Showing papers in "Hydrology and Earth System Sciences in 2019"

Technical note: Inherent benchmark or not? Comparing Nash-Sutcliffe and Kling-Gupta efficiency scores

Abstract: . A traditional metric used in hydrology to summarize model performance is the Nash–Sutcliffe efficiency (NSE). Increasingly an alternative metric, the Kling–Gupta efficiency (KGE), is used instead. When NSE is used, NSE = 0 corresponds to using the mean flow as a benchmark predictor. The same reasoning is applied in various studies that use KGE as a metric: negative KGE values are viewed as bad model performance, and only positive values are seen as good model performance. Here we show that using the mean flow as a predictor does not result in KGE = 0, but instead KGE = 1 - √ 2 ≈ - 0.41 . Thus, KGE values greater than −0.41 indicate that a model improves upon the mean flow benchmark – even if the model's KGE value is negative. NSE and KGE values cannot be directly compared, because their relationship is non-unique and depends in part on the coefficient of variation of the observed time series. Therefore, modellers who use the KGE metric should not let their understanding of NSE values guide them in interpreting KGE values and instead develop new understanding based on the constitutive parts of the KGE metric and the explicit use of benchmark values to compare KGE scores against. More generally, a strong case can be made for moving away from ad hoc use of aggregated efficiency metrics and towards a framework based on purpose-dependent evaluation metrics and benchmarks that allows for more robust model adequacy assessment.

Daily evaluation of 26 precipitation datasets using Stage-IV gauge-radar data for the CONUS

Abstract: . New precipitation ( P ) datasets are released regularly, following innovations in weather forecasting models, satellite retrieval methods, and multi-source merging techniques. Using the conterminous US as a case study, we evaluated the performance of 26 gridded (sub-)daily P datasets to obtain insight into the merit of these innovations. The evaluation was performed at a daily timescale for the period 2008–2017 using the Kling–Gupta efficiency (KGE), a performance metric combining correlation, bias, and variability. As a reference, we used the high-resolution (4 km) Stage-IV gauge-radar P dataset. Among the three KGE components, the P datasets performed worst overall in terms of correlation (related to event identification). In terms of improving KGE scores for these datasets, improved P totals (affecting the bias score) and improved distribution of P intensity (affecting the variability score) are of secondary importance. Among the 11 gauge-corrected P datasets, the best overall performance was obtained by MSWEP V2.2, underscoring the importance of applying daily gauge corrections and accounting for gauge reporting times. Several uncorrected P datasets outperformed gauge-corrected ones. Among the 15 uncorrected P datasets, the best performance was obtained by the ERA5-HRES fourth-generation reanalysis, reflecting the significant advances in earth system modeling during the last decade. The (re)analyses generally performed better in winter than in summer, while the opposite was the case for the satellite-based datasets. IMERGHH V05 performed substantially better than TMPA-3B42RT V7, attributable to the many improvements implemented in the IMERG satellite P retrieval algorithm. IMERGHH V05 outperformed ERA5-HRES in regions dominated by convective storms, while the opposite was observed in regions of complex terrain. The ERA5-EDA ensemble average exhibited higher correlations than the ERA5-HRES deterministic run, highlighting the value of ensemble modeling. The WRF regional convection-permitting climate model showed considerably more accurate P totals over the mountainous west and performed best among the uncorrected datasets in terms of variability, suggesting there is merit in using high-resolution models to obtain climatological P statistics. Our findings provide some guidance to choose the most suitable P dataset for a particular application.

Towards learning universal, regional, and local hydrological behaviors via machine learning applied to large-sample datasets

Abstract: . Regional rainfall–runoff modeling is an old but still mostly outstanding problem in the hydrological sciences. The problem currently is that traditional hydrological models degrade significantly in performance when calibrated for multiple basins together instead of for a single basin alone. In this paper, we propose a novel, data-driven approach using Long Short-Term Memory networks (LSTMs) and demonstrate that under a “big data” paradigm, this is not necessarily the case. By training a single LSTM model on 531 basins from the CAMELS dataset using meteorological time series data and static catchment attributes, we were able to significantly improve performance compared to a set of several different hydrological benchmark models. Our proposed approach not only significantly outperforms hydrological models that were calibrated regionally, but also achieves better performance than hydrological models that were calibrated for each basin individually. Furthermore, we propose an adaption to the standard LSTM architecture, which we call an Entity-Aware-LSTM (EA-LSTM), that allows for learning catchment similarities as a feature layer in a deep learning model. We show that these learned catchment similarities correspond well to what we would expect from prior hydrological understanding.

Seasonal origins of soil water used by trees

Abstract: . Rain recharges soil water storages and either percolates downward into aquifers and streams or is returned to the atmosphere through evapotranspiration. Although it is commonly assumed that summer rainfall recharges plant-available water during the growing season, the seasonal origins of water used by plants have not been systematically explored. We characterize the seasonal origins of waters in soils and trees by comparing their midsummer isotopic signatures ( δ2H ) to seasonal isotopic cycles in precipitation, using a new seasonal origin index. Across 182 Swiss forest sites, xylem water isotopic signatures show that summer rain was not the predominant water source for midsummer transpiration in any of the three sampled tree species. Beech and oak mostly used winter precipitation, whereas spruce used water of more diverse seasonal origins. Even in the same plots, beech consistently used more winter precipitation than spruce, demonstrating consistent niche partitioning in the rhizosphere. All three species' xylem water isotopes indicate that trees used more winter precipitation in drier regions, potentially mitigating their vulnerability to summer droughts. The widespread occurrence of winter isotopic signatures in midsummer xylem implies that growing-season rainfall may have minimally recharged the soil water storages that supply tree growth, even across diverse humid climates (690–2068 mm annual precipitation). These results challenge common assumptions concerning how water flows through soils and is accessed by trees. Beyond these ecological and hydrological implications, our findings also imply that stable isotopes of δ18O and δ2H in plant tissues, which are often used in climate reconstructions, may not reflect water from growing-season climates.

