Showing papers in "Hydrology and Earth System Sciences Discussions in 2021"

Evaluation of Random Forest for short-term daily streamflow forecast in rainfall and snowmelt driven watersheds

Abstract: In the past decades, data-driven Machine Learning (ML) models have emerged as promising tools for short-term streamflow forecasts Among other qualities, the popularity of ML for such applications is due to the methods' competitive performance compared with alternative approaches, ease of application, and relative lack of strict distributional assumptions Despite the encouraging results, most applications of ML for streamflow forecast have been limited to watersheds where rainfall is the major source of runoff In this study, we evaluate the potential of Random Forest (RF), a popular ML method, to make streamflow forecast at 1-day lead time at 86 watersheds in the Pacific Northwest These watersheds span climatic conditions and physiographic settings and exhibit varied contributions of rainfall and snowmelt to their streamflow Watersheds are classified into three hydrologic regimes: rainfall-dominated, transisent, and snowmelt-dominated based on the timing of center of annual flow volume RF performance is benchmarked against Naive and multiple linear regression (MLR) models, and evaluated using four metrics Coefficient of determination, Root mean squared error, Mean absolute error, and Kling-Gupta efficiency Model evaluation metrics suggest RF performs better in snowmelt-driven watersheds Largest improvement in forecasts, compared to benchmark models, are found among rainfall-driven watersheds We obtain Kling–Gupta Efficiency (KGE) scores in the range of 062–099 RF performance deteriorates with increase in catchment slope and increase in soil sandiness We note disagreement between two popular measures of RF variable importance and recommend jointly considering these measures with the physical processes under study These and other results presented provide new insights for effective application of RF-based streamflow forecasting

Towards Hybrid Modeling of the Global Hydrological Cycle

Abstract: . Progress in machine learning in conjunction with the increasing availability of relevant Earth observation data streams may help to overcome uncertainties of global hydrological models due to the complexity of the processes, diversity, and heterogeneity of the land surface and subsurface, as well as scale-dependency of processes and parameters. In this study, we exemplify a hybrid approach to global hydrological modeling that exploits the data-adaptiveness of machine learning for representing uncertain processes within a model structure based on physical principles like mass conservation. Our H2M model simulates the dynamics of snow, soil moisture, and groundwater pools globally at 1o spatial resolution and daily time step where simulated water fluxes depend on an embedded recurrent neural network. We trained the model simultaneously against observational products of terrestrial water storage variations (TWS), runoff, evapotranspiration, and snow water equivalent with a multi-task learning approach. We find that H2M is capable of reproducing the key patterns of global water cycle components with model performances being at least on par with four state-of-the-art global hydrological models. The neural network learned hydrological responses of evapotranspiration and runoff generation to antecedent soil moisture state that are qualitatively consistent with our understanding and theory. Simulated contributions of groundwater, soil moisture, and snowpack variability to TWS variations are plausible and within the large range of traditional GHMs. H2M indicates a somewhat stronger role of soil moisture for TWS variations in transitional and tropical regions compared to GHMs. Overall, we present a proof of concept for global hybrid hydrological modeling in providing a new, complementary, and data-driven perspective on global water cycle variations. With further increasing Earth observations, hybrid modeling has a large potential to advance our capability to monitor and understand the Earth system by facilitating a data-adaptive yet physically consistent, joint interpretation of heterogeneous data streams.

Assessing the dependence structure between oceanographic, fluvial, and pluvial flooding drivers along the United States coastline

Abstract: . Flooding is of particular concern in low-lying coastal zones that are prone to flooding impacts from multiple drivers: oceanographic (storm surge and wave), fluvial (excessive river discharge), and/or pluvial (surface runoff). In this study, we analyse for the first time the compound flooding potential along the contiguous United States (CONUS) coastline from all flooding drivers, using observations and reanalysis datasets. We assess the overall dependence from observations by using Kendall’s rank correlation coefficient (τ) and tail (extremal) dependence (χ). Geographically, we find highest dependence between different drivers at locations in the Gulf of Mexico, southeast, and southwest coasts. Regarding different driver combinations, the highest dependence exists between surge-waves, followed by surge-precipitation, surge-discharge, waves-precipitation, and waves-discharge. We also perform a seasonal dependence analysis (tropical vs extra-tropical season), where we find higher dependence between drivers during the tropical season along the Gulf and parts of the East coast and stronger dependence during the extra-tropical season on the West coast. Finally, we compare the dependence structure of different combinations of flooding drivers using observations and reanalysis data and use the Kullback–Leibler (KL) Divergence to assess significance in the differences of the tail dependence structure. We find, for example, that models underestimate the tail dependence between surge-discharge on the East and West coasts and overestimate tail dependence between surge-precipitation on the East coast, while they underestimate it on the West coast. The comprehensive analysis presented here provides new insights on where compound flooding potential is relatively higher, which variable combinations are most likely to lead to compounding effects, during which time of the year (tropical versus extra-tropical season) compound flooding is more likely to occur, and how well reanalysis data captures the dependence structure between the different flooding drivers.

Future water temperature of rivers in Switzerland under climate change investigated with physics-based models

Abstract: . Rivers are ecosystems highly sensitive to climate change and projected future increase in air temperature is expected to increase the stress for these ecosystems. Rivers are also an important socio-economical factor. In addition to changes in water availability, climate change will impact the temperature of rivers. This study presents a detailed analysis of river temperature and discharge evolution over the 21st century in Switzerland, a country covering a wide range of Alpine and lowland hydrological regimes. In total, 12 catchments are studied. They are situated both in the lowland Swiss Plateau and the Alpine regions and cover overall 10 % of the country’s area. This represents the so far largest study of climate change impacts on river temperature in Switzerland. The impact of climate change is assessed using a chain of physics-based models forced with the most recent climate change scenarios for Switzerland including low, mid, and high emissions pathways. A clear warming of river water is modelled during the 21st century, more pronounced for the high emission scenarios and toward the end of the century. For the period 2030–2040, median warming in river temperature of +1.1 °C for Swiss Plateau catchments and of +0.8 °C for Alpine catchments are expected compared to the reference period 1990–2000 (similar for all emission scenarios). At the end of the century (2080–2090), the median annual river temperature increase ranges between +0.9 °C for low emission and +3.5 °C for high emission scenarios for both Swiss Plateau and Alpine catchments. At the seasonal scale, the warming on the Swiss Plateau and in the Alpine regions exhibits different patterns. For the Swiss Plateau, the spring and fall warming is comparable to the warming in winter, while the summer warming is stronger but still moderate. In Alpine catchments, only a very limited warming is expected in winter. A marked discharge increase in winter and spring is expected in these catchments due to enhanced snowmelt and a larger fraction of liquid precipitation. Accordingly, the period of maximum discharge in Alpine catchments, currently occurring during mid-summer, will shift to earlier in the year by a few weeks (low emission) or almost two months (high emission) by the end of the century. In summer, the marked discharge reduction in Alpine catchments for high emission scenarios leads to an increase in sensitivity of water temperature to low discharge, which is not observed in the Swiss Plateau catchments. In addition, an important soil warming is expected due to glacier and snow cover decrease. These effects combined lead to a summertime river warming of +6.0 °C in Alpine catchments by the end of the century for high emission scenarios. Two metrics are used to show the adverse effects of river temperature increase both on natural and human systems. All results of this study along with the necessary source code are provided with this manuscript.

