Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Journal of Hydrology 636 (2024) 131287

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Characterizing streamflow regimes using a distributed model for


sustainable resource management in the humid tropics
Ayron M. Strauch a, b, *, Yu-Fen Huang b, Yin-Phan Tsang b
a
Commission on Water Resource Management, Department of Land and Natural Resources, State of Hawai‘i, 1151 Punchbowl, Rm 227, Honolulu, HI 96813, Hawai‘i,
USA
b
Department of Natural Resources and Environmental Management, University of Hawai‘i at Mānoa, 1910 East-West Rd, Sherman 101, Honolulu, HI 96822, Hawai‘i,
USA

A R T I C L E I N F O A B S T R A C T

This manuscript was handled by Andras Bar­ Estimating streamflow regimes on small tropical islands is challenging due to limited observational data, poor
dossy, Editor-in-Chief accuracy of remote sensing, and the limitations of coarse-scale climate models, restricting our understanding of
hydrological processes and the management of water resources. The inadequacy of data has also affected the
Keywords: development of instream flow standards (i.e., environmental flows, e-flows) needed to protect ecological and
Humid tropics
cultural values central to island communities. Estimates of water availability for agriculture, hydropower, and
SWAT model
drinking water supply are also critical for informed investments in water resource planning. We utilized high-
Streamflow
Instream flow standards resolution (250 m) gridded daily rainfall data to parameterize the Soil and Water Assessment Tool (SWAT) to
Sustainability characterize streamflow regimes at ungauged locations in a watershed on Kaua‘i Island. The resultant flow
statistics were then used to describe the daily, seasonal, and annual availability of surface water to evaluate the
implications of potential e-flow scenarios on water available for run-of-river hydropower. Compared to rainfall
data derived from a single station within the watershed, the gridded daily rainfall product increased the accuracy
of model output from not satisfactory to good. Model techniques using high resolution rainfall data may sub­
stitute for long-term hydrological data collection across complex tropical environments and provide data for
making sustainable management decisions at ungauged locations.

1. Introduction increase the uncertainty in the consequences of water management


decisions.
The global demand for water to support agriculture, hydropower, While the availability of streamflow records greatly improves the
and drinking water supply is similarly playing out on remote tropical quantification of hydrologic processes, monitoring streamflow at all
islands which face great constraints on land and water availability locations is not feasible, and other methods are required to develop the
(March et al., 2003). On tropical islands, the availability to meet these data necessary to classify flow regimes. Physically-based hydrologic
demands is limited by watershed and aquifer size, climate patterns, and models can be utilized to simulate streamflow, if input data are available
accessibility across varied landscapes (Craig, 2003). The challenges of with sufficient reliability—specifically rainfall and fog drip (Cuo et al.,
characterizing streamflow regimes on tropical islands mirror the prob­ 2006; Juvik and Ekern, 1978). A lack of reliable rainfall data leads to
lems associated with measuring rainfall: high intensity, episodic rainfall poor agreement between modeled and observed flow estimates. Char­
rapidly produces overland flow (Strauch et al., 2018; Christian et al., acterizing natural flow regimes has become increasingly important for
2019); short flow paths and steep gradients efficiently transport runoff, understanding the consequences of resource management decisions on
resulting in steeply sloped hydrographs (time to peak flow within hours) freshwater ecosystems (Arthington and Pusey, 2003; Poff et al., 1997).
compared to continental systems (Tomlinson & De Carlo, 2003); and This is especially true where tradeoffs among water uses have broad
regional differences in hydrogeology affect uncertainty in surface water- consequences for the local and regional community (Cai et al., 2018).
groundwater interactions (Izuka et al., 2018). Each of these processes Limitations in the availability of observational data across the tropics
affect our ability to estimate streamflow at ungauged locations and has affected estimates of rainfall and streamflow (Spangler et al., 2017;

* Corresponding author.
E-mail address: astrauch@hawaii.gov (A.M. Strauch).

https://doi.org/10.1016/j.jhydrol.2024.131287
Received 1 November 2022; Received in revised form 13 April 2024; Accepted 17 April 2024
Available online 11 May 2024
0022-1694/© 2024 Published by Elsevier B.V.
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Wohl et al., 2012). Multiple hydrological models have been used to es­ opportunity to generate power from renewable water resources without
timate streamflow regimes in the tropics (Wohl et al., 2012). We utilized using scarce land for utility scale wind or solar power. This type of hy­
the Soil and Water Assessment Tool (SWAT), a semi-distributed, physi­ dropower relies on the withdrawal of water from streams at high ele­
cally based hydrological model (Arnold et al., 1998). The SWAT was vations to power a turbine at a lower elevation, taking advantage of the
previously used successfully for water resource management applica­ head produced by steep elevation gradients and reliable water supply
tions, including the modeling of climate change impacts on streamflow (Anderson et al., 2014; Lorentis et al., 2010). Such systems avoid the
and water yield, erosion and sediment modeling, land use change, the inundation of scarce land and its associated production of greenhouse
assessment of management practices on streamflow and sediment yield gases (Prairie et al., 2017).
(Fukunaga et al., 2015; Gassman et al., 2007). Further, Leta et al. (2016) The objective of this study was to generate natural flow regimes in a
successfully used SWAT to estimate hydrological processes in a similarly tropical watershed for a system of run-of-river stream diversions based
complex watershed in Hawai‘i. on high resolution rainfall data to test the consequences of specific e-
With new insights in the hydrogeology of volcanic islands and the flows on the availability of water for hydropower. We first compared
development of computationally advanced models, streamflow can be modeled streamflow values generated using a single rainfall station
estimated at ungauged locations (Strauch et al., 2017; Izuka et al., located within the watershed to the high-resolution daily rainfall
2018). However, the extent and accuracy of data serving as model inputs product available from Longman et al. (2019). Then, using mean daily
can affect the reliability of the model results (Guan et al., 2020). The flow quantified over a 20-year period (1995–2014), we determined: 1)
spatial and temporal pattern of rainfall in the tropics are influenced by the annual and seasonal characteristics of the natural flow regime in
both large-scale climate trends and micro-climate patterns that develop each of these headwater tributaries; 2) low-flow duration characteristics
over heterogeneous landscapes (Mair and Fares, 2011; Frazier and that might inform the development of e-flows; and 3) the availability of
Giambelluca, 2017). Due to the stochastic nature of daily rainfall, large water for hydropower production under various e-flow scenarios. A
variations in elevation, slope, and aspect increases the spatial variability discussion of how these data may help provide equitable management of
driven by orographic processes, rain shadow effects, or trade wind surface water follows.
strength, adding much uncertainty (Buytaert et al., 2006). This makes
the interpolation of daily rainfall either from point measurements or 2. Methods
remote sensing intrinsically difficult and the estimation of streamflow at
ungauged locations uniquely challenging (Newman et al., 2019). Ordi­ 2.1. Watershed characteristics
nary kriging methods used to interpolate rainfall from point measure­
ments can be effective across complex topographic features at longer (i. Run-of-river hydropower makes an important contribution to the
e., monthly, annual) time scales (Mair and Fares, 2011; Frazier et al., total energy portfolio of the Island of Kaua‘i. Wailua is a 136.23 km2
2016), but complications may develop at shorter time scales (i.e., daily) watershed on the eastern flank of the Wai‘ale‘ale mountain, bordered on
and results may be less reliable (Shen, 2001; Ly et al., 2011). In Hawai‘i, the north by the Makaleha Mountains and the south by the Kilohana
the high diversity of rainfall patterns are due to a combination of Volcano on the Island of Kaua‘i (22◦ 05′N 159◦ 30′W) within the Ha­
persistent trade winds generating orographic rainfall and fog drip, steep waiian island chain (Fig. 1). It forms the northern portion of the Lı̄hu‘e
elevation gradients, complex topography, and the inversion in the Basin, a unique geologic depression that also includes the Huleia,
temperature lapse rate that influences cloud height, rainfall distribution, Nawiliwili and Hanama‘ulu watersheds (Izuka et al., 2018). Two sepa­
and relative humidity at various elevations (Giambelluca and Nullet, rate volcanic phases produced the geology of the region: the Waimea
1991; Nullet et al., 1995;Giambelluca et al., 2013). Canyon Series and the Kōloa volcanic series (Macdonald et al., 1960).
Historically, gridded rainfall was available at coarse spatial (0.5 Tholeiitic Waimea Canyon basalt formed during the initial shield-
degree x 0.5 degree) or temporal (monthly) resolutions (Hassell et al., building phase of the island and composes the lower layers of the
2017; Xie and Arkin, 1997; Frazier and Giambelluca, 2017). Such scales Lı̄hu‘e Basin (Gingerich, 1999). The two formations of this series are the
provide poor agreement for the processes that occur at smaller spatial Napali formation, which is composed of thin-bedded permeable lava
scales (catchments) and shorter time scales (hours to days) in tropical flows that originated at the caldera boundary, and the Haupu formation,
islands (Tomlinson and De Carlo, 2003; Biasutti and Yuter, 2013). which is composed of thickly massive lava flows accumulated in a large
Higher resolution (250 m) daily rainfall and temperature data can vastly pit crater in the southeast (Macdonald et al., 1960). Stream channel
improve the ability of models to simulate climate patterns across a range incision has exposed volcanic dikes formed by Waimea Canyon basalt,
of conditions needed to estimate real-world hydrological events (Hofstra which impound high-elevation groundwater, contributing to the base
et al., 2008; Guo et al., 2018). Recently available data for Hawai‘i has flow of streams (Gingerich, 1999). In the second volcanic phase, thick
created new possibilities for modeling watershed hydrology to improve lava flows of mostly massive alkalic olivine basalts of the Kōloa Volcanic
streamflow management (Longman et al., 2019). Series dominate the Lı̄hu‘e Basin (Macdonald et al., 1960). This stage of
There is also a need to establish environmental flows (e-flows) that rejuvenated shield building filled valleys, gorges, and depressions in the
protect freshwater resources from over exploitation (Acreman, 2016). In Waimea Canyon basalt with heterogeneous lava flows, ash, tuff, cinder,
Hawai‘i, the State of Hawai‘i Commission on Water Resource Manage­ and sediments as much as 300 m thick (Macdonald et al., 1960). While
ment (Hawai‘i CWRM) is mandated to establish instream flow standards some studies have shown that the contact between Kōloa Volcanics and
(i.e., e-flows) that protect identified instream values (e.g., ecosystem the Waimea Basalt occurs at about 150 m in elevation, there is evidence
services, traditional and customary practices, fish habitat, water quality, that the thickness of the Kōloa Volcanic layer is highly variable and
recreation, and navigation) across the Hawaiian islands (Sakoda, 2007). dependent on the shape of the eroded surface (Gingerich, 1999). This
Effective, informed management decisions to develop water for geology results in substantial groundwater contributions to streamflow
municipal, agricultural, or hydropower uses are thus limited by a lack of in the lower gradient reaches and affects deep groundwater routing
streamflow data characterizing natural flow regimes. within the SWAT model.
Hydroelectric power generation utilizes renewable surface water, Rainfall in the watershed is affected by east-northeasterly trade
reducing the dependence on fossil fuels and associated CO2 emissions. winds that push warm, moist air against the windward side of Wai‘a­
However, large reservoirs have many negative drawbacks in the tropics, le‘ale (elevation 1569 m) and up into higher elevations where cooler
including reductions to freshwater habitat, the trapping of sediment and temperatures persist. As the moist air cools, water condenses, and the air
organic matter leading to siltation (Falinski and Penn, 2018) and re­ mass releases precipitation. Near the coast, mean annual rainfall is as
ductions in downstream and estuarine productivity (Ezcurra et al., little as 1,270 mm but in the headwaters of Wailua, annual rainfall can
2019). By contrast, run-of-river hydropower, provides a unique exceed 10,000 mm per year (Giambelluca et al., 2013). As a result,