Selection of multi-model ensemble of general circulation models for the simulation of precipitation and maximum and minimum temperature based on spatial assessment metrics

Abstract: . The climate modelling community has trialled a large number of metrics for evaluating the temporal performance of general circulation models (GCMs), while very little attention has been given to the assessment of their spatial performance, which is equally important. This study evaluated the performance of 36 Coupled Model Intercomparison Project 5 (CMIP5) GCMs in relation to their skills in simulating mean annual, monsoon, winter, pre-monsoon, and post-monsoon precipitation and maximum and minimum temperature over Pakistan using state-of-the-art spatial metrics, SPAtial EFficiency, fractions skill score, Goodman–Kruskal's lambda, Cramer's V, Mapcurves, and Kling–Gupta efficiency, for the period 1961–2005. The multi-model ensemble (MME) precipitation and maximum and minimum temperature data were generated through the intelligent merging of simulated precipitation and maximum and minimum temperature of selected GCMs employing random forest (RF) regression and simple mean (SM) techniques. The results indicated some differences in the ranks of GCMs for different spatial metrics. The overall ranks indicated NorESM1-M, MIROC5, BCC-CSM1-1, and ACCESS1-3 as the best GCMs in simulating the spatial patterns of mean annual, monsoon, winter, pre-monsoon, and post-monsoon precipitation and maximum and minimum temperature over Pakistan. MME precipitation and maximum and minimum temperature generated based on the best-performing GCMs showed more similarities with observed precipitation and maximum and minimum temperature compared to precipitation and maximum and minimum temperature simulated by individual GCMs. The MMEs developed using RF displayed better performance than the MMEs based on SM. Multiple spatial metrics have been used for the first time for selecting GCMs based on their capability to mimic the spatial patterns of annual and seasonal precipitation and maximum and minimum temperature. The approach proposed in the present study can be extended to any number of GCMs and climate variables and applicable to any region for the suitable selection of an ensemble of GCMs to reduce uncertainties in climate projections.

Climate change, reforestation/afforestation, and urbanization impacts on evapotranspiration and streamflow in Europe.

Abstract: . Since the 1950s, Europe has undergone large shifts in climate and land cover. Previous assessments of past and future changes in evapotranspiration or streamflow have either focussed on land use/cover or climate contributions or on individual catchments under specific climate conditions, but not on all aspects at larger scales. Here, we aim to understand how decadal changes in climate (e.g. precipitation, temperature) and land use (e.g. deforestation/afforestation, urbanization) have impacted the amount and distribution of water resource availability (both evapotranspiration and streamflow) across Europe since the 1950s. To this end, we simulate the distribution of average evapotranspiration and streamflow at high resolution (1 km 2 ) by combining (a) a steady-state Budyko model for water balance partitioning constrained by long-term (lysimeter) observations across different land use types, (b) a novel decadal high-resolution historical land use reconstruction, and (c) gridded observations of key meteorological variables. The continental-scale patterns in the simulations agree well with coarser-scale observation-based estimates of evapotranspiration and also with observed changes in streamflow from small basins across Europe. We find that strong shifts in the continental-scale patterns of evapotranspiration and streamflow have occurred between the period around 1960 and 2010. In much of central-western Europe, our results show an increase in evapotranspiration of the order of 5 %–15 % between 1955–1965 and 2005–2015, whereas much of the Scandinavian peninsula shows increases exceeding 15 %. The Iberian Peninsula and other parts of the Mediterranean show a decrease of the order of 5 %–15 %. A similar north–south gradient was found for changes in streamflow, although changes in central-western Europe were generally small. Strong decreases and increases exceeding 45 % were found in parts of the Iberian and Scandinavian peninsulas, respectively. In Sweden, for example, increased precipitation is a larger driver than large-scale reforestation and afforestation, leading to increases in both streamflow and evapotranspiration. In most of the Mediterranean, decreased precipitation combines with increased forest cover and potential evapotranspiration to reduce streamflow. In spite of considerable local- and regional-scale complexity, the response of net actual evapotranspiration to changes in land use, precipitation, and potential evaporation is remarkably uniform across Europe, increasing by ∼ 35–60 km 3 yr −1 , equivalent to the discharge of a large river. For streamflow, effects of changes in precipitation ( ∼ 95 km 3 yr −1 ) dominate land use and potential evapotranspiration contributions ( ∼ 45–60 km 3 yr −1 ). Locally, increased forest cover, forest stand age, and urbanization have led to significant decreases and increases in available streamflow, even in catchments that are considered to be near-natural.

Unexplained hydrogen isotope offsets complicate the identification and quantification of tree water sources in a riparian forest

Abstract: . We investigated plant water sources of an emblematic refugial population of Fagus sylvatica (L.) in the Ciron river gorges in south-western France using stable water isotopes. It is generally assumed that no isotopic fractionation occurs during root water uptake, so that the isotopic composition of xylem water effectively reflects that of source water. However, this assumption has been called into question by recent studies that found that, at least at some dates during the growing season, plant water did not reflect any mixture of the potential water sources. In this context, highly resolved datasets covering a range of environmental conditions could shed light on possible plant–soil fractionation processes responsible for this phenomenon. In this study, the hydrogen ( δ2H ) and oxygen ( δ18O ) isotope compositions of all potential tree water sources and xylem water were measured fortnightly over an entire growing season. Using a Bayesian isotope mixing model (MixSIAR), we then quantified the relative contribution of water sources for F. sylvatica and Quercus robur (L.) trees. Based on δ18O data alone, both species used a mix of top and deep soil water over the season, with Q. robur using deeper soil water than F. sylvatica. The contribution of stream water appeared to be marginal despite the proximity of the trees to the stream, as already reported for other riparian forests. Xylem water δ18O could always be interpreted as a mixture of deep and shallow soil waters, but the δ2H of xylem water was often more depleted than the considered water sources. We argue that an isotopic fractionation in the unsaturated zone and/or within the plant tissues could underlie this unexpected relatively depleted δ2H of xylem water, as already observed in halophytic and xerophytic species. By means of a sensitivity analysis, we found that the estimation of plant water sources using mixing models was strongly affected by this δ2H depletion. A better understanding of what causes this isotopic separation between xylem and source water is urgently needed.

