Extreme runoff modeling is hindered by the lack of sufficient and relevant ground information and the low reliability of physically-based models. The authors propose to combine precipitation Remote Sensing (RS) products, Machine Learning (ML) modeling, and hydrometeorological knowledge to improve extreme runoff modeling. The approach applied to improve the representation of precipitation is the object-based Connected Component Analysis (CCA), a method that enables classifying and associating precipitation with extreme runoff events. Random Forest (RF) is employed as a ML model. We used 2.5 years of nearly-real-time hourly RS precipitation from the PERSIANN-CCS and IMERG-early run databases (spatial resolutions of 0.04 o and 0.1 o , respectively), and runoff at the outlet of a 3391 km 2-basin located in the tropical Andes of Ecuador. The developed models show the ability to simulate extreme runoff for the cases of long-duration precipitation events regardless of the spatial extent, obtaining Nash-Sutcliffe efficiencies (NSE) above 0.72. On the contrary, we found an unacceptable model performance for a combination of short-duration and spatially-extensive precipitation events. The strengths/weaknesses of the developed ML models are attributed to the ability/difficulty to represents complex precipitation-runoff responses.

Jeff Apple Benowitz1, Michael Everett Mann21 GeoSep Services, 1521 Pine Cone Road, Moscow, ID, USA2 Department of Earth, Environmental and Planetary Sciences, Brown University, Providence, RI, USACorresponding author: Jeff Apple Benowitz ([email protected])AbstractLong-lived magmatic arcs theoretically should migrate large trench perpendicular distances as convergent margin configurations and slab geometry vary over time, however many arc-magmatic belts are spatially localized over 10’s of millions of years. We document, by compiling published crystallization geochronology data for southern Alaska (6485 total bedrock and single grain detrital ages combined), that since ca. 100 Ma, arc magmatism has been localized along the Alaska Range suture zone, at times over 500-km inboard. However, since ca. 100 Ma incoming subducting slab characteristics, beneath mobile southern Alaska and convergent margin configurations, varied greatly and include both normal oceanic plate and oceanic plateau subduction, plate vector changes, oroclinal bending and reconfiguration of trench shape, terrane accretion, long distance translation and a Paleocene slab break off/slab window event. Therefore, it is inferred that inherited upper-plate lithospheric shape and heterogeneity must control in part the geometry of the subducting slab below a mobile southern Alaskan margin through hydrodynamic (viscous) mantle wedge “suction” forces. Additionally, crustal thickness heterogeneity may preferentially focus magma ascent through melt ponding along Moho offsets, and upper-plate lithospheric-scale strike-slip faults may be acting as passive and active conduits for arc magmatism. Inherited upper-plate controls on slab geometry could be a factor localizing arc magmatism along other long-lived convergent margin settings.

15 The increasing pressure on wetland resources continues to threaten the role wetlands play in 16 maintaining the ecological balance of watersheds. The Cuvette Centrale of the Congo is the 17 greatest intertropical peatland in the world. To fully understand its role in water resources and 18 ecological services linked to the quality of water and life in the basin, we first need to quantify 19 its role in the hydrological dynamics. To achieve this aim, we used the Soil and Water 20 Assessment Tool model (SWAT)-modified for tropical environments-in combination with 21 monthly discharge data. We analyzed water fluxes entering and flowing out of the Cuvette 22 Centrale of the Congo River Basin on a monthly time scale for the 2000-2012 period. The 23 model was calibrated, validated, and compared with discharge from gauging stations and 24 surface water elevation from radar altimetry. Results showed that upland runoff from the 25 Congo River was the highest contributor to the Cuvette Centrale (33 percent) followed closely 26 by efficient precipitation inside the Cuvette Centrale (31 percent) with right bank and left bank 27 tributaries contributing 25 percent and 11 percent respectively. We simulated monthly mean 28 interannual inflows of approximately 34,150 m 3 s-1 (88 billion m 3) with the main flood peaking 29 in November (45,310 m 3 s-1) and total outflows averaging around 39,860 m 3 s-1 (100 billion 30 m 3) peaking at 52,430 m 3 s-1 in December for the simulation period. We subsequently estimated 31 a negative monthly mean interannual variation of storage in the Cuvette Centrale wetlands in 32 the order of 5,700 m 3 s-1 suggesting that the Cuvette Centrale supplies the river during low 33 water periods. This highlights the important regulatory function of the Cuvette Centrale and 34 the need for protection of groundwater resources in order to maintain wetland water quantities 35 and quality. 36