2
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Fig. 1. Map of Wailua watershed and headwater tributary sub-basins (labeled), with U.S. Geological Survey stations (identification number), ditch/canal systems,
and hydropower plants, on the Island of Kaua‘i, Hawai‘i.

frequent and heavy rainfall is observed on the mountain slopes. The 2011). We divided the study watershed into 21 sub-basins (supple­
highest position in Wailua lies in the fog drip zone, a region of vegeta­ mental table 1) that were then further divided into hydrological
tion that intercepts moisture trapped below the temperature inversion response units (HRUs) based on geology, land use, soil type, and slope
(Scholl et al., 2002). While the temperature inversion zone typically (Neitsch et al., 2011). A greater number of smaller sub-basins were
extends from 2,100 m to 2,700 m, climate below the inversion can be nested at higher elevations to better represent topographic heteroge­
extremely variable, and difficult to estimate using a limited distribution neity (Fig. 1). ArcGIS compatible with SWAT 2012 was utilized to
of observations (Giambelluca and Nullet, 1992). manage the spatial data. The average (±standard deviation) size of each
The North Wailua Ditch traverses three headwater tributaries of the HRU was 0.0424 (±0.1467) km2.
Wailua River (drainage area = 136.2 km2), diverting water from each to We established the spatial characteristics with a 10 m x 10 m Digital
power the Upper Waiahi Hydropower Plant: Wai‘ale‘ale (drainage area Elevation Model (DEM) obtained from the USGS; 1:24,000 scale soil
= 4.56 km2), Waikoko (drainage area = 2.84 km2), and Ili‘ili‘ula maps from Soil Survey Geographic (SSURGO) database provided by the
(drainage area = 3.99 km2) (Fig. 1). The tailrace flow from this hydro­ U.S. Department of Agriculture Natural Resources Conservation Service
power plant is combined with the flow diverted from the Waiahi trib­ (USDA-NRCS), and 2010 land use data from National Oceanic and At­
utary to power the Lower Waiahi Hydropower Plant. mospheric Administration/Coastal Change Analysis Program (NOAA,
2010) for Kaua‘i (Table 1). The sub-basins utilized for estimated
2.2 Model set-up streamflow were all in the upper watersheds, above any agricultural or
urban land uses, and dominated by a mixture of native and non-native
The SWAT model is based on water balance that considers precipi­ wet forest species and wet cliff vegetation (see supplemental Fig. 1).
tation, surface runoff, baseflow, actual evapotranspiration, lateral flow, Based on the slope distributions of the entire watershed, five slope
percolation, and deep groundwater loss components (Neitsch et al., classes (the maximum number permissible in SWAT) were defined and

3
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Table 1 parameter-optimization by using the Sequential Uncertainty Fitting