Coastal and orographic effects on extreme precipitation revealed by weather radar observations

Abstract: . The yearly exceedance probability of extreme precipitation of multiple durations is crucial for infrastructure design, risk management and policymaking. Local extremes emerge from the interaction of weather systems with local terrain features such as coastlines and orography, however multi-duration extremes do not follow exactly the patterns of cumulative precipitation and are still not well understood. High-resolution information from weather radars could help us better quantifying their patterns, but traditional extreme-value analyses based on radar records were found too inaccurate for quantifying the extreme intensities for impact studies. Here, we propose a novel methodology for extreme precipitation frequency analysis based on relatively short weather radar records, and we use it to investigate coastal and orographic effects on extreme precipitation of durations between 10 minutes and 24 hours. Combining 11 years of radar data with 10-minute rain gauge data in the southeastern Mediterranean, we obtain estimates of the 1 in 100 years intensities with ~22 % standard error, which is lower than those obtained using traditional approaches on rain gauge data. We identify three distinct regimes, which respond differently to coastal and orographic forcing: short durations (~10 minutes), related to peak convective rain rates; hourly durations (~1 hours), related to the yield of individual convective cells; and long durations (~6–24 hours), related to the accumulation of multiple convective cells and to stratiform processes. At short and hourly durations, extreme return levels peak at the coastline, while at longer durations they peak corresponding to the orographic barriers. The distributions tail heaviness is rather uniform above the sea and rapidly changes in presence of orography, with opposing directions at short (decreasing tail heaviness, with a peak at hourly durations) and long (increasing) durations. These distinct effects suggest that short-scale hazards such as urban pluvial floods could be more of concern for the coastal regions, while longer-scale hazards such as flash floods could be more relevant in mountainous areas.

Water Resources Management, Technology, and Culture in Ancient Iran

Abstract: . Iran is one of the countries facing high water risk because of its geographical features, climate variations, and uneven distribution of water resources. Iranians have practiced different water management strategies at various periods following the region's geo-climatological features, needs, tools, available resources (surface water and groundwater), political stability, economic power, and socio-cultural characteristics. This study is a brief history of water management in Iran from pre-civilization times to the end of the Islamic Golden Age (1219 AD). This study pointed out geo-climatological features have consistently been crucial intrinsic properties controlling water regime, settlement patterns, and other socioeconomic issues. These factors caused the early agricultural communities to emerge in water-rich regions of northwestern, western, and southwestern Iran. By the 4th Millennium BC, while water access became more difficult as population growth, economic activity, and urbanization progress, water resources' systematic development appeared in west and southwest Iran under the Mesopotamian civilization. However, despite all benefits, Mesopotamian water-based technology and administration could not meet all water demands in Iran's arid regions. For these reasons, qanats were developed in Persia by the First Persian Empire (Achaemenid Empire). No doubt, the Achaemenids (550–330 BC) should be regarded as one of the early civilizations that emerged in a land without sufficient rainfall and major rivers. In this time, idle and marginal lands of Iran and neighboring regions of the Middle East, North Africa, and Central Asia could be cultivated through the spread of qanat technology, enabling large groups of peasants to increase crop yields and incomes. After a period of recession during the Seleucid Empire (312–63 BC) and the Parthian Empire (247 BC–224 AD), water resources development gained momentum in the Sassanid era (224–651). In this period, the progress of urbanization was expeditious. Consciously, water resources development in Khuzestan plains (Shushtar and Dezful) was crucial for agricultural intensification, economic expansion, and civilization development. The Sassanids wisely adapted Greek watermills to the complicated topography, limited water availability, and variable climate of Iran to produce food. Although the Iranians practiced a new era of water governance under the Sassanid rule (224–651 AD), chaotic Iran in the Late Sasanian and Early Islamic Period led to severe weaknesses in water-related sectors. After Islam's arrival, the Muslim rulers turned their attention from fighting to set up an Islamic civilization to break the socioeconomic stagnation. To achieve the goal, they opened their scientific doors to science and technology centers. Despite all efforts made during the 8th–12th century, the lack of creativity and investment in promoting water technologies, prioritizing political considerations over social benefits, occurring wars, and poor water management have induced the Iranians to lose their power in developing water resources. In today's Iran, the past water-related problems have aggravated by uneven climate change, population rise, rapid industrialization, urban development, and unprecedented changes in lifestyle. Undoubtedly, solving these problems and moving towards a better future is not possible without addressing the past.

Snowpack dynamics in the Lebanese mountainsfrom quasi-dynamically downscaled ERA5reanalysis updated by assimilating remotely-sensedfractional snow-covered area

Abstract: . The snowpack over the Mediterranean mountains constitutes a key water resource for the downstream populations. However, its dynamics have not been studied in detail yet in many areas, mostly because of the scarcity of snowpack observations. In this work, we present a characterization of the snowpack over the two mountain ranges of Lebanon. To obtain the necessary snowpack information, we have developed a 1 km regional scale snow reanalysis (ICAR_assim) covering the period 2010–2017. ICAR_assim was developed by means of ensemble-based data assimilation of MODIS fractional snow-covered area (fSCA) through the energy and mass balance model the Flexible Snow Model (FSM2), using the Particle Batch Smoother (PBS). The meteorological forcing data was obtained by a regional atmospheric simulation developed through the Intermediate Complexity Atmospheric Research model (ICAR) nested inside a coarser regional simulation developed by the Weather Research and Forecasting model (WRF). The boundary and initial conditions of WRF were provided by the ERA5 atmospheric reanalysis. ICAR_assim showed very good agreement with MODIS gap-filled snow products, with a spatial correlation of R = 0.98 in the snow probability (P(snow)), and a temporal correlation of R = 0.88 in the day of peak snow water equivalent (SWE)Similarly, ICAR_assim has shown a correlation with the seasonal mean SWE of R = 0.75 compared with in-situ observations from Automatic Weather Stations (AWS). The results highlight the high temporal variability of the snowpack in the Lebanon ranges, with differences between Mount Lebanon and Anti-Lebanon that cannot be only explained by its hypsography been Anti-Lebanon in the rain shadow of Mount Lebanon. The maximum fresh water stored in the snowpack is in the middle elevations approximately between 2200 and 2500 m. a.s.l. Thus, the resilience to further warming is low for the snow water resources of Lebanon due to the proximity of the snowpack to the zero isotherm.