Cumulate rocks record the magmatic and cooling processes during formation of Earth’s igneous crust. Extracting the information of these two processes from mineral records, however, is often convoluted by various extents of diffusive resetting during cooling subsequent to main stages of crystallization. Accordingly, for cumulate rocks at diffusive closure, the apparent “equilibrium” temperatures derived from geothermometers are generally lower than the crystallization temperatures. Using analytical or numerical models, geospeedometers can extract cooling rates from the closure temperatures (or profiles) but only if the initial temperatures are determined independently. Here I summarize the general framework of geothermometry and geospeedometry from a trace element perspective. The Mg and REE-based exchange geothermometers for mafic cumulate rocks are reviewed as examples of the geothermometer design. Based on the observed differential diffusive closures of Mg and REE in oceanic gabbros, I outline a general resolution to uniquely determine the initial crystallization temperature and cooling rate of a cumulate rock. The basic idea is further demonstrated using the recently developed Mg-REE coupled geospeedometer for mafic cumulate rocks. Finally, I use the Hess Deep gabbros as a case study to show that this two-element coupled geospeedometer is particularly useful to delineate the igneous accretion and cooling styles during crust formation. This two-element (or multi-element) coupled approach outlined here can also be readily extended for decoding comprehensive thermal histories of other petrological systems at various geological settings or other rocky planetary bodies.

The tectonic development of the Southeast Anatolian Orogenic Belt (SAOB) is closely related to the demise of the NeoTethys Ocean, which was located between the Arabian and Eurasian plates from the late Cretaceous to Late Miocene. The ocean contained several continental slivers and intra-oceanic magmatic arcs. The continental slivers represent narrow tectonic belts rifted off and drifted away from the Arabian Plate while the NeoTethyan Ocean and the back-arc basins were opened. Later they collided with one another during the branches of the oceans were eliminated. In these periods, the continental slivers were involved in the subduction zone and turned into metamorphic massifs. During the Late Cretaceous, the first collision occurred when an accretionary complex was thrust over the Arabian Plate’s leading edge. Despite the collision, the ocean survived in the North and Its northward subduction generated a new intra oceanic arc, which collided later with the northerly located continental slivers. In this period, the metamorphic massifs and the intra-oceanic arc front migrated to the South. The new magmatic arc collided with the southerly transported nappe package during the Late Eocene. The amalgamated nappe pile eventually obducted onto the Arabian Plate during the Late Miocene. The collision produced escape structures during the Neotectonic period.

Finding optimum balances between conflicting interests in multipurpose reservoirs often represents an important challenge for decision makers. This study assesses the use of different computational tools to obtain optimal reservoir operations applied to the Hatillo dam in the Dominican Republic. A multiobjective optimization approach is used, in which non-dominated sorting genetic algorithm II (NSGAII) and multi-objective evolutionary algorithm based on decomposition (MOEA/D) optimizers are applied to models that simulate reservoir operations. Three different Machine Learning (ML) models, namely, the multilayer perceptron (MLP), the radial basis network (RBN) and the linear function (LF), are employed to learn the current operation of the system. Subsequently, a general model is proposed to simulate daily reservoir operations (2009-2019), integrating water balances, physical constraints of the dam components and the ML models, the latter defining daily controlled discharges. In the optimization process, the ML parameters are the decision variables, while the objectives evaluated are irrigation, hydropower generation and flood control. The results are compared with the actual operation of the reservoir. Three dimensional Pareto fronts are obtained, from which, the wide variety of operations can be evidenced. The flood control objective was found to have a wide room for improvement over the current operation of the reservoir, and several of the solutions found improve the current operation for the three proposed objectives. The MLP models tend to generate the best results for this case study and the NSGAII optimizer generates the best optimization results.

