Hydrological Signatures Research Articles

Classifying annual daily hydrographs in Western North America using t‐distributed stochastic neighbour embedding

AbstractFlow regimes are critical for determining physical and biological processes in rivers and their classification and regionalization traditionally seeks to link patterns of flow to physiographic, climate and other information. There are many approaches to, and rationales for, catchment classification, with those focused on streamflow often seeking to relate a particular response characteristic to a physical property or climatic driver. Rationales include such topics as prediction in ungauged basins (PUB), and providing guidance for model selection in poorly understood hydrological systems. The annual daily hydrograph (ADH) is a first‐order easily visualized integrated expression of catchment function, and over many years the average ADH is a distinct hydrological signature that differentiate catchments from each other. In this study, we use t‐SNE, a state‐of‐the‐art technique of dimensionality reduction, to classify 17 110 ADHs for 304 reference catchments in mountainous Western North America. t‐SNE is chosen over other conventional methods of dimensionality reduction (e.g., PCA) as it presents greater separability of ADHs, which are projected on a 2D map where the similarities are evaluated according to their map distance. We then utilize a Deep Learning encoder to upgrade the non‐parametric t‐SNE to a parametric approach, enhancing its capability to address ‘unseen’ samples. Results showed that t‐SNE successfully clustered ADHs of similar flow regimes on the 2D map and allowed more accurate classification with KNN. In addition, many compact clusters on the 2D map in the coastal Pacific Northwest suggest information redundancy in the local streamflow network. The t‐SNE map provides an intuitive way to visualize the similarity of high‐dimensional data of ADHs, groups catchments with like characteristics, and avoids the reliance on subjective hydrometric indicators.

A multi‐objective approach to select hydrological models and constrain structural uncertainties for climate impact assessments

AbstractThe assessment of climate change impacts on water resources and flood risk is typically underpinned by hydrological models calibrated and selected based on observed streamflow records. Yet, changes in climate are rarely accounted for when selecting hydrological models, which compromises their ability to robustly represent future changes in catchment hydrology. In this paper, we test a simple framework for selecting an ensemble of calibrated hydrological model structures in catchments where contrasting climatic conditions have been observed. We start by considering 78 model structures produced with the FUSE modular modelling framework and rely on a Pareto scheme to select model structures maximizing model efficiency in both wet and dry periods. The application of this approach in three case study basins in Peru enables the identification of structures with good robustness, but also good performance according to hydrological signatures not used for model selection. We also highlight that some model structures that perform well according to traditional efficiency metrics have low performance in contrasting climates or suspicious internal states and fluxes. Importantly, the model selection approach followed here helps to reduce the spread in precipitation elasticities and temperature sensitivities, providing a clearer picture of future hydrological changes. Overall, this work demonstrates the potential of using contrasting climatic conditions in a multi‐objective framework to produce robust and credible simulations, and to constrain structural uncertainties in hydrological projections.

Explore Spatio‐Temporal Learning of Large Sample Hydrology Using Graph Neural Networks

AbstractStreamflow forecasting over gauged and ungauged basins play a vital role in water resources planning, especially under the changing climate. Increased availability of large sample hydrology data sets, together with recent advances in deep learning techniques, has presented new opportunities to explore temporal and spatial patterns in hydrological signatures for improving streamflow forecasting. The purpose of this study is to adapt and benchmark several state‐of‐the‐art graph neural network (GNN) architectures, including ChebNet, Graph Convolutional Network (GCN), and GraphWaveNet, for end‐to‐end graph learning. We explicitly represent river basins as nodes in a graph, learn the spatiotemporal nodal dependencies, and then use the learned relations to predict streamflow simultaneously across all nodes in the graph. The efficacy of the developed GNN models is investigated using the Catchment Attributes and MEteorology for Large‐sample Studies (CAMELS) data set under two settings, fixed graph topology (transductive learning), and variable graph topology (inductive learning), with the latter applicable to prediction in ungauged basins (PUB). Results indicate that GNNs are generally robust and computationally efficient, achieving similar or better performance than a baseline model trained using the long short‐term memory (LSTM) network. Further analyses are conducted to interpret the graph learning process at the edge and node levels and to investigate the effect of different model configurations. We conclude that graph learning constitutes a viable machine learning‐based method for aggregating spatiotemporal information from a multitude of sources for streamflow forecasting