Technical Note: Low cost stage-camera system for continuous waterlevel monitoring in ephemeral streams

Abstract: . Monitoring ephemeral and intermittent streams is a major challenge in hydrology. While direct field observations are best to detect spatial patterns of flow persistence, on site inspections are time and labor intensive and may be impractical in difficult-to-access environments. Motivated by latest advancements of digital cameras and computer vision techniques, in this work, we describe the development and application of a stage-camera system to monitor the water level in ungauged headwater streams. The system encompasses a consumer grade wildlife camera with near infrared (NIR) night vision capabilities and a white pole that serves as reference object in the collected images. Time-lapse imagery is processed through a computationally inexpensive algorithm featuring image quantization and binarization, and water level time series are filtered through a simple statistical scheme. The feasibility of the approach is demonstrated through a set of benchmark experiments performed in controlled and natural settings, characterized by an increased level of complexity. Maximum mean absolute errors between stage-camera and reference data are approximately equal to 2 cm in the worst scenario that corresponds to severe hydrometeorological conditions. Our preliminary results are encouraging and support the scalability of the stage-camera in future implementations in a wide range of natural settings.

Drivers of drought-induced shifts in the water balance through a Budyko approach

Abstract: . An inconsistent statistical relationship between precipitation and runoff has been observed between drought and non-drought periods, with less runoff usually observed during droughts than would be predicted using non-drought relationships. Most studies have examined these shifts using multi-linear regression models, which can identify correlations but are less appropriate for analyzing underlying hydrologic mechanisms. In this analysis, we show how the Budyko framework can be leveraged to quantify the impact of shifts in water allocation during drought using 30 years of data for 14 basins in California. We distinguish ″regime″ shifts, which result from changes in the aridity index along the same Budyko curve, from ″partitioning shifts″, which imply a change in the Budyko parameter ω and thus to the relationship among water-balance components that governs partitioning of available water. Regime shifts are primarily due to measurable climatic changes, making them predictable based on drought conditions. Partitioning shifts are related to nonlinear and indirect catchment feedbacks to drought conditions and are thus harder to predict a priori. We show that regime shifts dominate changes in absolute runoff during droughts, but that gains or losses due to partitioning shifts are still significant. We further discuss how basin characteristics and feedbacks correlate and may influence these shifts, finding that low aridity, high baseflow, a shift from snow to rain, and resilience of high-elevation runoff correlate to an increase in runoff as a fraction of precipitation during droughts. This new application of the Budyko framework can help identify mechanisms influencing catchment response to drought, with implications for water management in arid and drought-prone regions.

Dynamical forcings in heavy precipitation events over Italy: lessons from the HyMeX SOP1 campaign

Abstract: . The first Special Observation Period (SOP1) of HyMeX (Hydrological cycle in the Mediterranean eXperiment) was held in Fall 2012 and focused on heavy precipitation events (HPEs) and floods in the northwestern Mediterranean. Nine intensive observation periods (IOPs) involved the three Italian target areas (north-eastern Italy, NEI; Liguria and Tuscany, LT; central Italy, CI), enabling an unprecedented analysis of precipitation systems in these regions. In the present work, we highlight the major findings emerging from the HyMeX campaign and in the subsequent research activity over the three target areas, by means of conceptual models and through the identification of the relevant recursive mesoscale features. For NEI, two categories of events (Upstream and Alpine HPEs) have been identified, which differ mainly in the temporal evolution of the stability of the upstream environment and of the intensity of the impinging flow (i.e., the Froude number). The numerical simulation of convection in the Po Valley was found very sensitive to small changes in the environmental conditions, especially when they are close to the threshold between “flow-over” and “flow-around” regimes. Some mesoscale features (e.g., the presence of a shallow pressure minimum in the eastern Po Valley) were identified as fundamental to adequately simulate the detailed evolution of severe convective episodes. For LT, HyMeX SOP1 focused on orographically-enhanced precipitation over the Apennines and quasi-stationary mesoscale convective systems over the sea or close to the coast. For the latter category of events, associated with the majority of the recent HPEs in the area, local-scale or large-scale convergence lines appear fundamental to trigger and sustain convection. These lines are affected not only by the orography of the region, but also by perturbations induced by Sardinia and Corsica on the environmental flow. Cold pools formed via evaporation of precipitation also played a major role in determining the position of the trigger at later times. The accurate representation of the moisture structure below 2 km is the key to an accurate simulation of the timing and location of precipitation. For CI, a high low-level moisture content and marked low-level convergence over the sea were critical to support deep convection in IOPs affecting the Tyrrhenian coast. Also, an elevated moisture plume from the Tropics was observed to locally reinforce the intensity of the updrafts. For HPEs affecting the Adriatic regions, generally a cut-off low over the Tyrrhenian Sea induces intense Bora over the Adriatic basin. Low-level convergence triggers convection over the sea, while orographic uplift produces stratiform precipitation. The Adriatic Sea plays a critical role mainly through air-sea exchanges, which modify the characteristics of the flow and in turn the effect of the orographic forcing.