Estimating irrigation water use over the contiguous United States by combining satellite and reanalysis soil moisture data

Abstract: . Effective agricultural water management requires accurate and timely information on the availability and use of irrigation water. However, most existing information on irrigation water use ( IWU ) lacks the objectivity and spatiotemporal representativeness needed for operational water management and meaningful characterization of land–climate interactions. Although optical remote sensing has been used to map the area affected by irrigation, it does not physically allow for the estimation of the actual amount of irrigation water applied. On the other hand, microwave observations of the moisture content in the top soil layer are directly influenced by agricultural irrigation practices and thus potentially allow for the quantitative estimation of IWU. In this study, we combine surface soil moisture (SM) retrievals from the spaceborne SMAP, AMSR2 and ASCAT microwave sensors with modeled soil moisture from MERRA-2 reanalysis to derive monthly IWU dynamics over the contiguous United States (CONUS) for the period 2013–2016. The methodology is driven by the assumption that the hydrology formulation of the MERRA-2 model does not account for irrigation, while the remotely sensed soil moisture retrievals do contain an irrigation signal. For many CONUS irrigation hot spots, the estimated spatial irrigation patterns show good agreement with a reference data set on irrigated areas. Moreover, in intensively irrigated areas, the temporal dynamics of observed IWU is meaningful with respect to ancillary data on local irrigation practices. State-aggregated mean IWU volumes derived from the combination of SMAP and MERRA-2 soil moisture show a good correlation with statistically reported state-level irrigation water withdrawals (IWW) but systematically underestimate them. We argue that this discrepancy can be mainly attributed to the coarse spatial resolution of the employed satellite soil moisture retrievals, which fails to resolve local irrigation practices. Consequently, higher-resolution soil moisture data are needed to further enhance the accuracy of IWU mapping.

On the choice of calibration metrics for “high-flow” estimation using hydrologic models

Abstract: . Calibration is an essential step for improving the accuracy of simulations generated using hydrologic models. A key modeling decision is selecting the performance metric to be optimized. It has been common to use squared error performance metrics, or normalized variants such as Nash–Sutcliffe efficiency (NSE), based on the idea that their squared-error nature will emphasize the estimates of high flows. However, we conclude that NSE-based model calibrations actually result in poor reproduction of high-flow events, such as the annual peak flows that are used for flood frequency estimation. Using three different types of performance metrics, we calibrate two hydrological models at a daily step, the Variable Infiltration Capacity (VIC) model and the mesoscale Hydrologic Model (mHM), and evaluate their ability to simulate high-flow events for 492 basins throughout the contiguous United States. The metrics investigated are (1) NSE, (2) Kling–Gupta efficiency (KGE) and its variants, and (3) annual peak flow bias (APFB), where the latter is an application-specific metric that focuses on annual peak flows. As expected, the APFB metric produces the best annual peak flow estimates; however, performance on other high-flow-related metrics is poor. In contrast, the use of NSE results in annual peak flow estimates that are more than 20 % worse, primarily due to the tendency of NSE to underestimate observed flow variability. On the other hand, the use of KGE results in annual peak flow estimates that are better than from NSE, owing to improved flow time series metrics (mean and variance), with only a slight degradation in performance with respect to other related metrics, particularly when a non-standard weighting of the components of KGE is used. Stochastically generated ensemble simulations based on model residuals show the ability to improve the high-flow metrics, regardless of the deterministic performances. However, we emphasize that improving the fidelity of streamflow dynamics from deterministically calibrated models is still important, as it may improve high-flow metrics (for the right reasons). Overall, this work highlights the need for a deeper understanding of performance metric behavior and design in relation to the desired goals of model calibration.

Inundation analysis of metro systems with the storm water management model incorporated into a geographical information system: a case study in Shanghai

Abstract: . This study presents an integrated approach to evaluate inundation risks, in which an algorithm is proposed to integrate the storm water management model (SWMM) into a geographical information system (GIS). The proposed algorithm simulates the flood inundation of overland flows and in metro stations for each designed scenario. It involves the following stages: (i) determination of the grid location and spreading coefficient and (ii) an iterative calculation of the spreading process. In addition, an equation is proposed to calculate the inundation around a metro station and to predict the potential inundation risks of the metro system. The proposed method is applied to simulate the inundation risk of the metro system in the urban centre of Shanghai under 50-year, 100-year, and 500-year rainfall intensities. Both inundation extent and depth are obtained and the proposed method is validated with records of historical floods. The results demonstrate that in the case of a 500-year rainfall intensity, the inundated area with a water depth excess of 300 mm covers up to 5.16 km 2 . In addition, four metro stations are inundated to a depth of over 300 mm.

Representation and improved parameterization of reservoir operation in hydrological and land-surface models

Abstract: . Reservoirs significantly affect flow regimes in watershed systems by changing the magnitude and timing of streamflows. Failure to represent these effects limits the performance of hydrological and land-surface models (H-LSMs) in the many highly regulated basins across the globe and limits the applicability of such models to investigate the futures of watershed systems through scenario analysis (e.g., scenarios of climate, land use, or reservoir regulation changes). An adequate representation of reservoirs and their operation in an H-LSM is therefore essential for a realistic representation of the downstream flow regime. In this paper, we present a general parametric reservoir operation model based on piecewise-linear relationships between reservoir storage, inflow, and release to approximate actual reservoir operations. For the identification of the model parameters, we propose two strategies: (a) a “generalized” parameterization that requires a relatively limited amount of data and (b) direct calibration via multi-objective optimization when more data on historical storage and release are available. We use data from 37 reservoir case studies located in several regions across the globe for developing and testing the model. We further build this reservoir operation model into the MESH (Modelisation Environmentale-Surface et Hydrologie) modeling system, which is a large-scale H-LSM. Our results across the case studies show that the proposed reservoir model with both parameter-identification strategies leads to improved simulation accuracy compared with the other widely used approaches for reservoir operation simulation. We further show the significance of enabling MESH with this reservoir model and discuss the interdependent effects of the simulation accuracy of natural processes and that of reservoir operations on the overall model performance. The reservoir operation model is generic and can be integrated into any H-LSM.

Assessing the characteristics and drivers of compound flooding events around the UK coast

