1. Introduction
This paper presents projected hydrologic conditions under climate change in three small headwater basins in the Mataura River basin in southern New Zealand using a basin catchment hydrologic model and an ensemble of global climate model (GCM) output from phase 6 of the World Climate Research Programme’s Coupled Model Intercomparison Project (CMIP6).
Insights on how climate change will impact seasonal streamflow volume, flood-peak timing and magnitude, and summertime base flows in a river basin are essential to water managers and stakeholders involved with aquatic freshwater habitat requirements and water supply. Global climate models can be used in conjunction with hydrologic models to evaluate the sensitivity and effects of future climate change on freshwater resources in numerous applications throughout the world. In these studies, future air temperature and precipitation data simulated from multiple GCM runs using various greenhouse gas emission scenarios are used as input to a hydrologic model. Simulated output from the hydrologic model usually includes basin outlet streamflow along with other basin processes such as potential and actual evapotranspiration, snowpack accumulation and melt, interception, sublimation, infiltration, and groundwater flux. To assess the effects of climate change on mean annual runoff throughout the conterminous United States, Wolock and McCabe (1999) used a simple water-balance model and output from two atmospheric GCMs. However, their results were uncertain because they were mostly within the range of GCM decade-to-decade variability and GCM model error. To simulate hydrologic climate changes at a smaller watershed scale, downscaled GCM air temperature and precipitation data can be input to more detailed distributed watershed models that simulate streamflow in addition to other water and energy fluxes within a basin. Hamlet and Lettenmaier (1999) statistically downscaled precipitation and temperature data from four GCMs to evaluate surface-water response of the Columbia River basin. Using the grid-based variable infiltration capacity (VIC) hydrologic model, climate-change-driven simulations resulted in increased winter streamflow, reduced winter snow accumulation, and reduced spring and summer streamflow. Hay et al. (2011) and Markstrom et al. (2012) simulated watershed-scale responses to projected climate change over the twenty-first century in 14 selected basins across the United States. In their study, they used precipitation and air temperature output from five GCMs as input to a physically based distributed watershed model [Precipitation-Runoff Modeling System (PRMS)] for each of the basins. Simulated streamflow responses to climate change from that study were statistically compared to identify regional trends (Risley et al. 2011). PRMS has also simulated the spatial and temporal response of runoff to climate change in a complex large river basin (Chang and Jung 2010).
In New Zealand hydrologic catchment models have also been used to simulate hydroclimatic conditions under future climate change scenarios. Fowler (1999) simulated potential climate change impacts on water resources in the Auckland, New Zealand, region using a daily water balance model and emission scenario forcings from the First Assessment Report of the Intergovernmental Panel on Climate Change (IPCC). Both Poyck et al. (2011) and Gawith et al. (2012) used TopNet model and GCM-simulated climate data based on a “middle of the road” emission scenario forcing from the IPCC Fourth Assessment Report to simulate combined snow and streamflow conditions in the Clutha River basin in the South Island of New Zealand. Hendrikx et al. (2012), Hendrikx and Hreinsson (2012), and Hendrikx et al. (2013) simulated the potential impact of climate change on seasonal snow in New Zealand using a temperature index model and climate input using a “middle of the road” emission scenario forcing from the IPCC Fourth Assessment Report. Caruso et al. (2017) simulated climate change effects on hydropower operations in mountain headwater lakes in the South Island of New Zealand. Climate output from an ensemble of 12 GCMs from the IPCC Fourth Assessment Report were input to TopNet, which was linked with a hydropower lake balance model. Jobst (2017) and Jobst et al. (2018) used a fully distributed hydrological model, the Water Flow and Balance Simulation Model (WaSiM), to simulate potential impacts of climate change on the hydroclimate of an alpine catchment in the Clutha River basin in the South Island of New Zealand. Climate input data to WaSiM were simulated from GCMs and downscaled with regional climate models and were based on emission scenario forcings from the IPCC Fourth Assessment Report. A broad brushed view of New Zealand hydrologic conditions under twenty-first century climate change is presented in Collins (2020, 2021). TopNet was used, as a noncalibrated model, in these studies to project future hydrologic impacts at 43 862 river locations throughout the entire country for seven streamflow metrics. Future climate input data were based on emission scenario forcings from the IPCC Fifth Assessment Report (IPCC 2013; Ministry for the Environment 2018). Hydrological projections for the headwaters of the Mataura River indicate that by the end of the century mean-annual river discharge, mean-seasonal discharge (for winter, spring, and fall), and mean-annual flooding all increase with warming. Mean summer discharge and low-flow discharge, characterized by 7-day mean-annual low flow, remain stable (Collins 2020).
A major difference between CMIP6 GCMs and CMIP5 GCMs is increased climate sensitivity to changes in greenhouse emissions with some of the CMIP6 models (Hausfather 2019; Zelinka et al. 2020; Bjordal et al. 2020; Frey et al. 2017). Equilibrium climate sensitivity (ECS), used as one indicator of climate sensitivity, is the long-term response in global atmospheric air temperature that would occur with a doubling of atmospheric CO2 concentrations (Hausfather 2018). Before CMIP6 there was a general consensus from previous studies and assessments that true ECS ranged from 1.5° to 4.5°C (National Research Council 1979; IPCC 2013). However, ECS estimates from 15 GCMs of a selection of 40 CMIP6 GCMs were higher than 4.5°C, with the highest estimate at 5.6°C (Hausfather 2019). The increased sensitivity has been attributed to simulated cloud feedback processes in CMIP6 models that did not exist in earlier CMIP models (Highwood 2018; Hausfather 2018). The role of cloud feedback processes in climate change is undoubtedly complex. As global temperatures rise cloud processes can produce both positive (warming) and negative (cooling) climate feedbacks. Low-altitude clouds, which are high in liquid moisture, can have a negative feedback since they reflect sunlight back to space, whereas high-altitude clouds can trap more heat in the atmosphere as ice crystals transition to a liquid form under warming. CMIP6 models could be producing a net positive feedback due to increased land surface sunlight adsorption, decreased extratropical low-altitude cloud coverage, and decreased reflectivity due to decreased albedo and decreased water content. However, the inclusion of cloud feedback processes in CMIP6 simulations and their higher ECS estimates does not necessarily translate into more accurate climate change predictions (Zelinka et al. 2020; Scafetta 2022).