Data type and source (year of dataset publication) for SWAT model (SUFI2) algorithm through SWAT Calibration and Uncertainty Program
development. (Abbaspour et al., 2007). The SUFI2 algorithm requires a sufficient
Data type Data source number of iterations to converge to a stable solution. Based on other
Mean daily streamflow USGS 16068000 EB NF Wailua River nr Lihue
SWAT publications in Hawai‘i (Leta et al., 2016), we started with 300
Daily rainfall Longman et al. (2018) iterations. However, at 300 iterations, we found the objective function
Daily temperature (maximum, Longman et al. (2018) did not fully stabilized, so we extended to 500 iterations. Parameters
minimum) (see supplemental table 1) were initially adjusted to optimize runoff and
Digital elevation model (10 m) USGS (2019) National Elevation Dataset
peak flow conditions which occur regularly in the humid tropics and
Landcover NOAA Coastal Change Analysis Program (2010)
Soil Soil Survey Staff; USGS Soil Survey Geographic must be sufficiently characterized to produce reliable flow duration
Database (2017) curves. To improve model agreement during baseflow conditions, the
saturated hydraulic conductivity (SOL_K) was set to match its range
from observations based on USGS reports (Gingerich, 1999; Izuka, 2006;
assigned: 0-≤9%; 9-≤20 %; 20-≤31.5 %; 31.5-≤46 %; and > 46 %. Izuka and Gingerich, 1998). Model outputs were evaluated graphically
Climate variables included daily rainfall, maximum and minimum to compare hydrograph recessions and the extent modeled data deviated
temperatures, relative humidity, solar radiation, and wind speed from observed values (Fig. 2). Flow duration curves were also used to
(Table 1). Daily rainfall and temperature data for the period between visualize the distribution frequency of modeled flows compared to
1990 and 2014 were obtained from high-resolution gridded 250 m x observed flows (supplemental Fig. 3).
250 m (0.0625 km2) daily rainfall and temperature maps (Longman The calibration and validation included multiple statistical perfor­
et al., 2019). Longman et al. (2019) used an inverse distance weighted mance measures based on Moriasi et al. (2007): the Nash-Sutcliffe effi­
approach coupled with long-term climatological aid to interpolate daily ciency (NSE) (Nash & Sutcliffe, 1970); the correlation coefficient
rainfall data (Willmott and Robeson, 1995). If the HRU incorporated reported as coefficient of determination (r2) (Legates et al., 1999); root
more than one 250 m grid, we used the zonal mean (the function in ESRI mean square error (RMSE) (Hodson, 2022); and percent bias (PBIAS)
ArcGIS Desktop 10.4.1) of the daily rainfall for that HRU. Using the (Gupta et al., 1999). Performance evaluation criteria for daily and
built-in WXGEN weather generator model, relative humidity and solar annual flow metrics were assessed using Moriasi et al., (2015; table 9).
radiation were generated based on the daily rainfall and temperature, The upper and lower 95 % confidence intervals (CI) of simulated flow
while wind speed was generated independently using Sharpley and were then reported and the median model run and the best single
Williams (1990). The daily streamflow data were assessed using the R simulation were compared to observed flow characteristics (Table 3).
package, dataRetrieval (De Cicco et al., 2018). We used the median simulated value (n = 500) for reporting model daily
flow statistics for each sub-basin in the analysis as it most closely
resembled the observed low-flow values in the calibration period. We
2.3. Model calibration and validation
also simulated streamflow at a second USGS station (USGS 160,715,000
Left Branch Opaekaa Str nr Kapaa) located at a slightly lower elevation
We used data from the US Geological Survey (USGS) station
(140 m a.m.s.l.), a smaller catchment area (1.9425 km2), and less
16,068,000 on the East Branch of the North Fork Wailua (catchment
forested cover (68.4 %). For comparative purposes, the model was also
area = 16.17 km2; 153 m a.m.s.l.; 74.9 % evergreen forest) to calibrate
run utilizing a single rainfall station (USGS 220,356,159,281,401
and validate the model. We applied the model to estimate mean daily
1051.0 N Wailua Ditch Rain Gage) located within the watershed.
flow (MDF) at USGS 16068000 from 1995 to 2014. This time period
captures multiple positive and negative PDO phases and El Niño/La
Niña southern oscillations that influence wet and dry season rainfall in
Hawai‘i (Frazier et al., 2018; supplemental Fig. 2). The sub-basin of
USGS 16068000 has similar underlying geology, topography, and
vegetation cover to the sub-basins of interest. Additional aggregated
sub-basin characteristics are provided in Table 2. The newly available
rainfall and temperature products from Longman et al. (2019) begins in
1990. The period of daily flow records was split into three periods: a
warm-up period (1990–1995) to initialize the state variables, a cali­
bration period (1995–2010), and a validation period (2010–2014).
These periods were of sufficient length to capture the variability in
seasonal flows as well as the magnitude in low flow periods (supple­
mental table 3). 500 iterations of the model were calibrated by opti­
Fig. 2. Example of observed and modeled mean daily flow (m3s− 1) at USGS
mizing model stream discharge with respect to observed values from
16068000 during the calibration period for calendar years 2000 to 2002.
USGS station 16,068,000 for calendar years 1995 to 2010. We applied

Table 2
Location name, USGS gauge ID, location and catchment characteristics for sub-basin locations used in model calibration/validation and ungauged locations where the
model was applied.
Location name USGS gauge ID Period of latitude longitude Catchment area mean catchment Minimum maximum
Record (ha) slope (%) elevation (m) elevation (m)

EB NF Wailua 16068000 1912-present 22.07405 − 159.423 1609.39 45.35 157 991


River
NF Wailua River 16063000 1914–1985 22.06259 − 159.458 1184.82 58.19 218 1598
Wai‘ale‘ale 220427159300201 ungauged 22.06659 − 159.484 455.81 98.74 335 1598
Stream
Waikoko Stream 220356159281401 ungauged 22.05591 − 159.478 284.18 62.29 337 660
Ili‘ili‘ula Stream 220358159264301 ungauged 22.04631 − 159.488 397.78 93.73 324 1593
Waiahi Stream 16057900 2015–2019 22.02504 − 159.476 1046.02 39.74 247 772

4
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Table 3
Low-flow duration characteristics (m3s− 1) for the observed data at the USGS 16068000 stream gaging station and the lower 95 % confidence interval (CI), upper 95 %
CI, best simulation, and median model result for 500 iterations of the SWAT model run at a daily interval from 1995 to 2014, following a warm-up period from 1990 to
1995. [MDF = mean daily flow; L14d = lowest 14-day mean flow; CV = coefficient of variation; BFQ50 = median base flow].
MDF Q50 Q70 Q90 Q95 Q99 L14d CV BFQ50

Observed 1.240 0.742 0.586 0.374 0.311 0.263 0.612 1.633 0.650
Modeled Lower 95 % CI 0.733 0.540 0.414 0.253 0.202 0.184 0.644 1.552 0.450
Upper 95 % CI 1.319 0.901 0.673 0.407 0.307 0.264 0.736 1.358 0.770
Best Simulation 1.029 0.680 0.500 0.297 0.228 0.208 0.673 1.487 0.560
Median Model Run 1.023 0.749 0.557 0.322 0.250 0.224 0.700 1.430 0.620