Combined Impacts of Uncertainty in Precipitation and AirTemperature on Simulated Mountain System Recharge from anIntegrated Hydrologic Model

Abstract: . Mountainous regions act as the water towers of the world by producing streamflow and groundwater recharge, a function that is particularly important in semiarid regions. Quantifying rates of mountain system recharge is difficult, and hydrologic models offer a method to estimate recharge over large scales. These recharge estimates are prone to uncertainty from various sources including model structure and parameters. The quality of meteorological forcing datasets, particularly in mountainous regions, is a large source of uncertainty that is often neglected in groundwater investigations. In this contribution, we quantify the impact of uncertainty in both precipitation and air temperature forcing datasets on the simulated groundwater recharge in the mountainous watershed of the Kaweah River in California, USA. We make use of the integrated surface water – groundwater model, ParFlow.CLM and several gridded datasets commonly used in hydrologic studies, downscaled NLDAS-2, PRISM, Daymet, Gridmet, and TopoWx. Simulations indicate that across all forcing datasets, mountain front recharge is an important component of the water budget in the mountainous watershed accounting for 25–46 % of the annual precipitation, and ~90 % of the total mountain system recharge to the adjacent Central Valley aquifer. The uncertainty in gridded air temperature or precipitation datasets, when assessed individually, results in similar ranges of uncertainty in the simulated water budget. Variations in simulated recharge to changes in precipitation (elasticities) and air temperature (sensitivities) are larger than 1 % change in recharge per 1 % change in precipitation or 1-degree C change in temperature. The total volume of snowmelt is the primary factor creating the high water budget sensitivity; and snowmelt volume is influenced by both precipitation and air temperature forcings. The combined effect of uncertainty in air temperature and precipitation on recharge is additive, and results in uncertainty levels roughly equal to the sum of the individual uncertainties. Mountain system recharge pathways including mountain block recharge, mountain aquifer recharge, and mountain front recharge are less sensitive to changes in air temperature than changes in precipitation. Mountain front and mountain block recharge are more sensitive to changes in precipitation than other recharge pathways. The magnitude of uncertainty in the simulated water budget reflects the importance of developing high qualify meteorological forcing datasets in mountainous regions.

Patterns and dynamics of dissolved organic carbon exports from ariparian zone of a temperate, forested catchment

Abstract: . Export of dissolved organic carbon (DOC) from riparian zones (RZs) is an important, but poorly understood component of temperate catchment carbon budgets. This paper delineates explicit DOC source zones within the RZ of a small forested catchment in central Germany, and identifies and quantifies their dominant DOC export mechanism at high spatio-temporal resolution. Stream water DOC samples from differing hydrological situations were compared to riparian DOC groundwater and surface water samples and classified chemically (via Fourier-transform ion cyclotron resonance mass spectrometry) and spatially via a small-scale topographic analysis of the RZ at a resolution of 1 m. Explicit water fluxes from the resulting riparian DOC source zones were then simulated by a physically-based, fully-integrated numerical flow model (HydroGeoSphere). Chemical classification revealed two distinct DOC pools (DOCI and DOCII) in the RZ. The comparison of stream and riparian water samples indicated a predominant export of DOCI during wet conditions and high groundwater levels. The two DOC pools were spatially separated and mapped using a threshold value in high-resolution topographical wetness index (TWIHR). Hydrological modelling revealed that surface runoff from DOCI source zones with high TWIHR values dominated overall discharge generation and therefore DOC export. Although corresponding to only 15 % of the area in the studied RZ, the high TWIHR zones provided in total 1.5 times the load of DOC from the remaining 85 % of the area associated with the DOCII pool. Our results suggest that surface DOC export can play a dominant role for DOC export in RZs with overall low topographic relief and should be considered in DOC export models. We propose that proxies of spatial heterogeneity (here: TWIHR) can delineate the most active riparian source zones and provide a meaningful basis for improved model conceptualization of surficial DOC export.

Satellite observations reveal thirteen years of reservoir filling strategies, operating rules, and hydrological alterations in the Upper Mekong River Basin

Abstract: . The current situation in the Lancang–Mekong River Basin is emblematic of the issues faced by many transboundary basins around the world: riparian countries prioritize national water-energy policies and provide limited information on how major infrastructures are operated. In turn, such infrastructures and their management become a source of controversy. Here, we turn our attention to the Upper Mekong River, or Lancang, where a system of eleven mainstream dams controls about 55 % of the annual flow to Northern Thailand and Laos. Yet, assessing their actual impact is a challenging task because of the chronic lack of data on reservoir storage and dam release decisions. To overcome this challenge, we focus on the ten largest reservoirs and leverage satellite observations to infer 13-year time series of monthly storage variations. Specifically, we use area-storage curves (derived from a Digital Elevation Model) and time series of water surface area, which we estimate from Landsat images through a novel algorithm that removes the effects of clouds and other disturbances. We also use satellite radar altimetry data (Jason) to validate the results obtained from satellite imagery. Our results describe the evolution of the hydropower system and highlight the pivotal role played by Xiaowan and Nuozhadu reservoirs, which make up to ~85 % of the total system's storage in the Lancang River Basin. We show that these two reservoirs were filled in only two years, and that their operations did not change in response to the drought that occurred in the region in 2019–2020. Deciphering these operating strategies could help enrich existing monitoring tools and hydrological models, thereby supporting riparian countries in the design of more cooperative water-energy policies.

Revealing a significant isotopic offset between plant water and its sources using a global meta-analysis

Abstract: . Isotope-based approaches to study plant water sources rely on the assumption that root water uptake and within-plant water transport are non-fractionating processes. However, a growing number of studies have reported offsets between plant and source water stable isotope composition, for a wide range of ecosystems. These isotopic offsets can result in the erroneous attribution of source water used by plants and potential overestimations of groundwater uptake by the vegetation. We conducted a global meta-analysis to quantify the magnitude of these plant-source water isotopic offsets and explore whether their variability could be explained by either biotic or abiotic factors. Our database compiled 112 studies, spanning arctic to tropical biomes that reported the dual water isotope composition (δ2H and δ18O) of plant (stem) and source water, including soil water. We calculated 2H offsets in two ways: a line conditioned excess (LC-excess) that describes the 2H deviation from the local meteoric water line, and a soil water line conditioned excess (SW-excess), that describes the deviation from the soil water line, for each sampling campaign within each study. We tested for the effects of climate (air temperature and soil water content), soil class and plant traits (growth form, leaf habit, wood density and parenchyma fraction and mycorrhizal habit) on LC-excess and SW-excess. Globally, stem water was more depleted in 2H than soil water (SW-excess

Optimizing a backscatter forward operator using Sentinel-1 data over irrigated land