The purpose of this paper is to illustrate plausible hydrologic responses to future climate change in three small headwater basins under varying climate sensitivities and emissions scenarios. The modeling application used precipitation and air temperature input data from 10 CMIP6 GCMs with ECS values ranging from 1.8° to 5.6°C. Despite uncertainty in these datasets, the range of hydrologic responses can be assessed and ranked by users of those results for their reasonableness and plausibility.
2. Study area
The Mataura River basin covers an area of approximately 5400 km2 in the South Island of New Zealand (Hughes et al. 2011). The study area for this paper are three headwater basins located upstream of the Mataura River at Parawa (801 km2), Waikaia River at Piano Flat (493 km2), and Waimea Stream at Mandeville (398 km2) (Fig. 1, Table 1). Computed mean elevations in the Mataura River at Parawa, Waikaia River at Piano Flat, and Waimea Stream at Mandeville basins are approximately 750, 1160, and 190 m above sea level, respectively.
Mataura River headwater basins in southern New Zealand (the map was created by J. Mangano).
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Mataura River headwater basin stream gauges used for model calibration and climate change simulations (data source: Environment Southland Regional Council; http://envdata.es.govt.nz/index.aspx?c=flow).
New Zealand South Island weather patterns are dominated by westerly airflows, which create a wet-to-dry gradient over the island from west to east. Mean monthly air temperatures in Gore, New Zealand, centrally located within the Mataura River basin, range from approximately 14.0°C in January to 5.0°C in July. Higher-elevation locations, farther inland from Gore, exhibit greater seasonal temperature extremes. Mean annual precipitation typically ranges from 950 mm in Gore to over 1200 mm in the surrounding northern uplands. On a seasonal basis, precipitation is relatively consistent throughout the basin. Highest and lowest precipitation totals occur during the summer and winter, respectively (Hughes et al. 2011; Wilson 2008).
Land use in the three headwater basins includes agricultural pasture, grass, and forest lands. The two higher-elevation basins, Mataura River at Parawa and Waikaia River at Piano Flat basins are characterized as mostly alpine and high-country tussock grassland (Wilson 2008). The lower-elevation Waimea Stream at Mandeville basin is flatter and contains agricultural pasture lands. Soils in Mataura River at Parawa and Waikaia River at Piano Flat are loamy and well drained. However, the Waimea Stream at Mandeville is dominated by clay and loam soils.
Water resource management is driven by the Mataura Water Conservation Order (Ministry for the Environment 1997) limiting surface water and groundwater extraction within the Mataura River catchment. Hydrology of the Mataura River at Parawa and Waikaia River at Piano Flat basins is associated mainly with surface-water processes with surface water-groundwater interactions limited to riverbed–groundwater bed exchange through a high yield shallow groundwater system (Wilson 2008). Hydrology in the Waimea stream is impacted by regional groundwater flow within the quaternary aquifer underlying the Waimea River catchment. This is augmented through groundwater recharge (∼1.6 m3 s−1) to the river between the Mataura River at Parawa and Riversdale township (Wilson 2008).
3. Modeling methods
a. Time series data
Measured historic daily precipitation and minimum and maximum air temperature data from the Mataura River basin were needed to calibrate the hydrologic model and for bias adjustments to the GCM output. Although actual measured data from locations inside the three study basins were unavailable, interpolated meteorological data at nine locations (Table 2; Fig. 1) are available from the National Institute of Water and Atmospheric Research (NIWA) Virtual Climate Station Network (https://cliflo.niwa.co.nz; accessed 6 June 2022). Additional parameters from these climate data locations included daily total global solar radiation and daily total Penman potential evapotranspiration, which were also used in the hydrologic model calibration.
Mataura River headwater basin climate data sites used for hydrologic model calibration and climate data bias corrections (data source: National Institute of Water and Atmospheric Research Virtual Climate Station Network; https://cliflo.niwa.co.nz; accessed 6 June 2022).
Continuous daily streamflow data were measured at three stream gauges by Environment Southland Regional Council (http://envdata.es.govt.nz/index.aspx?c=flow; accessed 23 May 2021), which were used to calibrate the hydrologic model (Table 1.) Benchmark statistics computed from 28 years of streamflow records (from 1 April 1990 to 31 March 2018) characterize how streamflow varies in magnitude and in seasonality (Table 3). Mean annual runoff yield for the Mataura River at Parawa, Waikaia River at Piano Flat, and Waimea Stream at Mandeville basins for this period are 710, 784, and 276 mm, respectively. The lower runoff yield in Waimea Stream is attributed to less precipitation and greater evapotranspiration than the other two basins. In the higher-elevation basins (Mataura River at Parawa and Waikaia River at Piano Flat) the high streamflows occur in the spring (September–November) because they are dominated by snowmelt runoff (Fig. 2). However, in the rain-dominated Waimea Stream at Mandeville basin, high streamflows occur during the winter (June–August).
Runoff yield in the headwater basins from 2005 to 2007, computed from measured daily streamflow data (source: Environment Southland Regional Council; http://envdata.es.govt.nz/index.aspx?c=flow).