This study intends to integrate heterogeneous remote sensing observations and hydrological modelling into a simple framework to monitor hydrological variables in the poorly gauged Congo River basin (CRB). It focuses on the possibility to retrieve effective channel depths and discharges all over the basin in near real time (NRT). First, this paper discusses the complexity of calibrating and validating a hydrologic–hydrodynamic model (namely the MGB model) in the CRB. Next, it provides a twofold methodology for inferring discharge at newly monitored virtual stations (VSs, crossings of a satellite ground track with a water body). It makes use of remotely sensed datasets together with in-situ data to constrain, calibrate and validate the model, and also to build a dataset of stage/discharge rating curves (RCs) at 709 VSs distributed all over the basin. The model was well calibrated at the four gages with recent data (Nash-Sutcliffe Efficiency, NSE> 0.77). The satisfactory quality of RCs basin-wide (mean NSE between simulated discharge and rated discharge at VSs, NSEmean = 0.67) is an indicator of the overall consistency of discharge simulations even in ungauged upstream sub-basins. This RC dataset provides an unprecedented possibility of NRT monitoring of CRB hydrological state from the current operational satellite altimetry constellation. The discharges estimated at newly monitored locations proved to be consistent with observations. They can be used to increase the temporal sampling of water surface elevation (WSE) monitoring from space with no need for new model runs. The RC located under the fast sampling orbit of the SWOT satellite, to be flown in 2022, will be used to infer daily discharge in major contributors and in the Cuvette Centrale, as soon as data is released.

A multi-model hydrological assessment in the Congo Basin is performed to assess water availability conditions for historical and future periods (1913–2099). With models limited by scarce in situ observations, a combination of GRACE satellite data and soil-moisture-based drought indices is shown to be capable of estimating water budget, streamflow, and drought and storage variability. Changes in land use and land cover played a role in modifying the hydrologic responses but were found to be within the uncertainties of other inputs, including weather, soil, and model parameters. Seasonal and annual variability in total water storage anomalies (TWSAs) and the modified Palmer drought severity index (MPDSI) display a good correlation with each other. A selected set of global climate models is used to characterize the future temperature and precipitation patterns. It is expected that subbasin-scale variability in future temperature and precipitation increases will result in increased evapotranspiration, decreased runoff, and more drought events in the Congo Basin.

Fragments of former oceans are commonly observed in mountain belts: blueschists and eclogites, on the one hand, and ophiolites, on the other hand, are all that remains of ancient oceanic lithosphere. Though volumetrically subordinate, they provide essential insights into past geodynamics and into the processes involved in the formation and destruction of oceanic lithosphere. This contribution apprehends these two types of oceanic fragments jointly and shows the advantage of doing so for understanding the dynamics of oceanic convergence, i.e. subduction and obduction. We examine the intimate relationships between blueschists/eclogites and ophiolites, as well as the similarities and differences in the mechanisms leading to their preservation. While the extensive, unmetamorphosed true ophiolites markedly differ from fragments of oceanic lithosphere offscraped from the slab during subduction, at shallow or great depths, both types record the mechanical behavior and ‘hiccups’ of the subduction plate boundary. Their preservation also highlights the importance of the evolution of the subduction regime through time, from the onset of intra-oceanic subduction to the cessation of continental subduction.

Skillful forecasts of weather phenomena in numerical models begin with the most accurate set of initial conditions achievable from observational datasets. The process of combining observations with numerical model predictions is called data assimilation. This chapter describes the types of observations available for data assimilation in models that predict the transport, fate, and impacts of smoke pollution. Observation properties needed for effective data assimilation are identified based on experiences with a variety of observation types in data assimilation experiments, compiled from the published literature. The second half of the chapter surveys the data assimilation methodologies that have been applied to smoke aerosols, and describes specific problems associated with the smoke observations that require innovative techniques in data assimilation. The chapter concludes by providing an outlook for future research and development in data assimilation for smoke prediction models. Data assimilation for prediction of smoke is an emerging area of development that promises to greatly improve forecast skill as new datasets and techniques are applied.