Abstract: . Worldwide, the amount of water used for agricultural purposes is rising and the quantification of irrigation is becoming a crucial topic. Because of the the limited availability of in situ observations, an increasing number of studies is focusing on the synergistic use of models and satellite data to detect and quantify irrigation. The parameterization of irrigation in large scale Land Surface Models (LSM) is improving, but it is still hampered by the lack of information about dynamic crop rotations or the extent of irrigated areas, and the mostly unknown timing and amount of irrigation. On the other hand, remote sensing observations offer an opportunity to fill this gap as they are directly affected by, and hence potentially able to detect, irrigation. Therefore, combining LSMs and satellite information through data assimilation can offer the optimal way to quantify the water used for irrigation. The aim of this study is to optimize a land modeling system, consisting of the Noah-MP LSM, coupled with a backscatter observation operator, over irrigated land in order to simulate backscatter predictions. This is a first step towards building a reliable data assimilation system to ingest level-1 Sentinel-1 observations. In this context, we tested how well modeled soil moisture and vegetation estimates from the Noah-MP LSM running within the NASA Land Information System (LIS), with or without irrigation simulation, are able to capture the signal of high-resolution Sentinel-1 backscatter observations over the Po river Valley, an important agricultural area in Northern Italy. Next, aggregated 1-km Sentinel-1 backscatter observations were used to calibrate a Water Cloud Model (WCM) as observation operator using simulated soil moisture and Leaf Area Index estimates. The WCM was calibrated with and without activating an irrigation scheme in Noah-MP and considering two different cost functions. Results demonstrate that activating an irrigation scheme provides the optimal calibration of the WCM, even if the irrigation estimates are inaccurate. The Bayesian optimization is shown to result in the best unbiased calibrated system, with minimal chance of having error cross correlations between the model and observations. Our time series analysis further confirms that Sentinel-1 is able to track the impact of human activities on the water cycle, highlighting its potential to improve irrigation, soil moisture and vegetation estimates via future data assimilation.

Can we use precipitation isotope outputs of Isotopic General Circulation Models to improve hydrological modeling in large mountainous catchments on the Tibetan Plateau

Abstract: . Issues related to large uncertainty and parameter equifinality have posed big challenges for hydrological modeling in cold regions where runoff generation processes are particularly complex. Tracer-aided hydrological models coupling modules to simulate the transportation and fractionation of water stable isotope are increasingly used to constrain parameter uncertainty and refine the parameterizations of specific hydrological processes in cold regions. However, commonly unavailability of site sampling of spatially-distributed precipitation isotope hampers the practical applications of tracer-aided models in large scale catchments. This study, taken the precipitation isotope data (isoGSM) derived from the Isotopic General Circulation Models (iGCM) as an example, explored its utility in driving a tracer-aided hydrological model in the Yarlung Tsangpo River basin (YTR, around 2 × 105 km2) on the Tibetan Plateau (TP). The isoGSM product was first corrected based on the biases between gridded precipitation isotope estimates and limited site sampling measurements. Model simulations driven by the corrected isoGSM data were then compared with those forced by spatially interpolated precipitation isotope from site sampling measurements. Our results indicated that: (1) spatial precipitation isotope derived from the isoGSM data helped to reduce modeling uncertainty and improve parameter identifiability in a large mountainous catchment on the TP, in comparison to a calibration method using discharge and snow cover area fraction without any information of water isotope; (2) model parameters estimated by the corrected isoGSM data presented higher transferability to nested sub-basins and produced higher model performance in the validation period than that estimated by the interpolated precipitation isotope data from site sampling measurements; (3) model calibration procedure forced by the corrected isoGSM data successfully rejected parameter sets that overestimated glacier melt contribution and gave more reliable contributions of runoff components, indicating the corrected isoGSM data served as a better choice to provide informative spatial precipitation isotope than the interpolated data from site sampling measurements at macro scale. This work suggested plausible utility of combining isoGSM data with measurements from a sparse sampling network in improving hydrological modeling in large mountainous catchments.

Toward hyper-resolution global hydrological models including human activities: application to Kyushu Island, Japan

Abstract: . Global hydrological models that include human activities are powerful tools for assessing water availability and use at global and continental scales. Such models are typically applied at a spatial resolution of 30 arcminutes (approximately 50 km). In recent years, some 5-arcminute (9-km) applications have been reported, but with numerous technical challenges, including the validation of calculations for more than a million grid cells and the conversion of simulation results into meaningful information relevant to water resource management. Here, the H08 global water resources model was applied in two ways to Kyushu Island in Japan at resolution of 1 arcminute (2 km), and the detailed results were compared. One method involved feeding interpolated global meteorological and geographic data into the default global model (GLB; in accordance with previous high-resolution applications). For the other method, locally derived boundary conditions were input to the localized model (LOC; this method can be easily extended and applied to other regions, at least across Japan). The results showed that GLB cannot easily reproduce the historical record, especially for variables related to human activities (e.g., dam operation and water withdrawal). LOC is capable of estimating natural and human water balance components at daily time scales and providing reliable information for regional water resource assessment. The results highlight the importance of improving data preparation and modeling methods to represent water management and use in hyper-resolution global hydrology simulations.

A combined use of in situ and satellite-derived observations to characterize surface hydrology and its variability in the Congo River Basin

Abstract: . The Congo River Basin (CRB) is the second largest river system in the world, but its hydroclimatic characteristics remain relatively poorly known. Here, we jointly analyze a large record of in situ and satellite-derived observations, including long term time series of Surface Water Height (SWH) from radar altimetry (a total of 2,311 virtual stations) and surface water extent (SWE) from a multi-satellite technique to better characterize CRB surface hydrology and its variability. Firstly, we show that SWH from radar altimetry multi-missions agree well with in situ water stage at various locations, with root mean square deviation varying from 10 cm (with Sentinel-3A) to 75 cm (with European Remote Sensing-2). SWE from multi-satellite observations also shows a good behavior over a ~25-year period against in situ observations from sub-basin to basin scale. Both datasets help to better characterize the large spatial and temporal variability of hydrological patterns across the basin, with SWH exhibiting annual amplitude of more than 5 m in the northern sub-basins while Congo main-stream and Cuvette Centrale tributaries vary in smaller proportions (1.5 m to 4.5 m). Furthermore, SWH and SWE help better illustrate the spatial distribution and different timings of the CRB annual flood dynamic and how each sub-basin and tributary contribute to the hydrological regime at the outlet of the basin (the Brazzaville/Kinshasa station), including its peculiar bi-modal pattern. Across the basin, we jointly use SWH and SWE to estimate time lag and water travel time to reach the Brazzaville/Kinshasa station, ranging from 0–1 month in its vicinity downstream the basin up to 3 months in remote areas and small tributaries. Northern sub-basins and the central Congo region highly contribute to the large peak in December–January while the southern part of the basin supplies water to both hydrological peaks, in particular to the moderate one in April–May. The results are supported using in situ observations at various locations in the basin. Our results contribute to a better characterization of the hydrological variability in the CRB and represent an unprecedented source of information for hydrological modeling and to study hydrological processes over the region.