Abstract: . In low-lying coastal regions, flooding arises from oceanographic (storm surges plus tides and/or waves), fluvial (increased river discharge), and/or pluvial (direct surface run-off) sources. The adverse consequences of a flood can be disproportionately large when these different sources occur concurrently or in close succession, a phenomenon that is known as “compound flooding”. In this paper, we assess the potential for compound flooding arising from the joint occurrence of high storm surge and high river discharge around the coast of the UK. We hypothesise that there will be spatial variation in compound flood frequency, with some coastal regions experiencing a greater dependency between the two flooding sources than others. We map the dependence between high skew surges and high river discharge, considering 326 river stations linked to 33 tide gauge sites. We find that the joint occurrence of high skew surges and high river discharge occurs more frequently during the study period (15–50 years) at sites on the south-western and western coasts of the UK (between three and six joint events per decade) compared to sites along the eastern coast (between zero and one joint events per decade). Second, we investigate the meteorological conditions that drive compound and non-compound events across the UK. We show, for the first time, that spatial variability in the dependence and number of joint occurrences of high skew surges and high river discharge is driven by meteorological differences in storm characteristics. On the western coast of the UK, the storms that generate high skew surges and high river discharge are typically similar in characteristics and track across the UK on comparable pathways. In contrast, on the eastern coast, the storms that typically generate high skew surges are mostly distinct from the types of storms that tend to generate high river discharge. Third, we briefly examine how the phase and strength of dependence between high skew surge and high river discharge is influenced by the characteristics (i.e. flashiness, size, and elevation gradient) of the corresponding river catchments. We find that high skew surges tend to occur more frequently with high river discharge at catchments with a lower base flow index, smaller catchment area, and steeper elevation gradient. In catchments with a high base flow index, large catchment area, and shallow elevation gradient, the peak river flow tends to occur several days after the high skew surge. The previous lack of consideration of compound flooding means that flood risk has likely been underestimated around UK coasts, particularly along the south-western and western coasts. It is crucial that this be addressed in future assessments of flood risk and flood management approaches.

Multi-site calibration and validation of SWAT with satellite-based evapotranspiration in a data-sparse catchment in southwestern Nigeria

Abstract: . The main objective of this study was to calibrate and validate the eco-hydrological model Soil and Water Assessment Tool (SWAT) with satellite-based actual evapotranspiration (AET) data from the Global Land Evaporation Amsterdam Model (GLEAM_v3.0a) and from the Moderate Resolution Imaging Spectroradiometer Global Evaporation (MOD16) for the Ogun River Basin (20 292 km 2 ) located in southwestern Nigeria. Three potential evapotranspiration (PET) equations (Hargreaves, Priestley–Taylor and Penman–Monteith) were used for the SWAT simulation of AET. The reference simulations were the three AET variables simulated with SWAT before model calibration took place. The sequential uncertainty fitting technique (SUFI-2) was used for the SWAT model sensitivity analysis, calibration, validation and uncertainty analysis. The GLEAM_v3.0a and MOD16 products were subsequently used to calibrate the three SWAT-simulated AET variables, thereby obtaining six calibrations–validations at a monthly timescale. The model performance for the three SWAT model runs was evaluated for each of the 53 subbasins against the GLEAM_v3.0a and MOD16 products, which enabled the best model run with the highest-performing satellite-based AET product to be chosen. A verification of the simulated AET variable was carried out by (i) comparing the simulated AET of the calibrated model to GLEAM_v3.0b AET, which is a product that has different forcing data than the version of GLEAM used for the calibration, and (ii) assessing the long-term average annual and average monthly water balances at the outlet of the watershed. Overall, the SWAT model, composed of the Hargreaves PET equation and calibrated using the GLEAM_v3.0a data (GS1), performed well for the simulation of AET and provided a good level of confidence for using the SWAT model as a decision support tool. The 95 % uncertainty of the SWAT-simulated variable bracketed most of the satellite-based AET data in each subbasin. A validation of the simulated soil moisture dynamics for GS1 was carried out using satellite-retrieved soil moisture data, which revealed good agreement. The SWAT model (GS1) also captured the seasonal variability of the water balance components at the outlet of the watershed. This study demonstrated the potential to use remotely sensed evapotranspiration data for hydrological model calibration and validation in a sparsely gauged large river basin with reasonable accuracy. The novelty of the study is the use of these freely available satellite-derived AET datasets to effectively calibrate and validate an eco-hydrological model for a data-scarce catchment.

Using paired catchments to quantify the human influence on hydrological droughts

Abstract: . Quantifying the influence of human activities, such as reservoir building, water abstraction, and land use change, on hydrology is crucial for sustainable future water management, especially during drought. Model-based methods are very time-consuming to set up and require a good understanding of human processes and time series of water abstraction, land use change, and water infrastructure and management, which often are not available. Therefore, observation-based methods are being developed that give an indication of the direction and magnitude of the human influence on hydrological drought based on limited data. We suggest adding to those methods a “paired-catchment” approach, based on the classic hydrology approach that was developed in the 1920s for assessing the impact of land cover treatment on water quantity and quality. When applying the paired-catchment approach to long-term pre-existing human influences trying to detect an influence on extreme events such as droughts, a good catchment selection is crucial. The disturbed catchment needs to be paired with a catchment that is similar in all aspects except for the human activity under study, in that way isolating the effect of that specific activity. In this paper, we present a framework for selecting suitable paired catchments for the study of the human influence on hydrological drought. Essential elements in this framework are the availability of qualitative information on the human activity under study (type, timing, and magnitude), and the similarity of climate, geology, and other human influences between the catchments. We show the application of the framework on two contrasting case studies, one impacted by groundwater abstraction and one with a water transfer from another region. Applying the paired-catchment approach showed how the groundwater abstraction aggravated streamflow drought by more than 200 % for some metrics (total drought duration and total drought deficit) and the water transfer alleviated droughts with 25 % to 80 %, dependent on the metric. Benefits of the paired-catchment approach are that climate variability between pre- and post-disturbance periods does not have to be considered as the same time periods are used for analysis, and that it avoids assumptions considered when partly or fully relying on simulation modelling. Limitations of the approach are that finding a suitable catchment pair can be very challenging, often no pre-disturbance records are available to establish the natural difference between the catchments, and long time series of hydrological data are needed to robustly detect the effect of the human activities on hydrological drought. We suggest that the approach can be used for a first estimate of the human influence on hydrological drought, to steer campaigns to collect more data, and to complement and improve other existing methods (e.g. model-based or large-sample approaches).

Assessing the performance of global hydrological models for capturing peak river flows in the Amazon basin

Abstract: . Extreme flooding impacts millions of people that live within the Amazon floodplain. Global hydrological models (GHMs) are frequently used to assess and inform the management of flood risk, but knowledge on the skill of available models is required to inform their use and development. This paper presents an intercomparison of eight different GHMs freely available from collaborators of the Global Flood Partnership (GFP) for simulating floods in the Amazon basin. To gain insight into the strengths and shortcomings of each model, we assess their ability to reproduce daily and annual peak river flows against gauged observations at 75 hydrological stations over a 19-year period (1997–2015). As well as highlighting regional variability in the accuracy of simulated streamflow, these results indicate that (a) the meteorological input is the dominant control on the accuracy of both daily and annual maximum river flows, and (b) groundwater and routing calibration of Lisflood based on daily river flows has no impact on the ability to simulate flood peaks for the chosen river basin. These findings have important relevance for applications of large-scale hydrological models, including analysis of the impact of climate variability, assessment of the influence of long-term changes such as land-use and anthropogenic climate change, the assessment of flood likelihood, and for flood forecasting systems.