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Daily mean streamflow benchmark statistics for Mataura River headwater basins for water years from 1 Apr 1990 to 31 Mar 2018; MALF = mean annual low flow; MAF = mean annual flow. Boldface type indicates the season of highest streamflow.
b. Hydrologic model
1) Description
The hydrologic model used in this paper is PRMS, version 5.0, which was developed by the U.S. Geological Survey (USGS) (Leavesley et al. 1983; Markstrom et al. 2015; Regan and LaFontaine 2017). PRMS is a deterministic, distributed-parameter, physical process-based modeling system that can simulate an array of basin water and energy processes such as snowpack, solar radiation, evapotranspiration, surface-water, and groundwater on a daily time step (Fig. 3). The model is often used to characterize the streamflow response of a basin to changes in long-term climate or land-use conditions. The minimum input data required to run PRMS are daily precipitation and minimum and maximum daily air temperature. In PRMS, a basin can be discretized into smaller spatial units having similar hydrologic response conditions. Called hydrologic response units (HRUs), they are delineated by the modeler as part of the calibration process. Although HRUs are not created from a fixed criterion or method, they are usually defined from land surface physical characteristics that can include elevation, slope, aspect, vegetation type and cover, soil characteristics, geology, drainage boundaries, distribution of precipitation, temperature, solar radiation, and flow direction, which are the discretion of the modeler (Markstrom et al. 2015). Runoff generated from upland HRUs can be routed through a stream network to the basin outlet. Daily streamflow data measured at the basin outlet are typically used to calibrate a PRMS model when compared with simulated streamflows. A partial bibliography of PRMS applications is available online (https://water.usgs.gov/water-resources/software/PRMS/PRMS-Bibliography.pdf).
Model schematic diagram of the PRMS hydrologic model. [This figure was used with permission from Koczot et al. (2005)].
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
2) Calibration
Basin water and energy processes are simulated in PRMS using model coefficients (or parameters) that define the physical characteristics of the basin. Many of these parameters, such as elevation, slope, aspect, vegetation type and cover, soils, and geology, are measured from geographic information system datasets and remain unadjusted during the model calibration. These data are available from Manaaki Whenua–Landcare Research, Ltd. (https://www.landcareresearch.co.nz; accessed 6 June 2022). Data and publications are accessed through Landcare’s Our Environment website (https://ourenvironment.scinfo.org.nz/maps-and-tools/app/; accessed 6 June 2022.) While other PRMS model parameters, mostly derived from empirically based research, were adjusted manually and automatically in the calibration process (Table S1 in the online supplemental material). Solar radiation and potential evapotranspiration (PET) model parameters were calibrated before calibrating the runoff and snow parameters since the runoff and snow parameterization in PRMS is dependent on solar radiation and PET processes and not vice versa. Mean monthly values computed from daily global solar radiation and PET daily data from the climate data points in the three basins for 1990–2018 were matched with corresponding mean monthly values computed from simulated daily output (Figs. 4 and 5). An optimization program was used to calibrate runoff and snow parameters using a multiple-objective, stepwise calibration strategy and a shuffled complex evolution global search algorithm (Hay and Umemoto 2006; Duan et al. 1993). Parameters sensitive to water balance, daily flow timing, high flows and low flows were calibrated against measured daily streamflow data separately in a stepwise manner (Table S1).
Measured and simulated mean-monthly global solar radiation for Mataura River at Parawa basin (1990–2018).
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Measured and simulated mean-monthly potential evapotranspiration for Mataura River at Parawa basin (1990–2018).
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
For this paper three separate PRMS models were constructed and calibrated for the basins upstream of stream gauges at Mataura River at Parawa, Waikaia River at Piano Flat, and Waimea Stream at Mandeville (Table 1). The models for Mataura River at Parawa, Waikaia River at Piano Flat, and Waimea Stream at Mandeville were discretized into 299, 133, and 196 HRUs, respectively. Daily air temperature and precipitation data (1994–2018) from nine separate climate data sites were used to calibrate and validate the models (Table 2). Each basin had three climate data sites at varying elevations (Fig. 1). HRUs in each basin were grouped with their nearest climate data site. Daily air temperature and precipitation input data supplied to the HRUs were adjusted by user defined lapse rates in PRMS to compensate for elevation differences between the climate data sites and the HRUs. A common calibration period from 1 April 1994 to 31 March 2006 was used for all three models. Measured and simulated mean monthly streamflow for Mataura River at Parawa basin (1994–2006) and daily streamflows for 5 years of this period are shown in Figs. 6 and 7, respectively. The calibration period was preceded by a model spinup period (1990–94). The period from 1 April 2006 to 31 March 2018 was used for model validation. Model performance statistics [Nash–Sutcliffe efficiency, coefficient of determination, percent bias, and RMSE–standard deviation (SDEV) ratio] were computed for the period from 1994 to 2018 (Table 4). Nash–Sutcliffe efficiency and coefficient of determination values greater than 0.5 are generally considered satisfactory (Moriasi et al. 2007). Percent bias values ±10% are considered very good (Moriasi et al. 2007). RMSE–SDEV ratio values between 0 and 0.50, 0.50 and 0.60, and 0.60 and 0.70 are considered very good, good, and satisfactory, respectively (Moriasi et al. 2007). Using these measures of performance, the Mataura River at Parawa model performed well. The Waimea Stream at Mandeville model performed very well, and the Waikaia River at Piano Flat model only performed satisfactorily.
Measured and simulated mean-monthly streamflow for Mataura River at Parawa basin (1994–2006).
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Measured and simulated daily-mean streamflow for Mataura River at Parawa basin (2001–06).
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Model performance statistics for Mataura River headwater basins for 1994–2018, computed from measured daily streamflow data (source: Southland Environment Regional Council; http://envdata.es.govt.nz/index.aspx?c=flow).
c. General circulation models
Daily precipitation and air temperature data representing possible future climate scenarios needed to run the PRMS models were derived from GCM output provided by the CMIP6 multimodel dataset archive (Eyring et al. 2016). Results from Coupled Model Intercomparison Projects provide critical research to regular assessment reports produced by the IPCC.
Because of uncertainty in climate modeling, it is desirable to use an ensemble of GCMs to assess future climate change. Ten GCMs were selected for this paper (Table 5). The ECS values of the 10 GCMs ranged from 1.8° to 5.6°. The lower and higher ECS groups are the GCMs below and above 4.5°, respectively. For each GCM the differences between the first and last 20 years of the century showed an increase in air temperature and precipitation that was generally consistent with the ECS values of the GCM (Fig. 8).
Air temperature and precipitation changes between the periods of 2001–20 and 2081–2100 for the Mataura River at Parawa basin using the SSP5-8.5 emissions scenario. ECS values (°C) are listed next to the GCM names. Results are computed from GCM datasets.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
CMIP6 GCMs used in the simulations. ECS is equilibrium climate sensitivity (°C). ECS values are from Hausfather (2019).