The National Oceanic and Atmospheric Administration (NOAA) research to operations (R2O) experiment called the Big Data Project (BDP) was envisioned as a scalable approach for disseminating exponentially increasing NOAA observation, model, and research datasets to the public using commercial cloud services. At the start of the project, during the concept development phase, it was unclear how the specifics might work so a spiral development approach was adopted. It was expected that the number of data sets would increase, and the data extent would grow to cover complete records of some holdings, and that format experimentation would be needed to determine optimal cloud offerings. This dissemination model would require a new way for the BDP and NOAA to engage with end-users, who could range from large enterprises to small businesses and individuals. The BDP was expected to change the game-not just by reaching a broad and diverse set of users but by encouraging new ones. As Dr. Kathy Sullivan, former NOAA Administrator under whom the BDP began, noted, “The agency’s aim is to ‘spur innovation’ and to explore how to create a ’global economic return on investment” (Konkel, 2015). This Chapter describes the journey of BDP as it developed, transitioned and evolved from an experiment to an operational enterprise function for NOAA, now known as NOAA Open Data Dissemination (NODD). Obstacles to the Public’s Use of NOAA Environmental Data NOAA’s mission is to understand and predict changes in climate, weather, oceans, and coasts, to share that knowledge and information with others, and to conserve and manage coastal and marine ecosystems and resources. The agency takes seriously the need for communication of NOAA’s research, data, and information for use by the Nation’s businesses and communities to allow preparation, response and resilience to sudden or prolonged changes in our natural systems. This includes climate predictions and projections; weather and water reports, forecasts and warnings; nautical charts and navigational information; and the continuous delivery of a range of Earth observations and scientific data sets for use by public, private, and academic sectors (NOAA About our agency, 2021).

Elements with variable valence state (i.e. redox-sensitive) often show contrasting mineral/melt partition coefficients as a function of oxygen fugacity (fO2) in magmatic systems. This is because trace-element incorporation into crystal lattices depends on the charge, size, and crystal-field stabilization energy of atoms, all of which differ greatly between oxidized and reduced species of the same element. This has two critical implications: (1) petrologic/ geochemical modelling of partitioning behavior of redox-sensitive trace-elements in magmatic systems requires some knowledge of their oxidation state, and (2) the oxidation state of magmatic systems may be inferred from partitioning relations of redox-sensitive trace elements preserved in mineral and melt phases of rapidly cooled magmas. The advantage of this oxybarometric approach is that mineral/melt partitioning relations are not sensitive to late stage degassing, charge-transfer on quenching, or surficial alteration. In this chapter we discuss the theoretical treatment of experimental mineral/melt partitioning data of redox-sensitive trace elements, and review aspects concerning the partitioning behavior of well-known redox-sensitive elements, including transition metals (Ti, V, Cr, Fe), rare earth elements (Ce, Eu), U, and siderophile elements (Mo, W, Re, and platinum group elements) under planetary magmatic fO2 conditions.

The Congo basin is one of the most hydrologically active and pristine locations with limited understanding of how precipitation changes impacts on stream flow dynamics and variations in catchment stores. Given that the basin is among the three prominent convective regions that dominates global rainfall climatology during transition seasons, historical space-time variability of rainfall (1901-2014) over the basin in relation to river discharge is analyzed in order to understand significant hydro-climatic shift. Based on advance multivariate analyses, the total variability of the leading modes (annual variations) of rainfall increased during the 1931-1960 (56.3%) and 1961-1990 (57.3%) periods compared to the 1901-1930 baseline period (51.3%). It varied less between 1991 and 2014 (55.4%) as opposed to the two climatological periods between 1931 and 1990. Furthermore, the total variability in the multi-annual rainfall signals declined from 16.5% at the start of the century (1901-1930) to 13.6% in the 1991-2014 period while the total variability accounted for by other short-term meteorological signals oscillated between 4.0% and 2.7% during the entire period. Between 1995 and 2010 there seems to be a change in the hydrological regimes of the Congo river as the cumulative departures of rainfall and discharge were in opposite directions. The considerable association of discharge with rainfall in catchments characterized by strong annual and seasonal amplitudes in rainfall implies that the wetland hydrology of the basin is largely nourished by rainfall, in addition to possible exchange of fluxes within the Congo floodplain wetlands. Notably, a significant proportion of changes in the dominant rainfall patterns is still not explained by those of river discharge. This information signals the threshold of complex hydrological processes in the region, and perhaps suggest the influence of anthropogenic contributions (e.g., deforestation) and strong multi-scale ocean-atmosphere phenomena as key secondary drivers of hydrologic variability.