Effects of univariate and multivariate bias correction on hydrological impact projections in alpine catchments

Abstract: . Alpine catchments show a high sensitivity to climate variation as they include the elevation range of the snow line. Therefore, the correct representation of climate variables and their interdependence is crucial when describing or predicting hydrological processes. When using climate model simulations in hydrological impact studies, forcing meteorological data are usually downscaled and bias corrected, most often by univariate approaches such as quantile mapping of individual variables, neglecting the relationships that exist between climate variables. In this study we test the hypothesis that the explicit consideration of the relation between air temperature and precipitation will affect hydrological impact modelling in a snow-dominated mountain environment. Glacio-hydrological simulations were performed for two partly glacierized alpine catchments using a recently developed multivariate bias correction method to post-process EURO-CORDEX regional climate model outputs between 1976 and 2099. These simulations were compared to those obtained by using the common univariate quantile mapping for bias correction. As both methods correct each climate variable's distribution in the same way, the marginal distributions of the individual variables show no differences. Yet, regarding the interdependence of precipitation and air temperature, clear differences are notable in the studied catchments. Simultaneous correction based on the multivariate approach led to more precipitation below air temperatures of 0 ∘ C and therefore more simulated snowfall than with the data of the univariate approach. This difference translated to considerable consequences for the hydrological responses of the catchments. The multivariate bias-correction-forced simulations showed distinctly different results for projected snow cover characteristics, snowmelt-driven streamflow components, and expected glacier disappearance dates. In all aspects – the fraction of precipitation above and below 0 ∘ C, the simulated snow water equivalents, glacier volumes, and the streamflow regime – simulations resulting from the multivariate-corrected data corresponded better with reference data than the results of univariate bias correction. Differences in simulated total streamflow due to the different bias correction approaches may be considered negligible given the generally large spread of the projections, but systematic differences in the seasonally delayed streamflow components from snowmelt in particular will matter from a planning perspective. While this study does not allow conclusive evidence that multivariate bias correction approaches are generally preferable, it clearly demonstrates that incorporating or ignoring inter-variable relationships between air temperature and precipitation data can impact the conclusions drawn in hydrological climate change impact studies in snow-dominated environments.

Spatiotemporal changes in aridity of Pakistan during 1901–2016

Abstract: . The changing characteristics of aridity over a larger spatiotemporal scale have gained interest in recent years due to climate change. The long-term (1901–2016) changes in spatiotemporal patterns of annual and seasonal aridity during two major crop growing seasons of Pakistan, Kharif and Rabi, are evaluated in this study using gridded precipitation and potential evapotranspiration (PET) data. The UNESCO aridity index was used to estimate aridity at each grid point for all the years between 1901 and 2016. The temporal changes in aridity and its associations with precipitation and PET are evaluated by implementing a moving window of 50 years of data with an 11-year interval. The modified Mann–Kendall (MMK) trend test is applied to estimate unidirectional change by eliminating the effect of natural variability of climate, and Pettitt's test is used to detect year of change in aridity. The results revealed that the climate over 60 % of Pakistan (mainly in southern parts) is arid. The spatial patterns of aridity trends show a strong influence of the changes in precipitation on the aridity trend. The increasing trend in aridity (drier) is noticed in the southwest, where precipitation is low during Kharif, while there is a decreasing trend (wetter) in the Rabi season in the region which receives high precipitation due to western disturbances. The annual and Kharif aridity is found to decrease (wetter) at a rate of 0.0001 to 0.0002 per year in the northeast, while Kharif and Rabi aridity are found to increase (drier) at some locations in the south at a rate of −0.0019 to −0.0001 per year. The spatial patterns of aridity changes show a shift from arid to the semi-arid (wetter) climate in annual and Kharif over a large area while showing a shift from arid to hyper-arid (drier) region during Rabi in a small area. Most of the significant changes in precipitation and aridity are observed in the years between 1971 and 1980. Overall, aridity is found to increase (drier) in 0.52 %, 4.44 % and 0.52 % of the area and decrease (wetter) in 11.75 %, 7.57 % and 9.66 % of the area for annual, Rabi and Kharif seasons respectively during 1967–2016 relative to 1901–1950.

Benchmarking the predictive capability of hydrological models for river flow and flood peak predictions across over 1000 catchments in Great Britain

Abstract: . Benchmarking model performance across large samples of catchments is useful to guide model selection and future model development. Given uncertainties in the observational data we use to drive and evaluate hydrological models, and uncertainties in the structure and parameterisation of models we use to produce hydrological simulations and predictions, it is essential that model evaluation is undertaken within an uncertainty analysis framework. Here, we benchmark the capability of several lumped hydrological models across Great Britain by focusing on daily flow and peak flow simulation. Four hydrological model structures from the Framework for Understanding Structural Errors (FUSE) were applied to over 1000 catchments in England, Wales and Scotland. Model performance was then evaluated using standard performance metrics for daily flows and novel performance metrics for peak flows considering parameter uncertainty. Our results show that lumped hydrological models were able to produce adequate simulations across most of Great Britain, with each model producing simulations exceeding a 0.5 Nash–Sutcliffe efficiency for at least 80 % of catchments. All four models showed a similar spatial pattern of performance, producing better simulations in the wetter catchments to the west and poor model performance in central Scotland and south-eastern England. Poor model performance was often linked to the catchment water balance, with models unable to capture the catchment hydrology where the water balance did not close. Overall, performance was similar between model structures, but different models performed better for different catchment characteristics and metrics, as well as for assessing daily or peak flows, leading to the ensemble of model structures outperforming any single structure, thus demonstrating the value of using multi-model structures across a large sample of different catchment behaviours. This research evaluates what conceptual lumped models can achieve as a performance benchmark and provides interesting insights into where and why these simple models may fail. The large number of river catchments included in this study makes it an appropriate benchmark for any future developments of a national model of Great Britain.