Simultaneous assimilation of water levels from river gauges and satellite flood maps for near-real time flood mapping

Abstract: . Hydro-meteo hazard Early Warning Systems (EWSs) are operating in many regions of the world to mitigate nuisance effects of floods. EWSs performances are majorly impacted by the computational burden and complexity affecting flood prediction tools, especially for ungauged catchments that lack adequate river flow gauging stations. Earth Observation (EO) systems may surrogate to the lack of fluvial monitoring systems supporting the setting up of affordable EWSs. But, EO data, constrained by spatial and temporal resolution limitations, are not sufficient alone, especially at medium-small scales. Multiple sources of distributed flood observations need to be used for managing uncertainties of flood models, but this is not a trivial task for EWSs. In this work, a near real-time flood modelling approach is developed and tested for the simultaneous assimilation of both water level observations and EO-derived flood extents. An integrated physically-based flood wave generation and propagation modelling approach, that implements a Ensemble Kalman Filter, a parsimonious geomorphic rainfall-runoff algorithm (WFIUH) and a Quasi-2D hydraulic algorithm, is proposed. A data assimilation scheme is tested that retrieves distributed observed water depths from satellite images to update 2D hydraulic modelling state variables. Performances of the proposed approach are tested on a flood event for the Tiber river basin in central Italy. The selected case study shows varying performances depending if local and distributed observations are separately or simultaneously assimilated. Results suggest that the injection of multiple data sources into a flexible data assimilation framework, constitute an effective and viable advancement for flood mitigation tackling EWSs data scarcity, uncertainty and numerical stability issues.

Potential effects of cryogenic extraction biases on inferences drawnfrom xylem water deuterium isotope ratios: case studies using stableisotopes to infer plant water sources

Abstract: . Recent studies have demonstrated that plant and soilwater extraction techniques can introduce biases and uncertainties in stable isotope analyses. Here we show how recently documented δ2H biases resulting from cryogenic vacuum distillation of water from xylem tissues may have influenced the conclusions of five previous studies, including ours, that have used δ2H to infer plant water sources. Cryogenic extraction biases that reduce xylem water δ2H will also introduce an artifactual evaporation signal in dual-isotope (δ2H vs. δ18O) analyses. Calculations that estimate the composition of the source precipitation of xylem waters by compensating for their apparent evaporation will amplify the bias in δ2H, and also introduce new biases in the δ18O of the inferred pre-evaporation source precipitation. Cryogenic extraction biases may substantially alter plant water source attributions if the spread in δ2H among the potential end members is relatively narrow. By contrast, if the spread in δ2H among the potential end members is relatively wide, the impact of cryogenic extraction biases will be less pronounced, and thus suggestions that these biases universally invalidate inferences drawn from plant water δ2H are unwarranted. Nonetheless, until reliable correction factors for cryogenic extraction biases become available, their potential impact should be considered in studies using xylem water isotopes.

65-year changes of annual streamflow volumes across Europe with a focus on the Mediterranean basin

Abstract: . Determining the spatio-temporal variability of annual streamflow volume plays a relevant role in hydrology for improving and implementing sustainable and resilient policies and practices of water resource management. This study investigates annual streamflow volume trends in a newly-assembled, consolidated and validated dataset of daily mean river flow records from more than 3,000 stations, which cover near-natural basins in more than 40 countries across Europe. Although the dataset contains streamflow time-series from 1850 to 2015 in some stations, the statistical analyses were carried out by including observations from 1950 to 2015 in order to have a consistent and reliable dataset over the continent. Trends were detected calculating the slope of Theil-Sen's line over the annual anomalies of streamflow volume. The results show annual streamflow volume trends emerged at European scale, with a marked negative tendency in Mediterranean regions (about −1 × 103 m3/(km2 year)) and a generally positive trend in northern ones (about 0.5 × 103 m3/(km2 year)). The annual streamflow volume trend patterns appear in agreement with the continental-scale climate change observations in response to climate change drivers. In the Mediterranean area, the declining of annual streamflow volumes started in 1965 and since early 80' volumes are consistently lower than the average. The spatio-temporal annual streamflow volume patterns observed in this work can help to contextualize short-term trends and regional studies already available in the scientific literature as well as to provide a valid benchmark for further accurate quantitative analysis on annual streamflow volumes.

Testing a maximum evaporation theory over saturated land: Implications for potential evaporation estimation

Abstract: . State-of-the-art evaporation models usually assume the net radiation (Rn) and surface temperature (Ts; or near-surface air temperature) to be independent forcings on evaporation. However, Rn depends directly on Ts via outgoing longwave radiation and this creates a physical coupling between Rn and Ts that extends to evaporation. In this study, we test a maximum evaporation theory originally developed for global ocean over saturated land surfaces, which explicitly acknowledges the interactions between radiation, Ts and evaporation. Similar to the ocean surface, we find a maximum evaporation (LEmax) emerges over saturated land that represents a generic trade-off between a lower Rn and a higher evaporation fraction as Ts increases. Compared with flux site observations at the daily scale, we show that LEmax corresponds well to observed evaporation under non-water-limited conditions and that the Ts at which LEmax occurs also corresponds with the observed Ts. Our results suggest that saturated land surfaces behave essentially the same as ocean surfaces at time scales longer than a day and further imply that the maximum evaporation concept is a natural attribute of saturated land surfaces, which can be the basis of a new approach to estimating evaporation.

Storylines of UK drought based on the 2010–2012 event

Abstract: . Spatially extensive multi-year hydrological droughts cause significant environmental stress. Given the impacts of climate change, the UK is expected to remain vulnerable to future multi-year droughts. Existing approaches to quantify hydrological impacts of climate change are often scenario-driven and may miss out plausible outcomes with significant impacts. Event-based storyline approaches aim to quantify storylines of how observed events could hypothetically have unfolded in alternative ways. This study uses the 2010–2012 drought, the most recent period of severe hydrological drought in the UK, as a basis, and analyses counterfactual storylines based on changes to 1) precondition severity, 2) temporal drought sequence, and 3) climate change. Evidence from multiple storylines shows that maximum intensity, mean deficit and duration of the 2010–2012 drought were highly conditioned by its meteorological preconditions, particularly for northern catchments at shorter time scales. Recovery time from progressively drier preconditions reflect both spatial variation in drought conditions and the role of physical catchment characteristics, particularly hydrogeology in the propagation of multi-year droughts. Two plausible storylines of an additional dry year with dry winter conditions repeated either before the observed drought or replacing the observed dramatic drought termination confirm the vulnerability of UK catchments to a three dry winter scenario. Applying the UKCP18 climate projections, we find that drought conditions worsen with global warming with a mitigation of drought conditions by wetter winters in northern catchments at high warming levels. Comparison of the storylines with a benchmark drought (1975–76) and a protracted multi-year drought (1989–93) shows that for each storyline, drought conditions could have matched and exceeded those experienced during the past droughts at catchments across the UK, particularly for southern catchments. The construction of storylines based on observed events can complement existing methods to stress test UK catchments against plausible unrealized droughts.