For each GCM, simulated precipitation and air temperature output representing two future economic and emission scenarios, called shared socioeconomic pathways (SSPs) (SSP2-4.5 and SSP5-8.5), were selected (Table S2 in the online supplemental material). SSPs represents scenarios of projected socioeconomic policies throughout the twenty-first century. These climate policies result in greenhouse gas emissions scenarios that have specified end-of-the-century climate change (radiative forcing) outcomes. Five different SSPs were created for the IPCC Sixth Assessment Report of Climate Change (IPCC 2021; Riahi et al. 2017). SSP2.4.5, represents “middle of the road” socioeconomic policies that result in a radiative forcing outcome in 2100 of 4.5 W m−2. SSP5-8.5, represents more extreme “Fossil-fueled development (Taking the highway)” socioeconomic policies that result in a more extreme radiative forcing outcome in 2100 of 8.5 W m−2.
Ten GCMs and 2 SSPs provided 20 datasets. Each dataset included daily precipitation and daily minimum and maximum air temperature time series for historic (1990–2014) and future (2015–2100) periods for the nine climate data sites in the headwater basins used to run the PRMS models (Table 2).
The GCM time series datasets were retrieved using OpenDap links provided by the World Climate Research Programme website (https://esgf-node.llnl.gov/projects/esgf-llnl/; accessed 26 May 2022). The USGS Geo Data Portal website (https://cida.usgs.gov/gdp/; accessed 26 May 2022) was used to download weighted-average time series files for each of the three headwater basins using the OpenDap links. Tables S3 and S4 in the online supplemental material include search descriptors to locate the GCM time series datasets on the World Climate Research Programme website and the dataset citations, respectively. The datasets were created by the institutions that operate the 10 GCMs (Table 5) in a daily time step format with a nominal resolution ranging from 100 to 500 km. With the exception of data bias adjustments, no additional spatial or temporal downscaling was performed on the datasets.
d. Data bias adjustments
Statistical and/or dynamic downscaling techniques are often used to adjust GCM output for bias before they are used as hydrologic model input (Murphy 2000). This can be necessary if the hydrologic model was calibrated with locally measured climate data. In contrast GCM output are simulated at a coarse grid scale, sometimes as large as 1° latitude and longitude. To rectify the bias, dynamic downscaling using regional climate modeling is preferable to statistical downscaling since atmospheric processes are highly nonlinear with respect to elevation. This can be significant in mountainous terrains, like the Mataura River basin, as shown by Rasmussen et al. (2011) who simulated current and future snowfall and snowmelt in mountainous Colorado headwater regions using high-resolution climate modeling. Prein et al. (2015) also make the case for high-resolution simulations using regional climate (convection permitting) models (grid spacing < 4 km). However, regional climate modeling is unavailable in many situations since it can require an inordinate amount of time, expense, and computational resources as compared with statistical techniques.
In lieu of dynamic downscaling, a simplistic statistical downscaling technique of adjusting bias in precipitation (or air temperature) GCM output is computing the mean of the differences between the GCM output and measured data in the historic period, when the two datasets overlap, and then using the mean to adjust future period values in the GCM output. This method is generally discouraged since it does not preserve the statistical moments of the measured time series. As a result, the adjusted GCM output will not realistically account for extreme events in the future period.
A more robust method of statistical bias adjustment is the “equidistant quantile matching” (EQM). This method uses differences between the cumulative distribution functions (CDFs) of the overlapping historic period time series to adjust the CDF of future period GCM climate output (Li et al. 2010). For this paper an approximation of this method was used to adjust the GCM air temperature and precipitation datasets. First, the percentiles of both the measured and GCM time series were computed using their period of overlap (1990–2014). For each k-value percentile, the difference between the measured and GCM percentiles was then computed. The GCM future period (2015–2100) time series was then adjusted by these differences depending on each value’s percent ranking in the time series. This procedure can be illustrated using hypothetical air temperature datasets (Fig. S1 in the online supplemental material). The same shape of the measured time series CDF is transferred to adjusted future GCM time series CDF. A similar CDF transfer function approach was used to adjust daily precipitation GCM output as described in Piani et al. (2010) (Fig. S2 in the online supplemental material). These data bias adjustments were made to all historic and future period GCM datasets for all nine climate data sites in the basin models.
4. Results and discussion
PRMS simulated a full suite of water and energy processes in the three headwater basins from 1990 to 2100 using all 10 GCMs under both SSP emission scenarios. Following a 4-yr model spinup (1990–93), the PRMS simulations included two periods used for comparison: a base period (1994–2014) and an end of the twenty-first century period (2081–2100). The means of the daily simulated climatic and hydrologic variables from PRMS output for the two periods were computed and compared (Table 6). (Temperature values are differences in degrees Celsius. All other values are percent changes.) Projected changes in all three basins were generally consistent with each other. This was somewhat expected since for some, though not all, of the GCMs all three basins were located within the same GCM grid cell.
Climatic and hydrologic changes between the periods of 1994–2014 and 2081–2100 for Mataura River headwater basins. Values are computed from PRMS model output and are multimodel means of all GCMs or by ECS group.
Maximum and minimum air temperature increases are greater under SSP5-8.5 than under SSP2-4.5 due to greater radiative forcing (Figs. 9 and 10). Toward to end of the century SSP5-8.5 multimodel mean air temperatures are approximately 2°–3°C greater than SSP2-4.5 air temperatures. Air temperatures of the high-ECS group are also approximately 2°–3°C greater than the low-ECS group. However, precipitation differences between the two SSPs in the low-ECS group were not as easily discernible (Fig. 11). In contrast, precipitation differences between the two SSPs for the high-ECS group noticeably diverge from each other in the latter half of the end of the century. Increased precipitation between is also seen on a monthly basis for both SSPs and both ECS groups (Fig. 12). With the exception of February, precipitation increases for all months over the base period.