Quantifying new water fractions and transit time distributions using ensemble hydrograph separation: theory and benchmark tests

Abstract: Decades of hydrograph separation studies have estimated the proportions of recent precipitation in streamflow using end-member mixing of chemical or isotopic tracers Here I propose an ensemble approach to hydrograph separation that uses regressions between tracer fluctuations in precipitation and discharge to estimate the average fraction of new water (eg, same-day or same-week precipitation) in streamflow across an ensemble of time steps The points comprising this ensemble can be selected to isolate conditions of particular interest, making it possible to study how the new water fraction varies as a function of catchment and storm characteristics Even when new water fractions are highly variable over time, one can show mathematically (and confirm with benchmark tests) that ensemble hydrograph separation will accurately estimate their average Because ensemble hydrograph separation is based on correlations between tracer fluctuations rather than on tracer mass balances, it does not require that the end-member signatures are constant over time, or that all the end-members are sampled or even known, and it is relatively unaffected by evaporative isotopic fractionation Ensemble hydrograph separation can also be extended to a multiple regression that estimates the average (or “marginal”) transit time distribution (TTD) directly from observational data This approach can estimate both “backward” transit time distributions (the fraction of streamflow that originated as rainfall at different lag times) and “forward” transit time distributions (the fraction of rainfall that will become future streamflow at different lag times), with and without volume-weighting, up to a user-determined maximum time lag The approach makes no assumption about the shapes of the transit time distributions, nor does it assume that they are time-invariant, and it does not require continuous time series of tracer measurements Benchmark tests with a nonlinear, nonstationary catchment model confirm that ensemble hydrograph separation reliably quantifies both new water fractions and transit time distributions across widely varying catchment behaviors, using either daily or weekly tracer concentrations as input Numerical experiments with the benchmark model also illustrate how ensemble hydrograph separation can be used to quantify the effects of rainfall intensity, flow regime, and antecedent wetness on new water fractions and transit time distributions

Future projections of temperature and mixing regime of European temperate lakes

Abstract: . The physical response of lakes to climate warming is regionally variable and highly dependent on individual lake characteristics, making generalizations about their development difficult. To qualify the role of individual lake characteristics in their response to regionally homogeneous warming, we simulated temperature, ice cover, and mixing in four intensively studied German lakes of varying morphology and mixing regime with a one-dimensional lake model. We forced the model with an ensemble of 12 climate projections (RCP4.5) up to 2100. The lakes were projected to warm at 0.10–0.11 ∘ C decade −1 , which is 75 %–90 % of the projected air temperature trend. In simulations, surface temperatures increased strongly in winter and spring, but little or not at all in summer and autumn. Mean bottom temperatures were projected to increase in all lakes, with steeper trends in winter and in shallower lakes. Modelled ice thaw and summer stratification advanced by 1.5–2.2 and 1.4–1.8 days decade −1 respectively, whereas autumn turnover and winter freeze timing was less sensitive. The projected summer mixed-layer depth was unaffected by warming but sensitive to changes in water transparency. By mid-century, the frequency of ice and stratification-free winters was projected to increase by about 20 %, making ice cover rare and shifting the two deeper dimictic lakes to a predominantly monomictic regime. The polymictic lake was unlikely to become dimictic by the end of the century. A sensitivity analysis predicted that decreasing transparency would dampen the effect of warming on mean temperature but amplify its effect on stratification. However, this interaction was only predicted to occur in clear lakes, and not in the study lakes at their historical transparency. Not only lake morphology, but also mixing regime determines how heat is stored and ultimately how lakes respond to climate warming. Seasonal differences in climate warming rates are thus important and require more attention.

Sensitivity of hydrological models to temporal and spatial resolutions of rainfall data

Abstract: . Rainfall is the most important input for rainfall–runoff models. It is usually measured at specific sites on a daily or sub-daily timescale and requires interpolation for further application. This study aims to evaluate whether a higher temporal and spatial resolution of rainfall can lead to improved model performance. Four different gridded hourly and daily rainfall datasets with a spatial resolution of 1 km × 1 km for the state of Baden-Wurttemberg in Germany were constructed using a combination of data from a dense network of daily rainfall stations and a less dense network of sub-daily stations. Lumped and spatially distributed HBV models were used to investigate the sensitivity of model performance to the spatial resolution of rainfall. The four different rainfall datasets were used to drive both lumped and distributed HBV models to simulate daily discharges in four catchments. The main findings include that (1) a higher temporal resolution of rainfall improves the model performance if the station density is high; (2) a combination of observed high temporal resolution observations with disaggregated daily rainfall leads to further improvement in the tested models; and (3) for the present research, the increase in spatial resolution improves the performance of the model insubstantially or only marginally in most of the study catchments.

High-resolution regional climate modeling and projection over western Canada using a weather research forecasting model with a pseudo-global warming approach

Abstract: . Climate change poses great risks to western Canada's ecosystem and socioeconomical development. To assess these hydroclimatic risks under high-end emission scenario RCP8.5, this study used the Weather Research Forecasting (WRF) model at a convection-permitting (CP) 4 km resolution to dynamically downscale the mean projection of a 19-member CMIP5 ensemble by the end of the 21st century. The CP simulations include a retrospective simulation (CTL, 2000–2015) for verification forced by ERA-Interim and a pseudo-global warming (PGW) for climate change projection forced with climate change forcing (2071–2100 to 1976–2005) from CMIP5 ensemble added on ERA-Interim. The retrospective WRF-CTL's surface air temperature simulation was evaluated against Canadian daily analysis ANUSPLIN, showing good agreements in the geographical distribution with cold biases east of the Canadian Rockies, especially in spring. WRF-CTL captures the main pattern of observed precipitation distribution from CaPA and ANUSPLIN but shows a wet bias near the British Columbia coast in winter and over the immediate region on the lee side of the Canadian Rockies. The WRF-PGW simulation shows significant warming relative to CTL, especially over the polar region in the northeast during the cold season, and in daily minimum temperature. Precipitation changes in PGW over CTL vary with the seasons: in spring and late autumn precipitation increases in most areas, whereas in summer in the Saskatchewan River basin and southern Canadian Prairies, the precipitation change is negligible or decreased slightly. With almost no increase in precipitation and much more evapotranspiration in the future, the water availability during the growing season will be challenging for the Canadian Prairies. The WRF-PGW projected warming is less than that by the CMIP5 ensemble in all seasons. The CMIP5 ensemble projects a 10 %–20 % decrease in summer precipitation over the Canadian Prairies and generally agrees with WRF-PGW except for regions with significant terrain. This difference may be due to the much higher resolution of WRF being able to more faithfully represent small-scale summer convection and orographic lifting due to steep terrain. WRF-PGW shows an increase in high-intensity precipitation events and shifts the distribution of precipitation events toward more extremely intensive events in all seasons. Due to this shift in precipitation intensity to the higher end in the PGW simulation, the seemingly moderate increase in the total amount of precipitation in summer east of the Canadian Rockies may underestimate the increase in flooding risk and water shortage for agriculture. The change in the probability distribution of precipitation intensity also calls for innovative bias-correction methods to be developed for the application of the dataset when bias correction is required. High-quality meteorological observation over the region is needed for both forcing high-resolution climate simulation and conducting verification. The high-resolution downscaled climate simulations provide abundant opportunities both for investigating local-scale atmospheric dynamics and for studying climate impacts on hydrology, agriculture, and ecosystems.