Assessment of water availability vulnerability in the Cerrado

Many basins in the Cerrado biome already face increases in water disputes caused by agricultural expansion without long-term planning and monitoring of natural resources. The amount and timing of water availability are crucial for the socioeconomic development of agricultural activities and climate change resilience enhancement in the Cerrado. This study relied on a dataset developed for the Cerrado to characterize its climate and current water availability status. Different climate indices and hydrological signatures were adopted to describe the energy and water budget, climate seasonality, flow magnitude, flow dynamics, and average duration of low-flow events. Risk analysis was carried out to classify the monthly risk severity and identify periods and regions under a higher pressure regarding water availability in the Cerrado. Overall, the water availability is higher in the western and southern regions and lower in the eastern and northern regions, which experience a transition to a humid and semiarid climate, respectively. The water resources in the Cerrado occur under the highest pressure in September and October. We also identified regions with the potential to increase sustainable water use. These regions exhibit a low ratio between the flow equal to or exceeded 95% of the time and the mean flow (Q95/Q), which indicates that sustainable water use may be improved with techniques related to flow regularization and rainwater/runoff harvesting.

A hydro-climatological outlook on the long-term availability of water resources in Cauvery river basin

The past few decades have witnessed massive shifts in land-use patterns, land management practices and water demand in the river basins of peninsular India. Changes in hydrologic regimes of different components of the water cycle pose immense challenge to water security at the catchment scale. This paper provides a comprehensive overview and a quantitative assessment of the long-term availability of water resources in the Cauvery river basin, a major river basin in peninsular India. The Cauvery has been a contentious river for decades regarding water sharing among its riparian states of Karnataka and Tamil Nadu. Moreover, the river basin is characterized by extensive regional variability in both surface water and groundwater distributions and has faced acute water management challenges at multiple space and time scales. A descriptive overview of the major water security challenges faced by the basin is presented along with a detailed synthesis of hydrological modelling approaches and statistical methods to assess the basin’s long-term water availability. The Variable Infiltration Capacity (VIC) model is implemented to evaluate the long-term water balance components at the catchment scale for the period 1951–2016. Decadal Land Use Land Cover (LULC) in the basin for the years 1985, 1995 and 2005 are analysed. The statistical trends in hydrometeorology and extreme climatology of the basin are evaluated at seasonal and annual scales. Seasonal flow behaviour and influence of climate and human activities on streamflow are also assessed. A drought duration analysis is performed to infer about the implications of spatial distribution of water availability across the basin. The results show significant trends in the water balance components and the basin’s hydro-climatology. The LULC analysis shows notable changes in land use patterns due to human activities. Alterations in the predictability and temporal variability of streamflow are observed after the construction of dams which may affect the downstream water availability. This research highlights the importance of comprehensive synthesis of hydrological modelling approaches with hydrological signatures to understand the dynamics of hydrological processes at catchment scale with an objective of providing insights for effective planning and allocation of water resources and judicious management of water demands.

Regionalization for Ungauged Catchments — Lessons Learned From a Comparative Large‐Sample Study

AbstractModel parameter values for ungauged catchments can be regionalized from hydrologically similar gauged catchments. Achieving reliable and robust predictions in ungauged catchments by regionalization, however, is still a major challenge. Here, we conduct a comparative assessment of 19 regionalization approaches based on previously published literature to contribute new insights into their performance in different geographic regions. The approaches use geographical information, physical catchment attributes, hydrological signatures, or a combination thereof to select donor catchments and to subsequently transfer their entire parameter sets to the ungauged receiver catchment. Each regionalization approach was tested in a leave‐one‐out cross‐validation with a bucket‐type catchment model (the HBV model) using 671 gauged catchments in the United States with a diverse hydroclimatology. We then evaluated regionalization performance for several hydrograph aspects, compared it against calibration and regionalization benchmarks, and linked it to catchment descriptors. The results of this large‐sample regionalization study can be summarized in three major lessons: (a) Catchments can benefit from a well‐chosen regionalization approach independent of their geographic region and independent of how well they can be modeled or regionalized at best. (b) Almost perfect donors exist for most catchments and an excellent relative model performance can be reached for most catchments with current regionalization approaches. This implies that there is considerable potential for improvement in the prediction in ungauged catchments. (c) The ranking of regionalization approaches depends on how the predicted hydrographs are evaluated. These findings indicate that a multi‐criteria evaluation is essential for a robust assessment of regionalization performance.