2.4. Data Analysis 3. Results

To verify that modeled daily rainfall approximated rainfall measured 3.1. Modeled versus Observed hydrological characteristics
at a specific location within a sub-basin, gridded daily rainfall were
compared to the USGS 220356159281401 1051.0 N Wailua Ditch Rain Mean daily rainfall at the USGS 220356159281401 station matched
Gage from 2001 to 2014 using RMSE, PBIAS and r2 analyses. Mean daily observed values well, with performance evaluation criteria ratings of
flow (MDF), median (Q50) flow and low-flow duration discharge sta­ good or very good (RMSE = 9.34; r2 = 0.87; PBIAS = -16.3 %) and es­
tistics (Q70 to Q99) were calculated for each sub-basin for the model timates of gridded daily rainfall was similarly effective during the dry
period (Poff et al., 1997). The lowest 14-day average flow (L14d) was season (r2 = 0.83) as the wet season (r2 = 0.87). Although modeled
determined for the model period by using a running average (Olden and rainfall generally underestimated actual rainfall at this location, the
Poff, 2003). Minimum monthly flow was also calculated to compare large elevation variability, ranging from 335 m to 635 m to the south
variation in minimum flow conditions over time. The median base flow interfluve (260 m linear distance) and to 650 m to the north interfluve
was determined by first partitioning mean daily flow into daily base flow (710 m linear distance) confounds the applicability of a single rainfall
and runoff through a statistical base flow separation procedure using the station to the mean daily rainfall for an entire grid. Thus, a 250 m grid is
Eckhardt filter method (Lim et al., 2005). Flow statistics were similarly not expected to perfectly predict rainfall for a single point but may still
calculated for wet (November to April) and dry (May to October) sea­ accurately represent rainfall averaged over the entire 250 m grid. Using
sons. Coefficient of variation (CV) was calculated as the ratio of the daily rainfall available from the single rainfall station, the SWAT model
standard deviation (SD) to the mean. MDF yield was calculated by results were considered not satisfactory (NSE = 0.09; RMSE = 3.01; r2 =
normalizing flow values by catchment area (supplemental Fig. 4). 0.65; PBIAS = -18.3 %).
Additional historic (1914–1985) streamflow data were available on At the calibration station (USGS 16068000), MDF for the single best
the North Fork Wailua River (NF Wailua) at USGS 16063000 below the model had a good agreement rating with observed values (NSE = 0.73;
confluence of three tributaries. Modeled flow duration statistics for this RMSE = 1.0560; r2 = 0.75; PBIAS = 17.02 %), however, the median
site were compared to observed statistics based on Cheng (2016). Model model result more closely matched the observed low-flow characteris­
results from 1995 to 2014 period were also compared to low-flow tics (Table 3), while still maintaining satisfactory agreement between
duration discharge estimates for the 1961–2019 period provided by overall observed and modeled output (NSE = 0.69; RMSE = 1.1198; r2
Cheng (2020) for each sub-basin based on partial-record measurements = 0.72; PBIAS = 17.47 %) for the 1995–2014 period. As expected,
and record augmentation. modeled flows underestimated peak MDF compared to observed flows,
resulting in a distribution of flows with greater negative skew (Fig. 2).
2.5. Comparisons to previous records of water diverted for hydropower Despite strong correlations and low RMSE, at an annual time scale, the
model performance was not satisfactory using either the best simulation
The North Wailua Ditch transports water diverted from the Wai‘a­ (NSE = -0.156; RMSE = 0.232; PBIAS = 17.0 %; r2 = 0.80) or the median
le‘ale, Waikoko, and Ili‘ili‘ula streams. USGS station 160,612,000 on the model result (NSE = -0.208; RMSE = 0.2374; r2 = 0.80; PBIAS = 17.47
North Wailua Ditch below the Waikoko diversion dam was active until %). This was due to the model consistently underestimating the
2002, providing a record of daily ditch flow from the combined diverted magnitude of high flow events.
flows of Wai‘ale‘ale and Waikoko streams. However, this record does not At USGS 16071500, the model performed satisfactorily (NSE = 0.65;
indicate when the intakes were plugged by debris or when the ditch was RMSE = 0.0777; r2 = 0.69), capturing some of the magnitude and
shut down for maintenance (i.e., resulting in less ditch flow despite the variability of streamflow for the 1995–2014 period, however, the model
availability of streamflow); rather it reflects the amount diverted during was not satisfactory based on PBIAS (23.8 %). In this smaller, lower
periods below ditch capacity and without operational constraints (e.g., elevation watershed, the model consistently over-represented the min­
when the diversions are operational). While the maximum designed imum annual flow (mean difference − 73.5 %), although it successfully
capacity of the ditch is 1.3144 m3s− 1 (30 million gallons per day) at the estimated the mean annual flow (mean difference = -2.7 %).
penstock (SSFM International, 2019), under real-world conditions at
USGS 160612000, the capacity is approximately 1.02 m3s− 1. Assuming a
ditch efficiency of 80 % for diverted flow rates typical of older irrigation 3.2. Modeled streamflow in headwater tributaries
systems, comparisons between diverted flow at USGS 160612000 and
80 % of modeled flow summed from Wai‘ale‘ale and Waikoko can be In the headwater streams, MDF for the 1995 to 2014 period varied
made for the overlapping period of record (1995–2002). Significant from 0.722 m3s− 1 in Wai‘ale‘ale, to 0.610 m3s− 1 in Ili‘ili‘ula, and 0.369
differences in mean daily flow values between modeled and actual ditch m3s− 1 in Waikoko. Median flow from 1995 to 2014 was approximately
flow at USGS 16061200 were tested with the non-parametric Mann- 30 % less than MDF (Table 4). Median base flow (BFQ50) was about 60 %
Whitney U test. Mean daily ditch flow available below the diversion on of observed Q50 at USGS 16068000, 57 % of Q50 in Wai‘ale‘ale, 54 % of
Ili‘ili‘ula Stream was then estimated for flows below ditch capacity Q50 in Waikoko, 77 % of Q50 in Ili‘ili‘ula, and 80.5 % of Q50 in Waiahi,
(1.3144 m3s− 1). indicating the relatively high contribution of groundwater to surface
flow. Observed MDF yield at USGS 16068000 (7.63 mm) was similar in
magnitude to MDF yield at NF Wailua (USGS 16063000), and as ex­
pected, less than modeled values in the headwater streams (Table 4).

5
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Table 4
Modeled low-flow duration characteristics (m3s− 1) for five sub-basins for the 1995–2014 model period in Wailua, Kaua‘i. [MDF = mean daily flow; Y = yield (mm);
L14d = lowest 14-day mean flow; CV = coefficient of variation; BFQ50 = median base flow].
location MDF MDF Y Q50 Q70 Q90 Q95 Q99 L14d CV BFQ50

Wai‘ale‘ale 0.722 0.0086 0.491 0.373 0.275 0.232 0.156 0.099 1.211 0.410
Waikoko 0.369 0.0022 0.256 0.184 0.120 0.098 0.070 0.045 1.187 0.200
Ili‘ili‘ula 0.610 0.0011 0.429 0.306 0.202 0.165 0.112 0.073 1.158 0.330
Waiahi 1.272 0.0011 1.006 0.760 0.512 0.416 0.285 0.197 0.946 0.810
NF Wailua at 198 m 0.840 0.0024 0.668 0.495 0.338 0.283 0.216 0.147 0.989 0.550

There were consistent trends in mean annual baseflow that mirrored 3.4. Total water available for hydropower
total flow among tributaries over time, with differences driven by
catchment area and mean annual rainfall (Fig. 3). MDF yield was Daily mean ditch flows using modeled streamflow were not statisti­
greatest in Wai‘ale‘ale and Ili‘ili‘ula, followed by Waikoko and Waiahi cally different from observed ditch flows between 1995 and 2002 (U =
headwater streams (Table 4). This was also true for BFQ50 yields (not 3772776, p = 0.23). Low flow (Q90) was approximately 60 % of median
shown). The L14d flow was lowest in Waikoko and greatest in Waiahi. At ditch flow for both observed and modeled values (Table 6). Below the
median flow, Wai‘ale‘ale Stream contributed 86 % of the flow at NF contribution from Ili‘ili‘ula Stream to the ditch, one standard deviation
Wailua. above modeled mean total ditch flow (1.23 m3s− 1) was also consistent
From 1995 to 2014 minimum mean monthly flow ranged from 0.090 with estimates of maximum penstock capacity of approximately 1.10
m3s− 1 to 0.768 m3s− 1 for Wai‘ale‘ale Stream, from 0.065 m3s− 1 to 0.603 m3s− 1 (KIUC, pers. comm).
m3s− 1 for Ili‘ili‘ula Stream, from 0.040 m3s− 1 to 0.374 m3s− 1 for Wai­
koko Stream and from 0.181 m3s− 1 to 1.317 m3s− 1 for Waiahi Stream 3.5. Environmental flows and water available for hydropower
(Fig. 4). Over this same period, monthly coefficient of variation ranged
from 0.484 to 6.254 for Wai‘ale‘ale Stream, from 0.475 to 5.551 for Water available for hydropower under various e-flow scenarios from
Ili‘ili‘ula Stream, from 0.484 to 5.378 for Waikoko Stream, and from each tributary is provided in Table 7. The most protective e-flow sce­
0.551 to 6.054 for Waiahi Stream. nario resulted in no water available for hydropower during very low-
flow conditions (<Q90) but still approximately 1.1 m3s− 1 of water
3.3. Seasonal trends in streamflow characteristics available for hydropower at mean flow and 0.57 m3s− 1 at Q50.