The Congo Basin exhibits tremendous heterogeneities, out of which it emerges as an intricate system where complexity will vary consistently over time and space. Increased complexity in the absence of adequate knowledge will always result in increased uncertainties. One way of simplifying this complexity is through an understanding of organisational relationships of the landscape features, which is termed here as catchment classification. The need for a catchment classification framework for the Congo Basin is obvious given the basin’s inherent heterogeneities, the ungauged nature of the basin, and the pressing needs for water resources management that include the quantification of current and future supplies and demands, which also encompass the impacts of future changes associated with climate and land use, as well as water resources operational policies. The need is also prompted by many local-scale management concerns within the basin. This study uses an a priori approach to determine homogenous climatic-physiographic regions that are expected to underline dominant hydrological processes characteristics. A set of 1740 catchment units are partitioned across the whole basin, based on a set of comprehensive criteria, including natural break of the elevation gradient (199 units), inclusion of socio-economic and anthropogenic systems (204 units), and water management units based on traditional nomenclature of the rivers within the basin (1337 units). The identified catchment units are used to assess existing datasets of the basin physical properties, necessary to derive descriptors of the catchments characteristics. An unsupervised classification, based on Hierarchical Agglomerative Cluster algorithm is used, that yields 11 homogenous groups that are consistent with the current perceptual understanding of the Congo Basin physiographic and climatic settings. These regions represent therefore an a priori classification that will be further used to derive functional relationships of the catchments, necessary to enable hydrological prediction and water management in the basin.

This work presents an integrated methodology for the assessment of threats on spring quality and quantity in poorly investigated Mediterranean semi-arid karst catchments in Lebanon. Pilot investigations, including 1) high-resolution monitoring of spring water and climate, 2) artificial tracer experiments, and 3) analysis of micropollutants in surface water, groundwater, and wastewater samples were conducted to assess flow and transport in three karst catchments of El Qachqouch, El Assal, and Laban springs. First, the high-resolution in-situ spring data allows the quantification of available water volumes, as well as their seasonal and yearly variability in addition to shortages and floodwaters. Moreover, the statistical analysis of hydrographs and chemographs helps assess the karst typology, spring type and hydrodynamic behavior (storage versus fast flow). Furthermore, a series of artificial tracer experiments provides information about key-transport parameters related to the intrinsic vulnerability of the pilot springs, while the analysis of micropollutants gives insight into the specific types of point source pollution as well as contaminant types and loads. On the one hand, the tracer experiments reveal that any potential contamination occurring in snow-governed areas can be observed at the spring for an extensive time due to its intermittent release by gradual snowmelt, even with enough dilution effect. On the other hand, the assessment of persistent wastewater indicators shows that springs in the lower catchment (including El Qachqouch) are highly vulnerable to a wide range of pollutants from point source (dolines and river) and diffuse percolation. Such contaminants breakthrough is challenging to predict because of the heterogenous duality of infiltration and flow, typical of karst systems. Finally, this set of investigations is essential for the proper characterization of poorly studied systems in developing areas, whereby results can be integrated into conceptual and numerical models to be used by decision-makers as support tools in science-evidenced management plans.