Simulated maximum daily air temperature for Mataura River at Parawa basin (1994–2100). Lines are multimodel means of all GCMs or by group.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
As in Fig. 9, but for minimum daily air temperature.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
As in Fig. 9, but for precipitation totals.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Simulated mean-monthly precipitation totals for Mataura River at Parawa basin. Lines are multimodel means of all GCMs or by group.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Projected changes to annual and monthly streamflow in the basins are similar to precipitation and followed a similar pattern. Projected changes in streamflow between the base period and last 20-yr periods in the century for SSP2-4.5 and SSP5-8.5 are under 15% for the low-ECS group (Table 6). However, they increase upward toward 50% in the high-ECS group. Although seasonal streamflow generally follows seasonal precipitation, decreased snowpack, and earlier melting impacts streamflow in the Mataura River at Parawa and Waikaia River at Piano Flat basins as the timing of annual peak flows shifts toward winter and away from spring (Fig. 13). A similar shift is not seen in the Waimea Stream since it is rain dominated. Projected snowpack in the Mataura River at Parawa and Waikaia River at Piano Flat basins decrease under both emissions scenarios (Fig. 14). The timing of annual peak snowpack also shifts toward winter and away from spring. (Snowpack in the Waimea Stream at Mandeville basin is typically insignificant.)
Measured and simulated mean daily streamflow for three headwater basins. Lines are multimodel means of the GCM groups except for measured data line.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Simulated mean-daily snow-water equivalence in two headwater basins. Lines are multimodel means. Results are spatially averaged over each basin.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
Annual evapotranspiration increases for all basins, SSPs, and ECS groups by the end of the century (Table 6). Monthly evapotranspiration increases for all months for the high-ECS groups. Evapotranspiration also increases during the winter and spring for the low-ECS groups. However, it increases, or decreases, minimally for the low-ECS groups during the summer and fall (Fig. 15).
Simulated mean-monthly actual evapotranspiration totals for Mataura River at Parawa basin. Lines are multimodel means of all GCMs or by group.
Citation: Earth Interactions 27, 1; 10.1175/EI-D-22-0018.1
The water budget components of the three headwater basins (precipitation, evapotranspiration, and streamflow) all increase in magnitude for both ECS groups and both emissions scenarios (Table 7). Streamflow consistently increased for all groups and scenarios since the increased precipitation was greater than the increased evapotranspiration losses. However, the water budget ratios remain fairly stationary throughout the century.
Water budget changes between the periods of 1994–2014 and 2081–2100 for Mataura River headwater basins. Values are computed from PRMS model output and are multimodel means by ECS group.
5. Uncertainty discussion
Major sources of uncertainty in this study include the GCMs, downscaling techniques, bias adjustments, emission scenarios, and the basin hydrologic model.
Relative to earlier CMIP releases, CMIP6 GCMs had a wider range of climate sensitivities. Five of the GCMs used in this paper were greater than 4.5°C, which previous assessments and studies based on paleoclimate data had considered to be the upper range of ECS values (National Research Council 1979; IPCC 2013). The PRMS results presented in this paper using GCM data from the high-ECS group could be considered less plausible than results from the low-ECS group. Zelinka et al. (2020) and Scafetta (2022) determined that CMIP6 GCMs with higher ECS values, which included cloud feedback processes not included in CMIP5 GCMs, did not necessarily translate into more accurate climate projections.
GCMs are limited in their representation of Earth’s physical processes and feedbacks between the atmosphere, ocean, and land bodies due to their model structure and resolution. For the 10 GCMs used in this study, gridcell spacing of the atmospheric component ranged from 1° or greater in latitude and longitude (Table S5 in the online supplemental material). At the latitude of the study area in southern New Zealand (45°), the distance between 2° longitude is about 160 km, which is a climate input gridcell size that is too coarse for many basin hydrologic modeling applications. As a consequence, elevation and possible orographic differences within the grid cell introduce additional uncertainty to the modeling process.
The GCM gridcell spacing can differ with the nominal resolution of the dataset (WCRP CMIP6 data users guide: https://pcmdi.llnl.gov/CMIP6/Guide/dataUsers.html, accessed 29 September 2022). While the numerical computations in a GCM are simulated at the GCM gridcell scale, these computed values are reported to a nominal resolution for output. For this study all of the datasets were downloaded at 100-, 250-, and 500-km nominal resolutions.
Uncertainty exists in the discrepancy between GCM climate output and “measured” climate data for the study area. Sometimes the discrepancy between them can even be greater than the amount projected climate change. Because of nonlinear atmospheric processes, particularly in mountainous terrain, it is preferable to use regional climate modeling to downscale GCM datasets to a finer scale. However, regional climate modeling is often unavailable because of their expense and demand on computational resources. As an alternative, statistical bias adjustment techniques, such as the EQM technique, can be used to adjust GCM data. However, this requires an assumption that the statistical cumulative distribution functions of the datasets for the historic and future periods will have the same shape and characteristics.
The emission scenarios used to drive the GCMs also adds uncertainty because they are based on an array of plausible future political, social, economic, and technological factors. Two different future emission scenarios were used in this study as a means of widening the range of possible outcomes. Although SSP2-4.5 is not the lowest impacting emission scenario in CMIP6, it is economically and politically plausible and has a median radiative forcing at the end of the century (4.5 W m−2). SSP5-8.5, on the other hand, is a plausible “worst case” emission scenario with a high radiative forcing at the end of the century (8.5 W m−2).
Additional sources of uncertainty in the modeling process were associated with the hydrologic model. Like all hydrologic models, PRMS output is the culmination of error from many levels in the modeling process, such as data, parameterization, and model structure. Data measurement error exists in the air temperature and precipitation input data and the streamflow calibration data. Parameterization error is a function of how well the models were calibrated and verified by the modeler. Structural error is related to how well the hydrologic model represents the physical, water, and energy processes in the basin. Milly and Dunne (2011) discuss some of the uncertainty associated with the Jensen–Haise method (Jensen and Haise 1963; Jensen et al. 1970) of simulating PET used in PRMS. The method is an empirically based algorithm that uses only air temperature and solar radiation as input, does not account for the daily variation of relative humidity, and assumes complete mixing of the atmosphere above the evaporation surface. As a consequence, it may overestimate the amount of energy available for PET under warming conditions. Another source of model uncertainty is the lack of feedback mechanisms between PRMS and climate change. The PRMS basin models were calibrated for 1994–2018 land surface conditions. Present-day runoff processes were assumed to remain constant throughout the twenty-first century. A more ideal watershed model would include dynamic land surface parameters that are adjusted in the future as the basin’s physical environment is altered by climate change.