In some water years, there was little difference in mean seasonal


4. Discussion
flow, and dry season flow occasionally exceeded wet season flow
(Fig. 5). Modeled flow underestimated observed wet season MDF at
Run-of-river hydropower can utilize relatively small amounts of
USGS 16068000 by 31 %, but only underestimated observed dry season
water taken collectively at high elevations to generate power using steep
MDF by 11 %. In the headwater streams, MDF was 10–40 % less in the
topographic gradients naturally found in the landscape. Such a system
dry season than the wet season, while BFQ50 did not vary much between
often combines diverted water from several smaller streams before the
seasons (Table 5). Compared to the wet season, mean daily flow in the
cumulative flow enters a forebay. However, dewatering streams can
dry season declined the most in Waikoko (12.0 %), followed by Ili‘ili‘ula
affect the viability of native aquatic biota, aesthetic and recreational
(9.7 %), Wai‘ale‘ale (9.6 %), and Waiahi (9.6 %). In some tributaries, the
uses, and traditional and customary practices that take place down­
L14d flow was lower in the wet season than the dry season, further
stream (Rosero-Lópex et al., 2020; Strauch et al., 2022). Improved un­
reinforcing the unpredictability of low flow conditions.
derstanding of the temporal and spatial variability of surface water is
critical for making water management decisions and reliable data are

Fig. 3. Modeled mean annual flow (top) (m3s− 1) and mean annual baseflow (bottom) from 1995 to 2014 for four headwater streams in Wailua, Kaua‘i.

6
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Fig. 4. Monthly minimum flow (m3s− 1) from 1995 to 2014 for four headwater streams in Wailua, Kaua‘i.

Fig. 5. Mean wet (November to April) and dry (May to October) season streamflow for three headwater tributaries by water year. (note: water year 2014 only
includes Oct to Dec 2014 data).

needed to evaluate the implications of water management decisions on 4.1. Challenges with modeling tropical watersheds
downstream flows (Strauch, 2020). Modeled estimates of streamflow for
headwater streams can be used to characterize flow regimes, understand Hydrological models provide important insights for managing water
the consequences of flow restoration to downstream reaches, and the resources when there is a lack of reliable data. However, large variations
availability of water for hydropower. in rainfall (Frazier et al., 2018; Timm & Diaz, 2009; Tomasella et al.,
2008), solar radiation (Longman et al., 2014), evapotranspiration
(Giambelluca et al., 2014), cloud water interception (Juvik & Nullet,

7
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Table 5
Wet (November to April) and dry (May to October) seasonal low-flow characteristics (m3s− 1) for observed and modeled locations from 1995 to 2014 in Wailua, Kaua‘i.
[MDF = mean daily flow; Y = yield (mm); L14d = lowest 14-day mean flow; CV = coefficient of variation; BFQ50 = median base flow].
Wet Season

location MDF Q50 Q70 Q90 Q95 Q99 L14d CV BFQ50

USGS 16068000 observed 1.537 0.838 0.637 0.432 0.379 0.297 0.263 1.147 0.590
USGS 16068000 modeled 1.194 0.839 0.602 0.401 0.319 0.236 0.224 0.814 0.570
Wai‘ale‘ale 0.759 0.485 0.378 0.267 0.210 0.134 0.099 1.041 0.400
Waikoko 0.393 0.255 0.182 0.116 0.088 0.064 0.045 0.976 0.190
Ili‘ili‘ula 0.641 0.414 0.296 0.192 0.145 0.102 0.073 0.960 0.330
Waiahi 1.337 1.007 0.748 0.483 0.368 0.247 0.197 0.779 0.790
NF Wailua at 198 m 0.891 0.694 0.524 0.337 0.266 0.199 0.147 0.826 0.510
Dry Season

location MDF Q50 Q70 Q90 Q95 Q99 L14d CV BFQ50

USGS 16068000 observed 0.948 0.680 0.547 0.419 0.371 0.314 0.288 1.708 0.740
USGS 16068000 modeled 0.855 0.672 0.529 0.383 0.325 0.266 0.249 1.624 0.700
Wai‘ale‘ale 0.686 0.497 0.369 0.281 0.250 0.193 0.141 1.331 0.420
Waikoko 0.346 0.259 0.186 0.122 0.104 0.080 0.068 1.323 0.200
Ili‘ili‘ula 0.579 0.438 0.318 0.209 0.181 0.130 0.109 1.295 0.330
Waiahi 1.208 1.005 0.769 0.532 0.458 0.338 0.253 1.061 0.830
NF Wailua at 198 m 0.789 0.640 0.465 0.339 0.298 0.244 0.206 1.096 0.580

1995), and underlying geology (Gingerich, 1999) add complexity to


Table 6 modeling watershed hydrology in the tropics (Cuo et al., 2006; Strauch
Low-flow (m3s− 1) duration characteristics for the North Wailua Ditch below
et al., 2017). This is especially true on young volcanic islands where
Waikoko Stream diversion at USGS 16061200 and the modeled combined flow
stream channel incision dissects varied geology that influences
of Wai‘ale‘ale and Waikoko Streams from 01 Jan 1995 to 29 Sept 2002; total
modeled ditch flow based on the contributions of the three main headwater groundwater-surface water interactions and where daily rainfall is both
tributaries from 01 Jan 1995 to 29 Sept 2002. extremely variable and unpredictable (Gingerich & Engott, 2012).
Rainfall-driven runoff events dominate surface flow of low-order trop­
USGS 16061200 North Wailua North Wailua Ditch at
Ditch below Waikoko Stream penstock
ical streams, with frequent extreme events driving steep changes in
Intake hydrographs (Craig, 2003), while simultaneously, the underlying hy­
parameter modeled flow observed flow total modeled flow
drogeology drives important groundwater contributions to base flows
(Bassiouni and Oki, 2012).
Mean (±SD) 0.583 0.586 0.876 (±0.358)
In this application, the SWAT model accurately depicted saturation-
flow (±0.358) (±0.150)
Q50 0.553 0.595 0.806 excess runoff, which is ideal for wet tropical forests as both soil satu­
Q70 0.464 0.530 0.644 ration and shallow groundwater saturation leads to the generation of
Q90 0.319 0.396 0.405 faster surface runoff. Use of a physically-based model has certain ad­
Q99 0.174 0.157 0.216 vantages over previously developed hydrologic models, especially in
[SD = standard deviation]. heterogeneous tropical environments (Akoko et al., 2021). However, the
SWAT model consistently under-predicted peak discharge of storm
events and over-predicted the recession limb of the hydrograph. This is
not unusual, as hydrologic models often have issues with predicting

Table 7
Water available for hydropower for various flow regime values (m3s− 1) and potential instream flow standard scenarios: 20 % of median baseflow (20 %BFQ50); 64 % of
median baseflow (64 %BFQ50); 100 % of median baseflow (100 %BFQ50) for three headwater streams in Wailua, Kauai. [MDF = mean daily flow; L14d = lowest 14-day
mean flow; BFQ50 = median base flow].
stream flow characteristic

Wai‘ale‘ale MDF Q50 Q70 Q90 Q95 Q99 L14d BFQ50

Natural flow 0.722 0.491 0.373 0.275 0.232 0.156 0.099 0.410
20 %BFQ50 0.640 0.409 0.291 0.193 0.150 0.074 0.017 0.328
50 %BFQ50 0.517 0.286 0.168 0.070 0.027 0.000 0.000 0.205
64 %BFQ50 0.460 0.228 0.111 0.012 0.000 0.000 0.000 0.148
Waikoko MDF Q50 Q70 Q90 Q95 Q99 L14d BFQ50
Natural flow 0.369 0.256 0.184 0.120 0.098 0.070 0.045 0.200
20 %BFQ50 0.329 0.216 0.144 0.080 0.058 0.030 0.005 0.160
50 %BFQ50 0.269 0.156 0.084 0.020 0.000 0.000 0.000 0.100
64 %BFQ50 0.241 0.128 0.056 0.000 0.000 0.000 0.000 0.072
Ili‘ili‘ula MDF Q50 Q70 Q90 Q95 Q99 L14d BFQ50
Natural flow 0.610 0.429 0.306 0.202 0.165 0.112 0.073 0.330
20 %BFQ50 0.544 0.363 0.240 0.136 0.099 0.046 0.007 0.264
50 %BFQ50 0.445 0.264 0.141 0.037 0.000 0.000 0.000 0.165
64 %BFQ50 0.399 0.217 0.095 0.000 0.000 0.000 0.000 0.119
Waiahi MDF Q50 Q70 Q90 Q95 Q99 L14d BFQ50
Natural flow 1.272 1.006 0.760 0.512 0.416 0.285 0.197 0.810
20 %BFQ50 1.110 0.844 0.598 0.350 0.254 0.123 0.035 0.648
50 %BFQ50 0.867 0.601 0.355 0.107 0.011 0.000 0.000 0.405
64 %BFQ50 0.753 0.488 0.242 0.000 0.000 0.000 0.000 0.292

