://doi.org/10.5194/egusphere-egu22-11777, 2022.

15:41–15:48
|
EGU22-12817
|
ECS
|
|
Virtual presentation
Mohamed Elbasheer and Gerald Corzo Perez

Sub-seasonal to seasonal (S2S) forecasting ranges from two weeks to two months. This range of time is significant and has a substantial socio-economic impact for many societal applications such as agriculture, food security and risk mitigation because it gives a reasonable margin of time for any management measure (for example, disaster or risk mitigation measures) that need one or two weeks to be implemented. However, the reliability of the S2S forecasting is still underdeveloped, and many studies and even competitions have been promoted to aim at the study of how can Machine learning and other techniques help. So, in this study we evaluate the accuracy and reliability of the ECMWF S2S precipitation forecasts focusing on the extreme events (above and below normal precipitation events) using three verification methods; Receiver operating characteristic curve (ROC), Reliability diagram and Ranked probability skill score (RPSS). For this evaluation, three regions are selected globally. In addition to the accuracy evaluation, we investigated the use of machine learning techniques such as k-nearest neighbors (k-NN), Logistic Regression (LR) and Multilayer Perceptron (MLP) to improve the accuracy and reliability of the ECMWF S2S forecast. To select the appropriate input variables for the machine learning models; An analysis of temporal and spatial continuity of the variables is done using the Pearson correlation coefficient for temporal correlation and the experimental variogram for spatial continuity.

How to cite: Elbasheer, M. and Perez, G. C.: Exploratory analysis of Sub-seasonal to Seasonal precipitation forecasting using Machine Learning Techniques, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12817, https://doi.org/10.5194/egusphere-egu22-12817, 2022.

15:48–15:55
|
EGU22-12896
|
ECS
|
Highlight
|
Virtual presentation
Laura Viviana Garzon Useche et al.

The erratic drought nature and its great spatiotemporal variability make conventional forecasting systems or stochastic forecasting systems very limited in terms of the monitoring of its dynamic characteristics. Therefore, this study proposes a dynamic forecasting methodology based on machine learning models by tracking the spatial and temporal characteristics of drought events.

This methodology consists of four main phases. 1) the drought spatiotemporal characteristics calculation and extraction such as spatial aggregations or extensions, geospatial properties (area, perimeter), centroid location and trajectory from their connectivity, which are generated following the contiguous drought area analysis (CDA) proposed by Corzo--- 2) feature engineering and dataset preparation, which is consolidated according to the hierarchy and relative importance of the associated predictor and predictor variables 3) implementation of an intelligent analysis method based on deep neural network architecture (CNN, LSTM) techniques, which combines spatial observation mediated by convolution integrated with temporal analysis for prediction. Thus generating primary results against the future propagation pattern or trajectory of a spatial unit. 4) Analyzing the various model performances based on statistical metrics, validation of the generated trajectories using the area under the curve (AUC) and receiver operating characteristic (ROC) and error approach as Root Mean Square Error (RMSE).

This methodology is presented using indexes derived from the ERA 5 reanalysis dataset as SPEI and SPI on the Central America dry corridor (1979-2020), where the performance of the intelligent system will be evaluated not only taking into account the statistical performance, but also in the identification and forecasting of those regions with major drought generation tendencies.

How to cite: Garzon Useche, L. V., Sánchez Hernández, K. A., Corzo Perez, G. A., and Santos Granados, G. R.: Machine Learning Techniques for Spatiotemporal drought patterns forecasting, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12896, https://doi.org/10.5194/egusphere-egu22-12896, 2022.