6. Summary
Global climate models can be used in conjunction with hydrologic models to assess the sensitivity and effects of future climate change on freshwater resources. Simulated future air temperature and precipitation data from multiple GCM runs using different greenhouse gas emission scenarios are then input to a hydrologic model. Simulated output from the hydrologic model typically includes streamflow at the basin outlet in addition to various basin processes such as potential and actual evapotranspiration loss, snowpack accumulation and melt, interception, sublimation, infiltration, and groundwater flux.
In this study the Precipitation-Runoff Modeling System was used to simulate the hydrologic response to climate change in three small headwater basins (398–801 km2) in the Mataura River basin in the South Island of New Zealand under varying climate sensitivities and emissions scenarios. The modeling application used precipitation and air temperature input data from 10 CMIP6 GCMs with ECS values ranging from 1.8° to 5.6°C. ECS values for one-half of the GCMs were greater than 4.5°C, which has been considered the upper range of ECS values according to previous assessments. All 10 GCM datasets included two CMIP6 greenhouse gas emissions scenarios: SSP2.4.5, represents moderate socioeconomic policies that result in a radiative forcing outcome in 2100 of 4.5 W per square meter; SSP5-8.5, represents greater fossil-fueled development socioeconomic policies that result in a radiative forcing outcome in 2100 of 8.5 W m−2.
All GCM air temperature and precipitation datasets were adjusted for bias using locally measured data and an approximation of the “equidistant quantile matching” (EQF) method. This method uses the differences between the cumulative distribution functions (CDFs) of the overlapping historic period time series to adjust the CDF of future GCM climate output.
PRMS is a deterministic, distributed-parameter, physical process-based modeling system that can simulate an array of basin water and energy processes such as snowpack, solar radiation, evapotranspiration, surface water, and groundwater on a daily time step. Model parameters were calibrated for three basin models using a multiple-objective, stepwise calibration strategy and a shuffled complex evolution global search algorithm for 1994–2006 against measured daily streamflow. Measured daily streamflows from 2007 to 2018 were used for validation.
Using bias-adjusted GCM datasets as input the calibrated PRMS models simulated climatic and hydrologic conditions for the entire twenty-first century. Maximum and minimum daily air temperatures increased between the base period (1994–2014) and the end of the century (2081–2100) for all GCM models, ECS groups, and emissions scenarios by varying amounts. Although total annual precipitation increased over the century for both ECS groups, it noticeably increased for the high-ECS group more than the low-ECS group. On a monthly basis precipitation significantly increased during the winter months for both ECS groups and both emissions scenarios. Precipitation for the high-ECS group increased for all months. However, summer precipitation did not increase with the low-ECS group. Changes in seasonal streamflow were similar to changes in seasonal precipitation with the high-ECS group showing the greatest increase. Annual snowpack in the higher-elevation basins decreased significantly. Evapotranspiration increased for almost all months for both ECS groups and both emissions scenarios with the exception of the low-ECS group with the SSP2-4.5 emissions scenario during the summer and fall months. Although evapotranspiration increased, streamflow increased for all groups and scenarios since the increased precipitation was greater than the increased evapotranspiration losses.
Major sources of uncertainty in this paper are the GCMs, higher climate sensitivities of some of the GCMs, downscaling, bias adjustment, emission scenarios, and the hydrologic model. One-half of the GCMs used in the paper have ECS values greater than 4.5°C, which earlier climate assessments considered to be the upper range of ECS values. Model results that used those GCM datasets could be considered less plausible. GCMs themselves are also limited in their representation of Earth’s physical processes and feedbacks between the atmosphere, ocean, and land bodies. The GCM gridded output data are spatially coarse. As a consequence, elevation and possible orographic differences within the grid cell introduce additional uncertainty to the modeling process. Emission scenarios used to drive the GCMs also adds uncertainty because they are based on a wide array of possible future political, social, economic, and technological factors. Simulated output from the hydrologic model, PRMS, added uncertainty since there is error in the data, calibration, and model structure.
Acknowledgments.
The authors thank Steve Markstrom of the U.S. Geological Survey and an anonymous reviewer for their reviews to Earth Interactions. We also thank Zach Freed of The Nature Conservancy in Oregon who provided funding for the publication cost of the paper. We also thank Joseph Mangano who provided cartographic assistance. The research in this paper was conducted independent of the U.S. Geological Survey. Their endorsement of findings in the paper is not implied or assumed.
Data availability statement.
Precipitation and air temperature data used in the PRMS calibration are available through the National Institute of Water and Atmospheric Research (NIWA) Virtual Climate Station Network (https://cliflo.niwa.co.nz; accessed 6 June 2022). Streamflow data used in the PRMS calibration are available through Environment Southland Regional Council (http://envdata.es.govt.nz/index.aspx?c=flow; accessed 10 June 2022). Land surface data used to create PRMS parameter files are available through Manaaki Whenua–Landcare Research, Ltd. (https://www.landcareresearch.co.nz; accessed 6 June 2022). Data and publications are accessed through Landcare’s Our Environment website (https://ourenvironment.scinfo.org.nz/maps-and-tools/app/; accessed 6 June 2022.) Final calibrated PRMS model input files are available from the author at pdxjcr@gmail.com. CMIP6 GCM precipitation and air temperature datasets are available for download through the World Climate Research Programme website (https://esgf-node.llnl.gov/projects/esgf-llnl/; accessed 26 May 2022; citations are given in Table S4 in the online supplemental material).