8
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

runoff flow conditions (Cuo et al., 2006; Strauch et al., 2017). This is making sound management decisions. Flow from Wai‘ale‘ale, Waikoko,
expected to occur in steep gradient watersheds where rainfall at daily ʻIliʻiliʻula, and Waiahi headwater streams each contribute water for run-
time steps is not likely to be representative of real-world conditions in of-river hydropower. In this study, there was much agreement between
the tropics (Biasutti and Yuter, 2013). Daily rainfall under-represents modeled and observed ditch flows at USGS 16061200 from 1995 to
the intensity of sub-daily rainfall and hydrological models will regu­ 2002. Total modeled flow available at the penstock of the Upper Waiahi
larly under predict the recession limb of the hydrograph (Cuo et al., Hydropower plant also closely agreed with data from 1936 to 1938 re­
2006). ported by Lı̄hu‘e Plantation with respect to Q50 (0.826 m3s− 1), but not
In places with limited data and access, hydrological modeling can Q90 (0.772 m3s− 1) (Territorial Planning Board, 1939). By identifying the
provide useful information that would otherwise not exist for meeting amount of water that can be reasonably obtained to generate hydro­
resource management decisions. The accuracy of hydrological model power, accurate estimates of the consequences of e-flows can be made.
outputs largely relies on the certainty of rainfall data (Guan et al., 2020). In addition to highly variable rainfall at daily intervals, the season­
The accuracy of rainfall estimates is critical to determine the amount ality and interannual variability of rainfall in the humid tropics com­
and timing of runoff, but the spatial distribution of rainfall stations in plicates the utilization of water resources (Strauch, 2020). Diaz &
the tropics is often poor (Wohl et al., 2012; Longman et al., 2018). High- Giambelluca (2012) found that typical dry season (May–October) con­
resolution daily rainfall products will improve the calibration of models ditions have a characteristic pattern across time scales in Hawai‘i: drier
to characterize streamflow regimes in data scarce regions in the tropics than normal periods have a strengthening of the trade winds to the south
where a lack of sufficient monitoring of hydrological processes has of the islands; while conversely, wetter than normal periods have
limited the development of e-flows (Wohl et al., 2012; Hairan et al., generally weaker trades to the south with a tendency for southwesterly
2021). Compared to rainfall data derived from a single station within the storms. The inconsistency and unpredictability of rainfall suggests that
watershed, the gridded daily rainfall product increased the accuracy of management decisions focus on investments in infrastructure that cap­
model output from not satisfactory to good. Such high-resolution data ture larger runoff events for peak power production. While there were
reduces the inaccuracies associated with various methods for estimating clear patterns in surface flow between seasons, reflecting the general
rainfall across highly varied topography using point data at widely increase in water availability during the wet season, seasonality was not
dispersed rainfall stations (Mair and Fares, 2011). always apparent, and runoff events often occurred throughout the year.
Other methods may be utilized to estimate low-flow duration char­ Extreme low-flow conditions were often lower in the wet season than the
acteristics (Vogel and Kroll, 1991). For example, Cheng (2020) used the dry season, further reinforcing the unpredictability rainfall (Feng et al.,
Maintenance of Variance.1 model to estimate low-flow duration 2013).
streamflow values in Southeast Kaua‘i and compared to the results
presented here, estimated Q50 as 0.736 m3s− 1 (43.4 % more), 0.210 4.4. Predicting surface water availability to support resource management
m3s− 1 (19.2 % less), and 0.311 m3s− 1 (28 % less), for Wai‘aleale, Wai­ decisions in a changing climate
koko, and Ili‘ili‘ula, respectively, and Q90 as 0.510 m3s− 1 (88.9 % more),
0.079 m3s− 1 (31.3 % less), 0.224 m3s− 1 (14.9 % more), respectively. Climate change adds additional uncertainty to predicting water
Generating daily flow values through hydrologic modeling at multiple availability for off-stream use in the tropics, especially in regions ex­
sub-basins simultaneously has additional advantages for understanding pected to experience shifts in the distribution of rainfall (Timm et al.,
the natural flow regime. 2015; Milly et al., 2005). Predicted increases in air and ocean temper­
atures for the eastern Pacific Region (Ruosteenoja et al., 2003), coupled
4.2. Low-flow characteristics across headwater basins with localized decreases in rainfall (Mimura et al., 2007; Timm and Diaz,
2009) are expected to affect seasonal patterns in precipitation (Mora
Accurate estimates of flow regimes at existing or proposed stream et al., 2015; Timm et al., 2013). There has been a consistent downward
diversions are critical to determining appropriate e-flows or for evalu­ trend in total rainfall coupled with an increase in drought probability for
ating the economics of water infrastructure, such as new dams or most of Hawai‘i, which makes estimation of low-flow characteristics
municipal water sources. Variability in rainfall in the humid tropics has more urgent (Chu et al., 2010). Further, downscaling projections for
amplified consequences for runoff and surface flow (Mair and Fares, Hawai‘i suggest 5 to 40 % future declines in rainfall depending on the
2010), especially in headwater streams where rapid changes in elevation season, climate scenario, and island, with a growing contrast between
and aspect influence orographic rainfall (Cao et al., 2007; Christian windward and leeward regions (Timm et al., 2015). The impact of water
et al., 2019). Further, groundwater contributions to surface flow from available for off-stream uses (e.g., agriculture, hydropower, drinking
saturated geologic features may contribute substantially to flow re­ water supply) is most critical under drought conditions, when proposed
gimes. At higher elevation tributaries, base flow yield was greatest in e-flows will have the greatest impact. Clarifying the consequences of
Waiahi stream followed by the Wai‘ale‘ale stream. By contrast, at a e-flow establishment is critical to effectively balance resource manage­
lower elevation (USGS 16063000), the stream cuts through the thickly ment (Strauch, 2020).
saturated geology of the Kōloa Volcanic Series, influencing surface Climate change is compounding water resource problems by exas­
water-ground water interactions (Izuka et al., 2015; Izuka & Gingerich, perating extreme rainfall and drought events, increasing the unpre­
1998). Overall, this complexity reflects the much older substrate age, dictability of flow regimes. In Hawai‘i, as much as 70 % of the year
stream channel incision, and erosion of the Wailua watershed relative to experiences subsidence inversion between 1200 m and 2400 m above
younger volcanic islands where runoff contributes a greater proportion sea level, reducing the vertical movement of trade winds (Blumenstock
to MDF (Strauch et al., 2015). and Price, 1967). This contributes to the spatial distribution of localized
climate, specifically the elevation band of the fog drip zone (usually
4.3. Estimating the availability of water for hydropower 1200–1800 m) and the transition zone (usually 1800–2400 m), both of
which contribute to the water budget (Giambelluca and Nullet, 1991).
Hydropower is becoming an increasingly important component of The strength and frequency of the trade wind inversion has shifted with
the energy portfolios of islands due to the high cost of importing fossil climate change, resulting in an average decline in rainfall of 6 % in the
fuels and the need for more carbon–neutral energy generation strategies dry season and 31 % in the wet season at higher elevations (Longman
(Kuang et al., 2016; Shukrullah and Naz, 2021). For run-of-river hy­ et al., 2015). Long-term declines in rainfall and fog drip are expected to
dropower, quantifying the availability of base flows in streams is critical reduce groundwater recharge and have been attributed to subsequent
for determining the appropriate size of infrastructure. Since reservoirs declines in stream baseflows, making estimates of water available for
are not involved, the quantification of flow regimes, is critical for off-stream use based on historic records challenging (Bassiouni & Oki,