Assessment of bottom-up satellite rainfall products on estimating river discharge and hydrologic signatures in Brazilian catchments

Satellite rainfall products are one of the most valuable tools for water resources monitoring in data-scarce regions, due to their low latency and quasi-global range. However, there are still uncertainties associated with rainfall products performance used to estimate hydrologic signatures in several regions, such as Brazil. Here, we investigate the performance of three rainfall products in estimating daily precipitation, daily river discharge, and hydrologic signatures over Brazil: the SM2RAIN-ASCAT and the GPM+SM2RAIN satellite products, and the ERA5 reanalysis product. We used a subset of 520 catchments from the Catchments Attributes for Brazil (CABra) dataset and the hydrologic modeling was carried out using the MISDc hydrologic model. Satellite-based products performed better than ERA5 for most Brazilian biomes in estimating daily precipitation when compared with ground observations used as reference. Daily river discharge was also better modeled with SM2RAIN-ASCAT and GPM+SM2RAIN. Hydrologic modeling presented low values of bias and >80% of catchments with KGE > 0.5 in calibration. Lastly, hydrologic signatures were well estimated by SM2RAIN-ASCAT and GPM+SM2RAIN, and for some biomes (Atlantic Forest, Cerrado, and Caatinga) they are better predictors than ground-based observations. We showed that there is a significant added value when using SM2RAIN-ASCAT and GPM+SM2RAIN products in tropical catchments, allowing a high-quality continuous water resources monitoring even in data-scarce regions. Besides, our findings pave the way for a better understanding of hydrologic extremes (droughts and floods) using these satellite rainfall products on multiple spatial and temporal scales.

Unravelling runoff processes in Andean basins in northern Ecuador through hydrological signatures

AbstractStreamflow is the runoff response integrated in space and time over a complex system involving climatic and catchment physiographic factors. In the Andes, accelerating runoff process understanding is hampered by the inability to quantify heterogeneity of surface and subsurface catchment properties. Here, we present a statistical approach based on regression models and correlation analysis that links hydrological signatures and catchment properties to unveil processes in a set of volcanic mountain catchments (latitude 0°30'N) in Ecuador. The catchments represent form and function diversity in the same hydrological unit. We found that despite of similar atmospheric‐water inputs the water yield in the north‐east region is about 5× larger than in the south‐west region and their flow regimes are asymmetric. The soil‐bedrock interface and lithology exert a first‐order control on hydrologic partitioning, and this allowed us to hypothesize two hydrological mechanisms. Firstly, in the north‐east region, the perennial streamflow is associated with seasonal rainfall patterns, and subsequent drainage processes taking place at the surface and subsurface level. The amount of streamflow is related to landform characteristics, high canopy density and root development of forest as well as water holding capacity of organic soils. From a mechanistic standpoint, the low concentration time, steep slopes and shallow infiltration limited by high‐consolidated deposits of sedimentary and volcanics suggest a lateral movement of the flow. Secondly, in the south‐west region the streamflow regime is mostly groundwater‐dependent and it becomes seasonally enhanced by rainfall. Larger seasonal variations of precipitation and temperature result into enhanced evapotranspiration in the drier months, limiting shallow soil infiltration. Under the soil layers, highly permeable pyroclastic deposits and andesitic lavas promote deep percolation. The results highlight the degree of dissimilarity of hydrological processes in Andean settings, but unravelling their complexity seems plausible using streamflow signatures and causal explanatory models.

Delineating modelling uncertainty in river flow indicators with representative parameter sets

For mitigating negative effects of floods and droughts, estimates of flow indicators and their uncertainties are essential. The recently introduced concept of the representative parameter sets (RPSs) enables modelling uncertainty to be represented in the flow frequency space at low computational cost, using a small subset of pre-selected model parameter sets. This concept is here adapted to assess hazards of three flow indicators: annual maximal flow, annual 7-day-average low flow, and annual mean flow. An additional in-depth analysis assesses the RPS transferability to other flow indicators and to hydrological signatures. RPS-based simulations are benchmarked with a random selection of parameter sets. The results show that i) RPSs can be successfully transferred between flow indicators with only a small drop in model performance; and ii) RPSs can be used to represent modelling uncertainty in hydrological signatures. The RPS concept has thus great potential for delineating modelling uncertainty of any environmental model.