Flood disasters have regularly been reported in the Congo Basin with significant damages to human lives, food production systems and infrastructure. Losses incurred by these damages are huge and represent a major challenge for economic expansion in developing nations. In the Congo River Basin, where availability of in-situ data is a significant challenge, new approaches are needed to investigate flood risks and enable effective management strategies. This study uses recently developed global flood prediction data in order to produce flood risk maps for the Congo River Basin, where flood information currently does not exist. Flood hazard maps that estimate fluvial flooding at a grid cell resolution of 3 arc-seconds (~ 90 m), gridded population density data of 1 arc-second (~ 30 m) spatial resolution, and a spatial layer of infrastructure dataset are used to addresses flood risk at the scale of the Congo Basin. The global flood data provides different return periods of exposure to flooding in the Congo Basin and identifies flood extents. The risk analysis results are presented in terms of the percentage of population and infrastructure at flood risk for six return periods (5, 10, 20, 50, 75 and 100 year). Of the 525 administrative territories, 374 are AGU Books

Understanding the impacts of climate on surface water hydrology is required to predict consequences and implications on freshwater habitats, ecological assets, and wetland functions. Although the Congo basin is considerably a freshwater-rich region, largely characterised by numerous water resources after the similitude of the Amazon basin, recent accounts of droughts in the basin are indications that even the most humid regions of the world can be affected by droughts and its impacts. Given the scarcity and limited availability of hydrological data in the region, GRACE (Gravity Recovery and Climate Experiment) observations are combined with model and SPEI (standardized precipitation evapotranspiration index) data to investigate the likelihood of such impacts on the Congo basin’s surface water hydrology. By integrating multivariate analysis with support vector machine regression (SVMR), this study provides some highlights on the characteristics (intensity and variability) of drought events and GRACE-derived terrestrial water storage (TWS) and the influence of global climate on the Congo river discharge. The southern section of the basin shows considerable variability in the spatial and temporal patterns of SPEI and extreme droughts over the Congo basin appear to have persisted with more than 40% coverage in 1994. However, there has been a considerable fall in drought intensities since 2007 and coincides with periods of strong positive anomalies in discharge (i.e., 2007-010). GRACE-derived TWS over the Congo basin is driven by annual fluctuations in rainfall (r = 0.81 at three months phase lag) and strong inter-annual variations of river discharge (r = 0.88, α= 0.05). Generally, results show that changes in the surface water variations (from gauge and model output) of the Congo basin is a key component of the GRACE water column. The outputs of the SVMR scheme indicate that global climate through sea surface temperature anomalies of the Atlantic (r = 0.79, α= 0.05), Pacific (r = 0.79, α= 0.05), and Indian (r = 0.74, α= 0.05) oceans are associated with fluctuations in the Congo river discharge, and confirm the importance of climatic influence on surface water hydrology in the Congo basin.

This chapter describes the variability of rainfall and river discharges in the Ogooué River basin (ORB) in recent decades (since 1940). Due to its location crossing the Equator, the ORB receives abundant precipitation that maintains one of the world’s best-preserved ecosystems. In contrast to neighboring forest basins that have been severely degraded because of deforestation, mining resources extraction, extensions of agricultural areas, and river transport, which is a crucial alternative to the cruel lack of road infrastructures, the ORB is experimenting with an exceptional conservation policy in the region. For example, the rural penetration rate in Gabon is about 1 inhabitant per km² and many studies report a deforestation rate close to 0%, with even full natural regeneration. However, the fluctuations of the standardized anomaly index of rainfall in the ORB show three main phases of variations: the first wet phase was characterized by abundant precipitations from 1940 to 1970, the second phase of the long-term mild drought was extended in the 1970s and 1980s and the final third phase presented a slight return of abundance in precipitation. Even though drought severity in the ORB was mainly weak, its effects in river discharges were very sensitive on seasonal and inter-annual scales. The pure equatorial regime of the ORB characterized by equal maximum floods in spring and autumn changed significantly from the difference between both maximum discharges of 13.5 % during the 1960s to 27.0 %, 38.4 %, 33.9 %, and 26.7 % for the 1970s, 1980s, 2000s and 2010s respectively. A brief comparison between the ORB and the Congo River basin showed that changes in the ORB are part of a regional process that Central Africa is undergoing with some spatial heterogeneities.