9
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

2012). In three of the four headwater tributaries, we observed declines resources depends on confidence in the consequences of management
in mean annual flow and mean annual baseflow for the 1995–2014 decisions. The availability of high-resolution rainfall and temperature
period (supplemental Fig. 3). Shifts in the seasonality of rainfall gener­ data reduced model uncertainty, fulfilling a critical gap in data avail­
ally reflect larger scale patterns in ocean circulation (e.g., Pacific ability needed to make reasonable management decisions to provide for
Decadal Oscillation, El Nino Southern Oscillation, Southern Oscillation both environmental flows and hydropower production.
Index). The use of high-resolution data inputs can greatly improve hy­
drologic modeling by better characterizing rainfall inputs (Biasutti and CRediT authorship contribution statement
Yuter, 2013; Yi et al., 2021). Temporally and spatially coarse rainfall
inputs limit the usefulness of modeling for small, highly heterogeneous Ayron M. Strauch: Conceptualization, Formal analysis, Funding
catchments, especially in the tropics (Tomlinson and De Carlo, 2003; acquisition, Methodology, Project administration, Visualization,
Strauch et al., 2017). In comparison to model outputs when rainfall was Writing – original draft, Writing – review & editing. Yu-Fen Huang:
represented by a single station, model uncertainty was greatly reduced Data curation, Methodology, Software, Validation, Writing – review &
with the higher resolution rainfall data. editing. Yin-Phan Tsang: Conceptualization, Investigation, Project
administration, Resources, Supervision, Validation, Writing – review &
4.5. Complexities involved with the implementation of environmental editing.
flows
Declaration of competing interest
Freshwater resources are increasingly threatened by groundwater
and surface water extraction, often with little regard to freshwater The authors declare that they have no known competing financial
ecosystems (Cooke et al., 2022). Accurate estimates of flow regimes for interests or personal relationships that could have appeared to influence
headwater streams can be used to implement e-flows that improve the work reported in this paper.
protection of aquatic communities (Acreman, 2016). Environmental
flows are necessary to protect instream values, however, the extent to Data availability
which any particular flow regime meets the recognized need for an e-
flow is not always quantified. For example, Gingerich and Wolf (2005) Data will be made available on request.
estimated the flow needed to provide sufficient habitat downstream of
low-head dams on Maui Island, but minimum flows do not necessarily Acknowledgements
equate to idealized habitat conditions that support all stages of aquatic
life cycles. For amphidromous species in Hawai‘i, sufficient habitat may The authors thank the University of Hawai‘i Department of Natural
be found at elevations below dams due to groundwater gains in Resources and Environmental Management and the State of Hawai‘i
streamflow, but reproduction may be dependent on the frequency of Commission on Water Resource Management for providing support to
freshet events, which may be muted by upstream diversions (Gingerich the authors.
and Wolff, 2005). Further, additional flow releases to support floodplain
connectivity or estuary health may improve recruitment of stream fauna Appendix A. Supplementary data
to nearshore environments (Jakubínský et al., 2021). While hydrological
connectivity may improve the movement of species among habitats, a Supplementary data to this article can be found online at https://doi.
singular e-flow value may not be sufficient to meet the behavioral needs org/10.1016/j.jhydrol.2024.131287.
of all biota (Arantes et al., 2019). Enhancements in hydrological and
ecological monitoring are likely to support improved management of References
water resources, especially in remote tropical islands (Dobriyal et al.,
2017). Abbaspour, K.C., Yang, J., Maximov, I., Siber, R., Bogner, K., Mieleitner, J.,
Srinivasan, R., 2007. Modelling hydrology and water quality in the pre-alpine/alpine
In past decisions, the State of Hawai‘i has established instream flow
Thur watershed using SWAT. J. Hydrol. 333 (2–4), 413–430. https://doi.org/
standards that are designed to (1) connect upstream and downstream 10.1016/J.JHYDROL.2006.09.014.
habitat (estimated at 20 % of BFQ50) or (2) protect 90 % of instream Acreman, M., 2016. Environmental flows—basics for novices. WIREs Water 3 (5),
622–628.
habitat (estimated at 64 % of BFQ50). The implementation of these
Akoko, G., Le, T.H., Gomi, T., Kato, T., 2021. A review of SWAT model application in
scenarios can be applied to each stream’s respective time-series data to Africa. Water 13 (9), 1313. https://doi.org/10.3390/w13091313.
examine the availability of water for hydropower. The consequences are Andierson, D., Moggridge, H., Warren, P., Shucksmith, J., 2014. The impacts of ‘run-of-
particularly noteworthy under low-flow situations, which are expected river’ hydropower on the physical and ecological condition of rivers. Water and
Environment Journal 29, 268–276.
to become more frequent and severe due based on trends in drought Arantes, C.C., Fitzgerald, D.B., Hoeinghaus, D.J., Winemiller, K.O., 2019. Impacts of
(Frazier et al., 2022). hydroelectric dams on fishes and fisheries in tropical rivers through the lens of
functional traits. Curr. Opin. Environ. Sustain. 37, 28–40.
Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic
5. Conclusion modeling and assessment part I: model development. J. Am. Water Resour. Assoc. 34
(1), 73–89. https://doi.org/10.1111/j.1752-1688.1998.tb05961.x.
Managing for the continued capacity of tropical landscapes to pro­ Arthington, A.H., Pusey, B.J., 2003. Flow restoration and protection in Australian rivers.
River Res. Appl. 19 (5–6), 377–395.
vide water for terrestrial, aquatic, and near-shore environments is a Bassiouni, M., Oki, D.S., 2012. Trends and shifts in streamflow in Hawai‘i. Hydrol.
natural resource challenge for island systems due to the high variability Process. 27 (10), 1484–1500.
in elevation, slope, aspect, geology, and rainfall across relatively small Biasutti, M., Yuter, S.E., 2013. Observed frequency and intensity of tropical precipitation
from instantaneous estimates. J. Geophys. Res. Atmos. 118 (17), 9534–9551.
catchments and aquifers. Run-of-river hydropower can provide renew­ Blumenstock, D.I., Price, S. 1967. Climate of Hawaii. Climate of the States. Climate of the
able energy without the need for large reservoirs. Investments in hy­ States No 60-51. U.S. Department of Commerce.
dropower in tropical islands to reduce dependence on imported fossil Territorial Planning Board. 1939. Summary of Records: Surface Water Resources of the
Territory of Hawai‘i, 1901-1938. Territory of Hawai‘i: Honolulu Star-Bulletin,
fuels is challenging due to the uncertainty of water supply. Further,
Honolulu.
water resource management decisions are often constrained by a lack of Buytaert, W., Celleri, R., Willems, P., De Bievre, B., Qyseure, G., 2006. Spatial and
hydrological data, which may be alleviated with improved modeling temporal rainfall variability in mountainous areas: A case study from the south
efforts. We used hydrologic modeling to estimate flow regimes in remote Ecuadorian Andes. J. Hydrol. 329, 413–421.
Cai, X., Wallington, K., Shafiee-Jood, M., Marston, L., 2018. Understanding and
locations that lack long-term continuous records to improve the man­ managing the food-energy-water nexus—opportunities for water resources research.
agement and utilization of water resources. Equitable management of Adv. Water Resour. 111, 259–273.