Consistency of satellite-based precipitation products in space and over time compared with gauge observations and snow- hydrological modelling in the Lake Titicaca region

Abstract: . This paper proposes a protocol to assess the space–time consistency of 12 satellite-based precipitation products (SPPs) according to various indicators, including (i) direct comparison of SPPs with 72 precipitation gauges; (ii) sensitivity of streamflow modelling to SPPs at the outlet of four basins; and (iii) the sensitivity of distributed snow models to SPPs using a MODIS snow product as reference in an unmonitored mountainous area. The protocol was applied successively to four different time windows (2000–2004, 2004–2008, 2008–2012 and 2000–2012) to account for the space–time variability of the SPPs and to a large dataset composed of 12 SPPs (CMORPH–RAW v.1, CMORPH–CRT v.1, CMORPH–BLD v.1, CHIRP v.2, CHIRPS v.2, GSMaP v.6, MSWEP v.2.1, PERSIANN, PERSIANN–CDR, TMPA–RT v.7, TMPA–Adj v.7 and SM2Rain–CCI v.2), an unprecedented comparison. The aim of using different space scales and timescales and indicators was to evaluate whether the efficiency of SPPs varies with the method of assessment, time window and location. Results revealed very high discrepancies between SPPs. Compared to precipitation gauge observations, some SPPs (CMORPH–RAW v.1, CMORPH–CRT v.1, GSMaP v.6, PERSIANN, and TMPA–RT v.7) are unable to estimate regional precipitation, whereas the others (CHIRP v.2, CHIRPS v.2, CMORPH–BLD v.1, MSWEP v.2.1, PERSIANN–CDR, and TMPA–Adj v.7) produce a realistic representation despite recurrent spatial limitation over regions with contrasted emissivity, temperature and orography. In 9 out of 10 of the cases studied, streamflow was more realistically simulated when SPPs were used as forcing precipitation data rather than precipitation derived from the available precipitation gauge networks, whereas the SPP's ability to reproduce the duration of MODIS-based snow cover resulted in poorer simulations than simulation using available precipitation gauges. Interestingly, the potential of the SPPs varied significantly when they were used to reproduce gauge precipitation estimates, streamflow observations or snow cover duration and depending on the time window considered. SPPs thus produce space–time errors that cannot be assessed when a single indicator and/or time window is used, underlining the importance of carefully considering their space–time consistency before using them for hydro-climatic studies. Among all the SPPs assessed, MSWEP v.2.1 showed the highest space–time accuracy and consistency in reproducing gauge precipitation estimates, streamflow and snow cover duration.

Expansion and contraction of the flowing stream network alter hillslope flowpath lengths and the shape of the travel time distribution

Abstract: . Flowing stream networks dynamically extend and retract, both seasonally and in response to precipitation events. These network dynamics can dramatically alter the drainage density and thus the length of subsurface flow pathways to flowing streams. We mapped flowing stream networks in a small Swiss headwater catchment during different wetness conditions and estimated their effects on the distribution of travel times to the catchment outlet. For each point in the catchment, we determined the subsurface transport distance to the flowing stream based on the surface topography and determined the surface transport distance along the flowing stream to the outlet. We combined the distributions of these travel distances with assumed surface and subsurface flow velocities to estimate the distribution of travel times to the outlet. These calculations show that the extension and retraction of the stream network can substantially change the mean travel time and the shape of the travel time distribution. During wet conditions with a fully extended flowing stream network, the travel time distribution was strongly skewed to short travel times, but as the network retracted during dry conditions, the distribution of the travel times became more uniform. Stream network dynamics are widely ignored in catchment models, but our results show that they need to be taken into account when modeling solute transport and interpreting travel time distributions.

Storage dynamics, hydrological connectivity and flux ages in a karst catchment: conceptual modelling using stable isotopes

Abstract: . We developed a new tracer-aided hydrological model that disaggregates cockpit karst terrain into the two dominant landscape units of hillslopes and depressions (with fast and slow flow systems). The new model was calibrated by using high temporal resolution hydrometric and isotope data in the outflow of Chenqi catchment in Guizhou Province of south-western China. The model could track hourly water and isotope fluxes through each landscape unit and estimate the associated storage and water age dynamics. From the model results we inferred that the fast flow reservoir in the depression had the smallest water storage and the slow flow reservoir the largest, with the hillslope intermediate. The estimated mean ages of water draining the hillslope unit, and the fast and slow flow reservoirs during the study period, were 137, 326 and 493 days, respectively. Distinct seasonal variability in hydroclimatic conditions and associated water storage dynamics (captured by the model) were the main drivers of non-stationary hydrological connectivity between the hillslope and depression. During the dry season, slow flow in the depression contributes the largest proportion (78.4 %) of flow to the underground stream draining the catchment, resulting in weak hydrological connectivity between the hillslope and depression. During the wet period, with the resulting rapid increase in storage, the hillslope unit contributes the largest proportion (57.5 %) of flow to the underground stream due to the strong hydrological connectivity between the hillslope and depression. Meanwhile, the tracer-aided model can be used to identify the sources of uncertainty in the model results. Our analysis showed that the model uncertainty of the hydrological variables in the different units relies on their connectivity with the outlet when the calibration target uses only the outlet information. The model uncertainty was much lower for the “newer” water from the fast flow system in the depression and flow from the hillslope unit during the wet season and higher for “older” water from the slow flow system in the depression. This suggests that to constrain model parameters further, increased high-resolution hydrometric and tracer data on the internal dynamics of systems (e.g. groundwater responses during low flow periods) could be used in calibration.