The Drying Regimes of Non‐Perennial Rivers and Streams

AbstractThe flow regime paradigm is central to the aquatic sciences, where flow drives critical functions in lotic systems. Non‐perennial streams comprise the majority of global river length, thus we extended this paradigm to stream drying. Using 894 USGS gages, we isolated 25,207 drying events from 1979 to 2018, represented by a streamflow peak followed by no flow. We calculated hydrologic signatures for each drying event and using multivariate statistics, grouped events into drying regimes characterized by: (a) fast drying, (b) long no‐flow duration, (c) prolonged drying following low antecedent flows, (d) drying without a distinctive hydrologic signature. 77% of gages had more than one drying regime at different times within the study period. Random forests revealed land cover/use are more important to how a river dries than climate or physiographic characteristics. Clustering stream drying behavior may allow practitioners to more systematically adapt water resource management practices to specific drying regimes or rivers.

Interactions of Microtopography, Slope and Infiltration Cause Complex Rainfall‐Runoff Behavior at the Hillslope Scale for Single Rainfall Events

AbstractMicrotopography (MT) can govern runoff dynamics as a net result of local heterogeneities in the flow paths and ponding. This in turn controls the development of the surface water layer that connects and flows downslope. It is therefore important to understand which microtopographic features affect runoff generation dynamics and its macroscopic—hillslope scale—hydrological signatures (e.g., hydrographs, runoff and infiltration volumes). In this study, we numerically solve 2D overland flow from a single rain pulse on 1,460 idealized hillslopes with different slopes and sinusoidal microtopographies and different infiltration capacities. We assess hydrodynamic distributions, hydrographs and hydrological indices to assess the effects of MT and infiltration on the (local) hydrodynamic and (larger scale) hydrologic responses in terms of surface runoff regimes. The results show that MT enhances infiltration and that infiltration and runoff depend in a strong non‐linear way on slope and the properties of MT. Three regimes of influence of MT were identified: one in which MT plays a negligible role but there is a high sensitivity to the infiltration capacity curve, a second regime in which hydrological partitioning is highly sensitive to MT and the infiltration capacity curve, and a third regime in which MT increases infiltration, but the response is insensitive to particular features, and more affected by the average slopes. The regimes are the product of the interplay between small (MT) and large scale (slope) properties. Furthermore, the results suggest that hydrological signatures can be interpreted and explained by the spatiotemporal variation of surface connectivity.

Precipitation Biases in the ECMWF Integrated Forecasting System

AbstractPrecipitation is a key component of the global water cycle and plays a crucial role in flooding, droughts, and water supply. One way to manage its socioeconomic effects is based on precipitation forecasts from numerical weather prediction (NWP) models, and an important step to improve precipitation forecasts is by diagnosing NWP biases. In this study, we investigate the biases in precipitation forecasts from the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (IFS). Using the IFS control forecast from 12 June 2019 to 11 June 2020 at 5219 stations globally, we show that in each of the boreal winter and summer half years, the IFS 1) has an average global wet bias and 2) displays similar bias patterns for forecasts starting at 0000 and 1200 UTC and across forecast days 1–5. These biases are dependent on observed (climatological) precipitation; stations with low observed precipitation have an IFS wet bias, while stations with high observed precipitation have an IFS dry bias. Southeast Asia has a wet bias of 1.61 mm day−1 (in boreal summer) and over the study period the precipitation is overestimated by 31.0% on forecast day 3. This is the hydrological signature of several hypothesized processes including issues specifying the IFS snowpack over the Tibetan Plateau, which may affect the mei-yu front. These biases have implications for IFS land–atmosphere feedbacks, river discharge, and for ocean circulation in the Southeast Asia region. Reducing these biases could lead to more accurate forecasts of the global water cycle.

Assessing watershed hydrological response to climate change based on signature indices

Abstract Due to the fact that one of the important ways of describing the performance of basins is to use the hydrological signatures, the present study investigates the effects of climate change using the hydrological signatures in Azarshahr Chay basin, Iran. To this end, the Canadian Earth system model (CanESM2) is first used to predict future climate change (2030–2059) under two Representative Concentration Pathways (RCP2.6 and RCP8.5). Six signature indices were extracted from flow duration curve (FDC) as follows: runoff ratio (RR), high-segment volume (FHV), low-segment volume (FLV), mid-segment slope (FMS), mid-range flow (FMM), and maximum peak discharge (DiffMaxPeak). These signature indices act as sorts of fingerprints representing differences in the hydrological behavior of the basin. The results indicate that the most significant changes in the future hydrological response are related to the FHV and FLV and FMS indices. The BiasFHV index indicates an increase in high discharge rates under the RCP8.5 scenario, compared to the baseline period, and also the RCP2.6 scenario. The mean annual discharge rate, however, is lower than the discharge rate under this scenario. Generally, for the RCP8.5 scenario, the changes in the signature indices in both high discharges and low discharges are significant.