In-stream Escherichia Coli Modeling Using high-temporal-resolution data with deep learning and process-based models

Abstract: . Contamination of surface waters through microbiological pollutants is a major concern to public health. Although long-term and high-frequency E. coli monitoring can help prevent diseases from fecal pathogenic microorganisms, this monitoring is time consuming and expensive. Process-driven models are an alternative method for determining fecal pathogenic microorganisms. However, process-based modeling still has limitations in improving the model accuracy because of the complex mechanistic relationships among hydrological and environmental variables. On the other hand, with the rise in data availability and computation power, the use of data-driven models is increasing. Therefore, in this study, we simulated the transport of Escherichia coli (E. coli) in a 0.6 km² tropical headwater catchment located in Lao PDR using a deep learning model and a process-based model. The deep learning model was built using the long short-term memory (LSTM) technique, whereas the process-based model was constructed using the Hydrological Simulation Program–FORTRAN (HSPF). First, we calibrated both models for surface as well as for subsurface flow. Then, we simulated the E. coli transport with 6 min time steps with both the HSPF and LSTM models. The LSTM provided accurate results for surface and subsurface flow, by showing 0.51 and 0.64 of Nash–Sutcliffe Efficiency (NSE), respectively, whereas the NSE values yielded by the HSPF were −0.7 and 0.59 for surface and subsurface flow. The simulated E. coli concentration from LSTM also improved, yielding an NSE of 0.35, whereas the HSPF showed an unacceptable performance, with an NSE value of −3.01. This is because of the limitation of HSPF in capturing the dynamics of E. coli with land-use change. The simulated E. coli concentration showed rise and drop patterns corresponding to annual changes in land use. This study shows the application of deep learning-based models as an efficient alternative to process-based models for E. coil fate and transport simulation at the catchment scale.

High-resolution drought simulations and comparison to soil moisture observations in Germany

Abstract: . The 2018–2020 consecutive drought events in Germany resulted in impacts related with several sectors such as agriculture, forestry, water management, industry, energy production and transport. A major national operational drought information system is the German Drought Monitor (GDM), launched in 2014. It provides daily soil moisture (SM) simulated with the mesoscale hydrological model (mHM) and its related soil moisture index at a spatial resolution of 4 × 4 km2. Key to preparedness for extreme drought events are high-resolution information systems. The release of the new soil map BUEK200 allowed to increase the model resolution to ~1.2 × 1.2 km2, which is used in the second version of the GDM. In this paper, we explore the ability to provide drought information on the one-kilometer scale in Germany. Therefore, we compare simulated SM dynamics using homogenized and deseasonalized SM observations to evaluate the high-resolution drought simulations of the GDM. These SM observations are obtained from single profile measurements, spatially distributed sensor networks, cosmic-ray neutron stations and lysimeters at 40 sites in Germany. The results show that the agreement of simulated and observed SM dynamics is especially high in the vegetation period (0.84 median correlation R) and lower in winter (0.59 median R). Lower agreement in winter results from methodological uncertainties in simulations as well as in observations. Moderate but significant improvements between the first and second GDM version to observed SM were found in correlations for autumn (+0.07 median R) and winter (+0.12 median R). The annual drought intensity ranking and the spatial structure of drought events over the past 69 years is comparable for the two GDM versions. However, the higher resolution of the second GDM version allows a much more detailed representation of the spatial variability of SM, which is particularly beneficial for local risk assessments. Furthermore, the results underline that nationwide drought information systems depend both on appropriate simulations of the water cycle and a broad, high-quality observational soil moisture database.

Disentangling Sources of Future Uncertainties for WaterManagement in Sub-Saharan River Basins

Abstract: . Water management in sub-Saharan African river basins is challenged by uncertain future climatic, social and economical patterns, potentially causing diverging water demands and availability, as well as by multi-stakeholder dynamics, resulting in evolving conflicts and tradeoffs. In such contexts, a better understanding of the sensitivity of water management to the different sources of uncertainty can support policy makers in identifying robust water supply policies balancing optimality and low vulnerability against likely adverse future conditions. This paper contributes an integrated decision-analytic framework combining optimization, robustness, sensitivity and uncertainty analysis to retrieve the main sources of vulnerability to optimal and robust reservoir operating policies across multi-dimensional objective spaces. We demonstrate our approach onto the lower Umbeluzi river basin, Mozambique, an archetypal example of sub-Saharan river basin, where surface water scarcity compounded by substantial climatic variability, uncontrolled urbanization rate, and agricultural expansion are hampering the Pequenos Lipompos dam ability of supplying the agricultural, energy and urban sectors. We adopt an Evolutionary Multi-Objective Direct Policy Search optimization approach for designing optimal operating policies, whose robustness against social, agricultural, infrastructural and climatic uncertainties is assessed via robustness analysis. We then implement the GLUE and PAWN uncertainty and sensitivity analysis methods for disentangling the main challenges to the sustainability of the operating policies and quantifying their impacts on the urban, agricultural and energy sectors. Numerical results highlight the importance of robustness analysis when dealing with uncertain scenarios, with optimal-non robust reservoir operating policies largely dominated by robust control strategies across all stakeholders. Furthermore, while robust policies are usually vulnerable only to hydrological perturbations and are able to sustain the majority of population growth and agricultural expansion scenarios, non-robust policies are sensitive also to social and agricultural changes, and require structural interventions to ensure stable supply.

The importance of ecosystem adaptation on hydrological model predictions in response to climate change

Abstract: . To predict future hydrological behavior in a changing world, often use is made of models calibrated on past observations, disregarding that hydrological systems, hence model parameters, will change as well. Yet, ecosystems likely adjust their root-zone storage capacity, which is the key parameter of any hydrological system, in response to climate change. In addition, other species might become dominant, both under natural and anthropogenic influence. In this study, we propose a top-down approach, which directly uses projected climate data to estimate how vegetation adapts its root-zone storage capacity at the catchment scale in response to changes in magnitude and seasonality of hydro-climatic variables. Additionally, the Budyko characteristics of different dominant ecosystems in sub-catchments are used to simulate the hydrological behavior of potential future land-use change, in a space-for-time exchange. We hypothesize that changes in the predicted hydrological response as a result of 2 K global warming are more pronounced when explicitly considering changes in the sub-surface system properties induced by vegetation adaptation to changing environmental conditions. We test our hypothesis in the Meuse basin in four scenarios designed to predict the hydrological response to 2 K global warming in comparison to current-day conditions using a process-based hydrological model with (a) a stationary system, i.e. no changes in the root-zone storage capacity of vegetation and historical land use, (b) an adapted root-zone storage capacity in response to a changing climate but with historical land use, and (c, d) an adapted root-zone storage capacity considering two hypothetical changes in land use from coniferous plantations/agriculture towards broadleaved forest and vice-versa. We found that the larger root-zone storage capacities (+34 %) in response to a more pronounced seasonality with drier summers under 2 K global warming strongly alter seasonal patterns of the hydrological response, with an overall increase in mean annual evaporation (+4 %), a decrease in recharge (−6 %) and a decrease in streamflow (−7 %), compared to predictions with a stationary system. By integrating a time-dynamic representation of changing vegetation properties in hydrological models, we make a potential step towards more reliable hydrological predictions under change.