Using R in hydrology: a review of recent developments and future directions

Abstract: . The open-source programming language R has gained a central place in the hydrological sciences over the last decade, driven by the availability of diverse hydro-meteorological data archives and the development of open-source computational tools. The growth of R's usage in hydrology is reflected in the number of newly published hydrological packages, the strengthening of online user communities, and the popularity of training courses and events. In this paper, we explore the benefits and advantages of R's usage in hydrology, such as the democratization of data science and numerical literacy, the enhancement of reproducible research and open science, the access to statistical tools, the ease of connecting R to and from other languages, and the support provided by a growing community. This paper provides an overview of a typical hydrological workflow based on reproducible principles and packages for retrieval of hydro-meteorological data, spatial analysis, hydrological modelling, statistics, and the design of static and dynamic visualizations and documents. We discuss some of the challenges that arise when using R in hydrology and useful tools to overcome them, including the use of hydrological libraries, documentation, and vignettes (long-form guides that illustrate how to use packages); the role of integrated development environments (IDEs); and the challenges of big data and parallel computing in hydrology. Lastly, this paper provides a roadmap for R's future within hydrology, with R packages as a driver of progress in the hydrological sciences, application programming interfaces (APIs) providing new avenues for data acquisition and provision, enhanced teaching of hydrology in R, and the continued growth of the community via short courses and events.

Scalable flood level trend monitoring with surveillance cameras using a deep convolutional neural network

Abstract: . In many countries, urban flooding due to local, intense rainfall is expected to become more frequent because of climate change and urbanization. Cities trying to adapt to this growing risk are challenged by a chronic lack of surface flooding data that are needed for flood risk assessment and planning. In this work, we propose a new approach that exploits existing surveillance camera systems to provide qualitative flood level trend information at scale. The approach uses a deep convolutional neural network (DCNN) to detect floodwater in surveillance footage and a novel qualitative flood index (namely, the static observer flooding index – SOFI) as a proxy for water level fluctuations visible from a surveillance camera's viewpoint. To demonstrate the approach, we trained the DCNN on 1218 flooding images collected from the Internet and applied it to six surveillance videos representing different flooding and lighting conditions. The SOFI signal obtained from the videos had a 75 % correlation to the actual water level fluctuation on average. By retraining the DCNN with a few frames from a given video, the correlation is increased to 85 % on average. The results confirm that the approach is versatile, with the potential to be applied to a variety of surveillance camera models and flooding situations without the need for on-site camera calibration. Thanks to this flexibility, this approach could be a cheap and highly scalable alternative to conventional sensing methods.

Mapping soil hydraulic properties using random-forest-based pedotransfer functions and geostatistics

Abstract: . Spatial 3-D information on soil hydraulic properties for areas larger than plot scale is usually derived using indirect methods such as pedotransfer functions (PTFs) due to the lack of measured information on them. PTFs describe the relationship between the desired soil hydraulic parameter and easily available soil properties based on a soil hydraulic reference dataset. Soil hydraulic properties of a catchment or region can be calculated by applying PTFs on available soil maps. Our aim was to analyse the performance of (i) indirect (using PTFs) and (ii) direct (geostatistical) mapping methods to derive 3-D soil hydraulic properties. The study was performed on the Balaton catchment area in Hungary, where density of measured soil hydraulic data fulfils the requirements of geostatistical methods. Maps of saturated water content (0 cm matric potential), field capacity ( −330 cm matric potential) and wilting point ( −15 000 cm matric potential) for 0–30, 30–60 and 60–90 cm soil depth were prepared. PTFs were derived using the random forest method on the whole Hungarian soil hydraulic dataset, which includes soil chemical, physical, taxonomical and hydraulic properties of some 12 000 samples complemented with information on topography, climate, parent material, vegetation and land use. As a direct and thus geostatistical method, random forest combined with kriging (RFK) was applied to 359 soil profiles located in the Balaton catchment area. There were no significant differences between the direct and indirect methods in six out of nine maps having root-mean-square-error values between 0.052 and 0.074 cm 3 cm −3 , which is in accordance with the internationally accepted performance of hydraulic PTFs. The PTF-based mapping method performed significantly better than the RFK for the saturated water content at 30–60 and 60–90 cm soil depth; in the case of wilting point the RFK outperformed the PTFs at 60–90 cm depth. Differences between the PTF-based and RFK mapped values are less than 0.025 cm 3 cm −3 for 65 %–86 % of the catchment. In RFK, the uncertainty of input environmental covariate layers is less influential on the mapped values, which is preferable. In the PTF-based method the uncertainty of mapping soil hydraulic properties is less computationally intensive. Detailed comparisons of maps derived from the PTF-based method and the RFK are presented in this paper.

The recent developments in cloud removal approaches of MODIS snow cover product

Abstract: . The snow cover products of optical remote sensing systems play an important role in research into global climate change, the hydrological cycle, and the energy balance. Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover products are the most popular datasets used in the community. However, for MODIS, cloud cover results in spatial and temporal discontinuity for long-term snow monitoring. In the last few decades, a large number of cloud removal methods for MODIS snow cover products have been proposed. In this paper, our goal is to make a comprehensive summarization of the existing algorithms for generating cloud-free MODIS snow cover products and to expose the development trends. The methods of generating cloud-free MODIS snow cover products are classified into spatial methods, temporal methods, spatio-temporal methods, and multi-source fusion methods. The spatial methods and temporal methods remove the cloud cover of the snow product based on the spatial patterns and temporal changing correlation of the snowpack, respectively. The spatio-temporal methods utilize the spatial and temporal features of snow jointly. The multi-source fusion methods utilize the complementary information among different sources among optical observations, microwave observations, and station observations.

Land–atmosphere interactions in the tropics – a review

Abstract: . The continental tropics play a leading role in the terrestrial energy, water, and carbon cycles. Land–atmosphere interactions are integral in the regulation of these fluxes across multiple spatial and temporal scales over tropical continents. We review here some of the important characteristics of tropical continental climates and how land–atmosphere interactions regulate them. Along with a wide range of climates, the tropics manifest a diverse array of land–atmosphere interactions. Broadly speaking, in tropical rainforest climates, light and energy are typically more limiting than precipitation and water supply for photosynthesis and evapotranspiration (ET), whereas in savanna and semi-arid climates, water is the critical regulator of surface fluxes and land–atmosphere interactions. We discuss the impact of the land surface, how it affects shallow and deep clouds, and how these clouds in turn can feed back to the surface by modulating surface radiation and precipitation. Some results from recent research suggest that shallow clouds may be especially critical to land–atmosphere interactions. On the other hand, the impact of land-surface conditions on deep convection appears to occur over larger, nonlocal scales and may be a more relevant land–atmosphere feedback mechanism in transitional dry-to-wet regions and climate regimes.