ADHI: the African Database of Hydrometric Indices (1950–2018)

Abstract. The African continent is probably the one with the lowest density of hydrometric stations currently measuring river discharge despite the fact that the number of operating stations was quite important until the 1970s. This new African Database of Hydrometric Indices (ADHI) provides a wide range of hydrometric indices and hydrological signatures computed from different sources of data after a quality control. It includes 1466 stations with at least 10 years of daily discharge data over the period 1950–2018. The average record length is 33 years, and 131 stations have complete records over 50 years. With this new dataset spanning most climatic regions of the African continent, several hydrometric indices have been computed, representing mean flow characteristics and extremes (low flows and floods), and are accessible to the scientific community. The database will be updated on a regular basis to include more hydrometric stations and longer time series of river discharge. The ADHI is available for download at: https://doi.org/10.23708/LXGXQ9 (Tramblay and Rouché, 2020).

Signatures of Hydrologic Function Across the Critical Zone Observatory Network

AbstractDespite a multitude of small catchment studies, we lack a deep understanding of how variations in critical zone architecture lead to variations in hydrologic states and fluxes. This study characterizes hydrologic dynamics of 15 catchments of the U.S. Critical Zone Observatory (CZO) network where we hypothesized that our understanding of subsurface structure would illuminate patterns of hydrologic partitioning. The CZOs collect data sets that characterize the physical, chemical, and biological architecture of the subsurface, while also monitoring hydrologic fluxes such as streamflow, precipitation, and evapotranspiration. For the first time, we collate time series of hydrologic variables across the CZO network and begin the process of examining hydrologic signatures across sites. We find that catchments with low baseflow indices and high runoff sensitivity to storage receive most of their precipitation as rain and contain clay‐rich regolith profiles, prominent argillic horizons, and/or anthropogenic modifications. In contrast, sites with high baseflow indices and low runoff sensitivity to storage receive the majority of precipitation as snow and have more permeable regolith profiles. The seasonal variability of water balance components is a key control on the dynamic range of hydraulically connected water in the critical zone. These findings lead us to posit that water balance partitioning and streamflow hydraulics are linked through the coevolution of critical zone architecture but that much work remains to parse these controls out quantitatively.

Analysis of geomorphological character of Machkula watershed (4D5B3) in Bhima lower Sub-basin

There has been an intensive search in recent years for suitable strategies to organize and classify the heterogeneous group of catchments that characterize landscape. The focus on testing the value of hydrological signatures is mostly derived from morphometric analysis of the watershed. The study of Machkula watershed (454.41 sq. km) involves its evaluation Interms of slope (0.28%), shape factor (11.04) and circulatory ratio (0.22) which influences the hydrological response. The steeper cross-slopes (0.27% to 3.04%) at its 3rd order network within flatter terrain as it appear at 6th order has greater impact on drainage pattern (2.93 to 4.83) at meso and micro scales of watersheds and derive proportional gravitational influences on runoff generation and sediment transportation. Its shape (11.04) at synoptic level under influence of topographic complexity Potray’s the shapes at 3rd order network (2.31 to 10.34) indicating its varied influence on runoff and sedimentation transport process.

TOSSH: A Toolbox for Streamflow Signatures in Hydrology

We present a Matlab toolbox to calculate hydrologic signatures, which are metrics that quantify streamflow dynamics. Signatures are widely used for catchment characterisation, hydrologic model evaluation, and assessment of instream habitat, but standardisation across applications and advice on signature selection is lacking. The toolbox provides accessible, standardised signature calculations, with clear information on methodological decisions and recommended parameter values. The toolbox implements three categories of signatures: basic signatures that describe the five components of a natural streamflow regime, signatures from benchmark papers, and an extended set of process-based signatures. The toolbox is designed for ease of use, including documentation, workflow scripts and example data to demonstrate implementation procedures, and visualisation options. We demonstrate the accuracy and robustness of the signature calculations by applying reproducible workflows to large streamflow datasets. The modular design of the toolbox allows for flexibility and easy future expansion. The toolbox is available from https://github.com/TOSSHtoolbox/TOSSH (https://doi.org/10.5281/zenodo.4313275).