REFERENCES
Bjordal, J., T. Storelvmo, K. Alterskjær, and T. Carlson, 2020: Equilibrium climate sensitivity above 5°C plausible due to state-dependent cloud feedback. Nat. Geosci., 13, 718–721, https://doi.org/10.1038/s41561-020-00649-1.
Caruso, B. S., R. King, S. Newton, and C. Zammit, 2017: Simulation of climate change effects on hydropower operations in mountain headwater lakes, New Zealand. River Res. Appl., 33, 147–161, https://doi.org/10.1002/rra.3056.
Chang, H., and I.-W. Jung, 2010: Spatial and temporal changes in runoff caused by climate change in a complex large river basin in Oregon. J. Hydrol., 388, 186–207, https://doi.org/10.1016/j.jhydrol.2010.04.040.
Collins, D. B. G., 2020: New Zealand river hydrology under late 21st century climate change. Water, 12, 2175, https://doi.org/10.3390/w12082175.
Collins, D. B. G., 2021: Hydrological sentinels and the relative emergence of climate change signals in New Zealand river flows. Hydrol. Sci. J., 66, 2146–2154, https://doi.org/10.1080/02626667.2021.1987439.
Duan, Q. Y., V. K. Gupta, and S. Sorooshian, 1993: Shuffled complex evolution approach for effective and efficient global minimization. J. Optim. Theory Appl., 76, 501–521, https://doi.org/10.1007/BF00939380.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fowler, A., 1999: Potential climate change impacts on water resources in the Auckland region (New Zealand). Climate Res., 11, 221–245, https://doi.org/10.3354/cr011221.
Frey, W. R., E. A. Maroon, A. G. Pendergrass, and J. E. Kay, 2017: Do Southern Ocean cloud feedbacks matter for 21st century warming? Geophys. Res. Lett., 44, 12 447–12 456, https://doi.org/10.1002/2017GL076339.
Gawith, D., D. G. Kingston, and H. McMillan, 2012: The effects of climate change on runoff in the Lindis and Matukituki catchments, central Otago, New Zealand. J. Hydrol., 51, 21–136.
Hamlet, A. F., and D. P. Lettenmaier, 1999: Effects of climate change on hydrology and water resources in the Columbia River basin. J. Amer. Water Resour. Assoc., 35, 1597–1623, https://doi.org/10.1111/j.1752-1688.1999.tb04240.x.
Hausfather, Z., 2018: Explainer: How scientists estimate ‘climate sensitivity.’ Carbon Brief—Clear on Climate, https://www.carbonbrief.org/explainer-how-scientists-estimate-climate-sensitivity.
Hausfather, Z., 2019: CMIP6: The next generation of climate models explained. Carbon Brief—Clear on Climate, https://www.carbonbrief.org/cmip6-the-next-generation-of-climate-models-explained.
Hay, L. E., and M. Umemoto, 2006: Multiple-objective stepwise calibration using Luca. U.S. Geological Survey Open-File Rep. 2006–1323, 28 pp., https://pubs.usgs.gov/of/2006/1323/.
Hay, L. E., S. L. Markstrom, and C. Ward-Garrison, 2011: Watershed-scale response to climate change through the twenty-first century for selected basins across the United States. Earth Interact., 15, https://doi.org/10.1175/2010EI370.1.
Hendrikx, J., and E. Ö. Hreinsson, 2012: The potential impact of climate change on seasonal snow in New Zealand: Part II—Industry vulnerability and future snowmaking potential. Theor. Appl. Climatol., 110, 619–630, https://doi.org/10.1007/s00704-012-0713-z.
Hendrikx, J., E. Ö. Hreinsson, M. P. Clark, and A. B. Mullan, 2012: The potential impact of climate change on seasonal snow in New Zealand: Part I—An analysis using 12 GCMs. Theor. Appl. Climatol., 110, 607–618, https://doi.org/10.1007/s00704-012-0711-1.
Hendrikx, J., C. Zammit, E. Ö. Hreinsson, and S. Becken, 2013: A comparative assessment of the potential impact of climate change on the ski industry in New Zealand and Australia. Climatic Change, 119, 965–978, https://doi.org/10.1007/s10584-013-0741-4.
Highwood, E., 2018: Guest post: Why clouds hold the key to better climate models. Carbon Brief—Clear on Climate, https://www.carbonbrief.org/guest-post-why-clouds-hold-key-better-climate-models.
Hughes, B., S. Harris, and P. Brown, 2011: Mataura Catchment Strategic Water Study. Liquid Earth, Aqualinc Research, and Harris Consulting Rep., 232 pp., https://www.es.govt.nz/repository/libraries/id:26gi9ayo517q9stt81sd/hierarchy/about-us/plans-and-strategies/strategies/mataura-catchment-strategic-water-study/documents/mataura_catchment_strategic_water_study2.pdf.
IPCC, 2013: Summary for policymakers. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al. Eds., Cambridge University Press, 3–29, https://www.ipcc.ch/site/assets/uploads/2018/02/WG1AR5_SPM_FINAL.pdf.
IPCC, 2021: Summary for policymakers. Climate Change 2021: The Physical Science Basis, V. Masson-Delmotte et al., Eds., Cambridge University Press, 3–32, https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_SPM.pdf.
Jensen, M. E., and H. R. Haise, 1963: Estimating evapotranspiration from solar radiation. J. Irrig. Drain. Div., 89, 15–41.
Jensen, M. E., D. C. N. Rob, and C. E. Franzoy, 1970: Scheduling irrigations using climate-crop-soil data. J. Irrig. Drain. Div., 96, 25–38, https://doi.org/10.1061/JRCEA4.0000699.
Jobst, A. M., 2017: The potential impacts of climate change on the hydro-climate of the Clutha/Mata-Au catchment. Ph.D. thesis, University of Otago, 366 pp., https://hdl.handle.net/10523/7334.
Jobst, A. M., D. G. Kingston, N. J. Cullen, and J. Schmid, 2018: Intercomparison of different uncertainty sources in hydrological climate change projections for an alpine catchment (upper Clutha River, New Zealand). Hydrol. Earth Syst. Sci., 22, 3125–3142, https://doi.org/10.5194/hess-22-3125-2018.