Climatic expression of rainfall on soil moisture dynamics in drylands

Abstract: . In drylands, characterised by water scarcity and frequent meteorological droughts, knowledge of soil moisture dynamics and its drivers (evapotranspiration, soil physical properties and the timing and sequencing of precipitation events) is fundamental to understanding changes in water availability to plants and human society, especially under a nonstationary climate. Given the episodic and stochastic nature of rainfall in drylands and the limited availability of data in these regions, we sought to explore what effects the temporal resolution of precipitation data has on soil moisture and how soil moisture distributions might evolve under different scenarios of climate change. Such information is critical for anticipating the impact of a changing climate on dryland communities across the globe, especially those that depend on rainfed agriculture and groundwater wells for drinking water for humans and livestock. A major challenge to understanding soil moisture in response to climate is the availability of precipitation datasets for dryland regions across the globe. Gridded precipitation data may only be available for daily or weekly time periods, even though rainstorms in drylands often occur on much shorter time scales, but it is currently unknown how this timescale mismatch might affect our understanding of soil moisture. Numerical modelling enables retrodiction or prediction of how climate translates into dynamically evolving moisture within the soil profile. It can be used to explore how climate data at different temporal resolutions affect these soil moisture dynamics, as well as to explore the influence of shifts in rainfall characteristics (e.g., storm intensity) under potential scenarios of climate change. This study uses Hydrus 1-D, to investigate the dynamics of soil moisture over a period of decades in response to the same underlying rainfall data resolved at hourly, daily, and weekly resolutions, as well as to step changes in rainfall delivery, which is expected under a warming atmosphere. We parameterised the model using rainfall, evaporative demand, and soils data from the semi-arid Walnut Gulch Experimental Watershed (WGEW) in SE Arizona, but we present the results as a generalized study of how rainfall resolution and shifts in rainfall intensity may affect dryland soil moisture at different depths. Our results indicate that hourly or better rainfall resolution captures the dynamics of soil moisture in drylands, and that critical information on soil water content, moisture availability to vegetation, actual evapotranspiration, and deep percolation of infiltrated water is lost when soil moisture modelling is driven by rainfall data at coarser temporal resolutions (daily, weekly). We further show that modest changes in rainfall intensity dramatically shift soil water content and the overall water balance. These findings are relevant to the prediction of soil moisture for crop yield forecasts, for adaptation to climate-related risks, and for anticipating the challenges of water scarcity and food insecurity in dryland communities around the globe, where available datasets are of low spatial and temporal resolution.

Analysing river network dynamics and active length - discharge relationship using water presence sensors

Abstract: Despite the importance of temporary streams for the provision of key ecosystem services, their experimental monitoring remains challenging because of the practical difficulties in performing accurate high-frequency surveys of the flowing portion of river networks In this study, about 30 electrical resistance (ER) sensors were deployed in a high relief 26 km2 catchment of the Italian Alps to monitor the spatio-temporal dynamics of the active river network during the fall of 2019 The set-up of the ER sensors was personalized to make them more flexible for the deployment in the field and more accurate under low flow conditions Available ER data were analyzed, compared to field based estimates of the nodes' persistency and then used to generate a sequence of maps representing the active reaches of the stream network with a sub-daily temporal resolution This allowed a proper estimate of the joint variations of active river network length (L) and catchment discharge (Q) during the entire study period Our analysis revealed a high cross-correlation between the statistics of individual ER signals and the flow persistencies of the cross sections where the sensors were placed The observed spatial and temporal dynamics of the actively flowing channels also revealed the diversity of the hydrological behaviour of distinct zones of the study catchment, which was attributed to differences in the catchment geology and stream-bed composition The more pronounced responsiveness of the total active length to small precipitation events as compared to the catchment discharge led to important hysteresis in the L vs Q relationship, thereby impairing the performances of a power-law model frequently used in the literature to relate these two quantities Consequently, in our study site the adoption of a unique power-law L-Q relationship to infer flowing length variability from observed discharges would underestimate the actual variations of L by 40% Our work emphasizes the potential of ER sensors for analysing spatio-temporal dynamics of active channels in temporary streams, discussing the major limitations of this type of technology emerging from the specific application presented herein

Regionalization of hydrological model parameters using gradient boosting machine

Abstract: . Regionalization of hydrological model parameters is key to hydrological predictions in ungauged basins. The commonly used multiple linear regression (MLR) method may not be applicable in complex and nonlinear relationships between model parameters and watershed properties. Moreover, most regionalization methods assume lumped parameters for each catchment without considering within-catchment heterogeneity. Here we incorporated the Penman-Monteith-Leuning (PML) equation into the Distributed Time-Variant Gain Model (DTVGM) to improve the mechanistic representation of the evapotranspiration process. We calibrated six key model parameters grid-by-grid across China using a multivariable calibration strategy, which incorporates spatiotemporal runoff and evapotranspiration (ET) datasets (0.25°, monthly) as reference. In addition, we used the gradient boosting machine (GBM), a machine learning technique, to portray the dependence of model parameters on soil and terrain attributes in four distinct climatic zones across China. We show that the modified DTVGM could reasonably estimate the runoff and ET over China using the calibrated parameters, but performed better in humid than arid regions for the validation period. The regionalized parameters by the GBM method exhibited better spatial coherence relative to the calibrated grid-by-grid parameters. In addition, GBM outperformed the stepwise MLR method in both parameter regionalization and gridded runoff simulations at national scale, though the improvement is not significant pertaining to watershed streamflow validation due to most of the watersheds being located in humid regions. We also revealed that the slope, saturated soil moisture content, and elevation are the most important explanatory variables to inform model parameters based on the GBM approach. The machine-learning-based regionalization approach provides an effective alternative to deriving hydrological model parameters by using watershed properties in ungauged regions.