Including Regional Knowledge Improves Baseflow Signature Predictions in Large Sample Hydrology

AbstractA catchment's hydrological response is controlled by climatic forcing and by the landscape through which water moves. Yet when we compare large samples of catchments, we often find climate to be the only good predictor of the hydrological response and a lot of variability is left unexplained. This contradicts extensive evidence from field and regional studies which shows the importance of catchment form (e.g., geology) on catchment hydrological processes, particularly on baseflow processes. We hypothesize that this is due to limitations in (a) the catchment attributes we use to inform our analyses and (b) the hydrological signatures we use to describe the hydrological response. To test these hypotheses, we use a large sample of catchment data across the contiguous United States. By reviewing literature from several U.S. regions, we show that region‐specific knowledge is underutilized in large sample studies. To organize the findings from these regions, we propose and apply a framework based on standardized perceptual models. Informed by these perceptual models, we use both available and newly calculated catchment attributes to show that baseflow signature predictions can be improved regionally. Multiple baseflow signatures are needed to better distinguish between different baseflow sources, such as the subsurface, surface water bodies, and snow. We conclude with pointing at potential future directions and argue that we should aim at a more systematic and hydrologically motivated selection of catchment attributes and hydrological signatures.

Explanation and Probabilistic Prediction of Hydrological Signatures with Statistical Boosting Algorithms

Hydrological signatures, i.e., statistical features of streamflow time series, are used to characterize the hydrology of a region. A relevant problem is the prediction of hydrological signatures in ungauged regions using the attributes obtained from remote sensing measurements at ungauged and gauged regions together with estimated hydrological signatures from gauged regions. The relevant framework is formulated as a regression problem, where the attributes are the predictor variables and the hydrological signatures are the dependent variables. Here we aim to provide probabilistic predictions of hydrological signatures using statistical boosting in a regression setting. We predict 12 hydrological signatures using 28 attributes in 667 basins in the contiguous US. We provide formal assessment of probabilistic predictions using quantile scores. We also exploit the statistical boosting properties with respect to the interpretability of derived models. It is shown that probabilistic predictions at quantile levels 2.5% and 97.5% using linear models as base learners exhibit better performance compared to more flexible boosting models that use both linear models and stumps (i.e., one-level decision trees). On the contrary, boosting models that use both linear models and stumps perform better than boosting with linear models when used for point predictions. Moreover, it is shown that climatic indices and topographic characteristics are the most important attributes for predicting hydrological signatures.

Time‐Varying Sensitivity Analysis Reveals Relationships Between Watershed Climate and Variations in Annual Parameter Importance in Regions With Strong Interannual Variability

AbstractClimate change impacts on hydroclimatology are becoming increasingly apparent around the world. It is unknown how annual variations in precipitation and air temperature alter the model‐inferred importance of hydrological processes and how this varies across watersheds. To examine this, we used parsimonious rainfall‐runoff model and applied time‐varying sensitivity analysis across 30 Californian watersheds for a 33‐year period (1981–2014). We calculated annual total order sensitivity indices for five performance metrics: Kling‐Gupta Efficiency (KGE) and model error in simulating hydrologic signatures (runoff ratio, slope of flow duration curve, baseflow index, and timing of streamflow centroid). Sensitivity of hydrological signatures to the parameters differed by signature, while parameter importance with respect to KGE was much more spatially and temporally variable. Variations in parameter sensitivity with respect to KGE were either correlated with air temperature (snow‐dominated sites in the Sierra Nevada) or precipitation (lower elevation sites < 1,300 m). Across error metrics and signatures, parameter sensitivity strongly differed between wet and dry years for a subset of our study sites. While parameter importance varied through time, parameter sensitivity variations across watersheds were much more pronounced. This suggests that parameter controls on model performance are much more a reflection of watershed properties as opposed to being dominantly shaped by shifts in precipitation and air temperature. These findings emphasize the importance of understanding simulated watershed responses to fluctuating annual conditions, as this conceptual knowledge is necessary to anticipate similarities and differences in response across even relatively proximal watersheds in the face of growing extreme conditions.