Koczot, K. M., A. E. Jeton, B. J. McGurk, and M. D. Dettinger, 2005: Precipitation-runoff processes in the Feather River Basin, northeastern California, with prospects for streamflow predictability, water years 1971–97. U.S. Geological Survey Scientific Investigations Rep. 2004–5202, 92 pp., https://pubs.usgs.gov/sir/2004/5202/.
Leavesley, G. H., R. W. Lichty, B. M. Troutman, and L. G. Saindon, 1983: Precipitation-Runoff Modeling System—User’s manual. U.S. Geological Survey Water-Resources Investigations Rep. 83–4238, 214 pp., https://pubs.er.usgs.gov/publication/wri834238.
Li, H., J. Sheffield, and E. F. Wood, 2010: Bias correction of monthly precipitation and temperature fields from intergovernmental panel on climate change AR4 models using equidistant quantile matching. J. Geophys. Res., 115, D10101, https://doi.org/10.1029/2009JD012882.
Markstrom, S. L., and Coauthors, 2012: Integrated watershed-scale response to climate change for selected basins across the United States. USGS Investigations Doc. 2011–5077, 143 pp., https://pubs.er.usgs.gov/publication/sir20115077.
Markstrom, S. L., R. S. Regan, L. E. Hay, R. J. Viger, R. M. Webb, R. A. Payn, and J. H. LaFontaine, 2015: PRMS-IV, the precipitation-runoff modeling system (version 4). USGS Techniques and Methods Doc. 6-B, 158 pp., https://doi.org/10.3133/tm6B7.
Milly, P. C. D., and K. A. Dunne, 2011: On the hydrologic adjustment of climate-model projections: The potential pitfall of potential evapotranspiration. Earth Interact., 15, https://doi.org/10.1175/2010EI363.1.
Ministry for the Environment, 1997: Water Conservation (Mataura River) Order 1997. New Zealand Parliamentary Counsel Office Doc., 10 pp., https://www.legislation.govt.nz/regulation/public/1997/0126/latest/DLM234002.html.
Ministry for the Environment, 2018: Climate Change Predictions for New Zealand: Atmospheric Projections Based on Simulations Undertaken for the IPCC 5th Assessment. 2nd ed. New Zealand Government, 131 pp., https://environment.govt.nz/assets/Publications/Files/Climate-change-projections-2nd-edition-final.pdf.
Moriasi, D. N., J. G. Arnold, M. W. Van Liew, R. L. Bingner, R. D. Harmel, and T. L. Veith, 2007: Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE, 50, 885–900, https://doi.org/10.13031/2013.23153.
Murphy, J., 2000: Predictions of climate change over Europe using statistical and dynamical downscaling techniques. Int. J. Climatol., 20, 489–501, https://doi.org/10.1002/(SICI)1097-0088(200004)20:5<489::AID-JOC484>3.0.CO;2-6.
National Research Council, 1979: Carbon Dioxide and Climate: A Scientific Assessment. National Academies Press, 34 pp., https://doi.org/10.17226/12181.
Piani, C., J. O. Haerter, and E. Coppola, 2010: Statistical bias correction for daily precipitation in regional climate models over Europe. Theor. Appl. Climatol., 99, 187–192, https://doi.org/10.1007/s00704-009-0134-9.
Poyck, S., J. Hendrikx, H. Mcmillan, E. Hreinsson, and R. Woods, 2011: Combined snow- and streamflow modelling to estimate impacts of climate change on water resources in the Clutha River, New Zealand. J. Hydrol., 50, 293–312.
Prein, A. F., and Coauthors, 2015: A review on regional convection-permitting climate modeling: Demonstrations, prospects, and challenges. Rev. Geophys., 53, 323–361, https://doi.org/10.1002/2014RG000475.
Rasmussen, R., and Coauthors, 2011: High-resolution coupled climate runoff simulations of seasonal snowfall over Colorado: A process study of current and warmer climate. J. Climate, 24, 3015–3048, https://doi.org/10.1175/2010JCLI3985.1.
Regan, R. S., and J. H. LaFontaine, 2017: Documentation of the dynamic parameter, water-use, stream and lake flow routing, and two summary output modules and updates to surface-depression storage simulation and initial conditions specification options with the Precipitation-Runoff Modeling System (PRMS). USGS Techniques and Methods Doc. 6-B8, 72 pp., https://doi.org/10.3133/tm6B8.
Riahi, K., and Coauthors, 2017: The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Global Environ. Change, 42, 153–168, https://doi.org/10.1016/j.gloenvcha.2016.05.009.
Risley, J. C., H. Moradkhani, L. Hay, and S. Markstrom, 2011: Statistical comparisons of watershed-scale response to climate change in selected basins across the United States. Earth Interact., 15, https://doi.org/10.1175/2010EI364.1.
Scafetta, N., 2022: Advanced testing of low, medium, and high ECS CMIP6 GCM simulations versus ERA5-T2m. Geophys. Res. Lett., 49, e2022GL097716, https://doi.org/10.1029/2022GL097716.
Wilson, K., 2008: Surface water and groundwater relationships in the Mataura catchment above Gore. Environment Southland Tech. Rep. 2008-03, 67 pp., https://www.es.govt.nz/repository/libraries/id:26gi9ayo517q9stt81sd/hierarchy/environment/water/groundwater/groundwater-monitoring/documents/groundwater-reports/surface-water-and-groundwater-relationships-in-the-mataura-catchment-above-gore-2008.pdf.
Wolock, D. M., and G. J. McCabe, 1999: Estimates of runoff using water-balance and atmospheric general circulation models. J. Amer. Water Resour. Assoc., 35, 1341–1350, https://doi.org/10.1111/j.1752-1688.1999.tb04219.x.
Zelinka, M. D., T. A. Myers, D. T. McCoy, S. Po-Chedley, P. M. Caldwell, P. Ceppi, S. A. Klein, and K. E. Taylor, 2020: Causes of higher climate sensitivity in CMIP6 models. Geophys. Res. Lett., 47, e2019GL085782, https://doi.org/10.1029/2019GL085782.