10
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Cao, G., Giambelluca, T.W., Stevens, D.E., Schroeder, T.A., 2007. Inversion variability in Gupta, H.V., Sorooshian, S., Yapo, P.O., 1999. Status of automatic calibration for
the Hawaiian trade wind regime. J. Clim. 20, 1145–1160. hydrologic models: comparison with multilevel expert calibration. J. Hydrol. Eng. 4
Cheng, C.L., 2016. Low-flow characteristics for streams on the Islands of Kaua‘i, O‘ahu, (2), 135–143.
Moloka‘i, Maui, and Hawai‘i, State of Hawai‘i. Scientific Investigations Report Hairan, M.H., Jamil, N.R., Azmai, M.N.A., Looi, L.J., Camara, M., 2021. Environmental
2016–5103. flow assessment of a tropical river system using hydrological index methods. Water
Cheng, C.L., 2020. Low-flow characteristics of streams from Wailua to Hanapēpē. US 13 (18), 2477. https://doi.org/10.3390/w13182477.
Geol. Surv. Sci. Investig. Rep. 2020–5128. Hassell, D., Gregory, J., Blower, J., Lawrence, B.N., Taylor, K.E., 2017. A data model of
Christian, J., Martin, J., McKay, S.K., Chappell, J., Pringle, C.M., 2019. Building a the climate and forecast metadata conventions (CF-1.6) with a software
hydrologic foundation for tropical watershed management. PLoS One 14 (3), implementation (cf-python v2.1). Geosci. Model Dev. 10, 4619.
e0213306. Hodson, T.O., 2022. Root-mean-square error (RMSE) or mean absolute error (MAE):
Chu, P.-S., Chen, Y.R., Schroeder, T.A., 2010. Changes in precipitation extremes in the When to use them or not. Geosci. Model Dev. 15, 5481–5487. https://doi.org/
Hawaiian Islands in a warming climate. J. Clim. 23, 4881–4900. 10.5194/gmd-15-5481-2022.
Cooke, S.J., Frempong-Manso, A., Piczak, M.L., Karathanou, E., Clavijo, C., Ajagbe, S.O., Hofstra, N., Haylock, M., New, M., Jones, P., Frei, C., 2008. Comparison of six methods
Akeredolu, E., Strauch, A.M., Piccolo, J., 2022. A freshwater perspective on the for the interpolation of daily, European climate data. J. Geophys. Res. 113, D21110.
United Nations decade for ecosystem restoration. Conservation Science and Practice, Izuka S.K., Gingerich, S.B. 1998. Ground Water in the Southern Lihue Basin, Kauai,
p. e12787. Hawaii. U.S. Geological Survey Water-Resources Investigation Report, (98–4031).
Craig, D.A., 2003. Geomorphology, development of running water habitats, and Izuka, S.K., Engott, J.A., Rotzoll, K., Bassiouni, M., Johnson, A.G., Miller, L.D., Mair, A.
evolution of black flies on Polynesian Islands. Bioscience 53, 1079–1093. 2018. Volcanic aquifers of Hawai‘i—Hydrogeology, water budgets, and conceptual
Cuo, I., Giambelluca, T.W., Ziegler, A.D., Nullet, M., 2006. Use of the distributed models (ver. 2.0). USGS. 2015-5164.
hydrology soil vegetation model to study road effects on hydrological processes in Izuka, S. (2006). Effects of Irrigation, Drought, and Ground-Water Withdrawals on
Pang Khum Experimental Watershed, northern Thailand. For. Ecol. Manage. 224, Ground-Water Levels in the Southern Lihue Basin, Kauai , Hawaii. Scientific
81–94. Investigations Report, (2006–5291).
De Cicco, L. A., Lorenz, D., Hirsch, R. M., & Watkins, W. (2018). dataRetrieval: R packages Jakubínský, J., Prokopová, P., Raška, P., Salvati, L., Bezak, N., Cudlín, O., Cudlín, P.,
for discovering and retrieving water data available from U.S. federal hydrologic web Purkyt, J., Vezza, P., Camporeale, C., Daněk, J., Pástor, M., Lepeška, T., 2021.
services. https://doi.org/10.5066/P9X4L3GE. Managing floodplains using nature-based solutions to support multiple ecosystem
Diaz, H.F., Giambelluca, T.W., 2012. Changes in atmospheric circulation patterns functions and services. WIREs. Water 8L e1545. https://doi.org/10.1002/
associated with high and low rainfall regimes in the Hawaiian Islands region on wat2.1545.
multiple time scales. Global Planet. Change 98–99, 97–108. Juvik, J.O., Ekern, P.C. 1978. A Climatology of Mountain Fog on Mauna Loa Hawai‘i
Dobriyal, P., Badola, R., Tuboi, C., Hussain, S.A., 2017. A review of methods for Island. Technical Report No. 118. Water Resources Research Center, University of
monitoring streamflow for sustainable water resource management. Appl Water Sci Hawaii at Manoa.
7, 2617–2628. Juvik, J.O., Nullet, D., 1995. Relationships between rainfall, cloud-water interception,
Ezcurra, E., Barrios, E., Excurra, P., Ezcurra, A., Vanderplank, S., Vidal, O., Villanueva- and canopy throughfall in a Hawaiian montane forest. In: Hamilton, L.S., Juvik, J.O.,
Almanza, L., Aburto-Oropeza, O., 2019. A natural experiment reveals the impact of Scatena, F.N. (Eds.), Tropical montane cloud forests. Springer-Verlag, New York,
hydroelectric dams on the estuaries of tropical rivers. Science. Advances 5, pp. 165–182.
eacu9875. Kuang, Y., Zhang, Y., Zhou, B., Li, C., Cao, Y., Li, L., Zeng, L., 2016. A review of
Falinski, K., Penn, D., 2018. Loss of reservoir capacity through sedimentation in Hawaii: renewable energy utilization in islands. Renew. Sustain. Energy Rev. 59, 504–513.
Management implications for the twenty-first century. Pac. Sci. 72 (1), 1–19. Legates, R., David, McCabe, J., Gregory, 1999. Evaluating the use goodness-of-fit;
Feng, X., Porporato, A., Rodriguez-Iturbe, I., 2013. Changes in rainfall seasonality in the measures in hydrologic and hydroclimatic model validation. Water Resour. Res. 35
tropics. Nat. Clim. Chang. 3, 811–815. (1), 233–241.
Frazier, A.G., Giambelluca, T.W., 2017. Spatial trend analysis of Hawaiian rainfall 1920 Leta, O.T., El-Kadi, A., Dulai, H., Ghazal, K.A., 2016. Assessment of climate change
to 2012. Int. J. Climatol. 37 (5), 2522–2531. impacts on water balance components of Heeia watershed in Hawaii. J. Hydrol.: Reg.
Frazier, A.G., Giambelluca, T.W., Diaz, H.F., Needham, H.L., 2016. Comparison of Stud. 8, 182–197.
geostatistical approaches to spatially interpolate month-year rainfall for the Lim, K.J., Engel, B.A., Tang, Z., Choi, J., Kim, K.-S., Muthukrishnan, S., Tripathy, D.,
Hawaiian Islands. Int. J. Climatol. 36, 1459–1470. 2005. Automated web GIS based hydrograph analysis tool, WHAT. J. Am. Water
Frazier, A.G., Timm, O.E., Giambelluca, T.W., Diaz, H.F., 2018. The influence of ENSO, Resour. Assoc. 41 (6), 1407–1416.
PDO, and PNA on secular rainfall variations in Hawaii. Clim. Dyn. 51, 2127–2140. Longman, R.J., Diaz, H.F., Giambelluca, T.W., 2015. Sustained increases in lower-
Frazier, A.G., Giardina, C.P., Giambelluca, T.W., Brewington, L., Chen, Y.-L., Chu, P.-S., tropospheric subsidence over the central tropical North Pacific drive a decline in
Fortini, L.B., Hall, D., Helweg, D.A., Keener, V.W., Longman, R.J., Lucas, M.P., high-elevation rainfall in Hawaii. J. Clim. 28, 8743–8759.
Mair, A., Oki, D.S., Reyes, J.J., Yelenik, S.G., Trauernicht, C., 2022. A century of Longman, R.J., Giambelluca, T.W., Alliss, R.J., Barnes, M.L. 2014. Temporal solar
drought in Hawai‘i: Geospatial analysis and synthesis across hydrological, radiation change at high elevation in Hawai‘i. Journal of Geophysical Research:
ecological, and socioeconomic scales. Sustainability 14, 12023. https://doi.org/ Atmospheres. JD021322.
10.3390/su141912023. Longman, R.J., Giambelluca, T.W., Nullet, M.A., Frazier, A.G., Kodama, K., Crausbay, S.
Fukunaga, D.C., Cecilio, R.A., Zanetti, S.S., Oliveira, L.T., Caiado, M.A.C., 2015. D., Krushelnycky, P.D., Cordell, S., Clark, M.P., Newman, A.J., Arnold, J.R., 2018.
Application of the SWAT hydrologic model to a tropical watershed at Brazil. Catena Compilation of climate data from heterogeneous networks across the Hawaiian
125, 206–213. Islands. Nature Scientific Data 5, 180012.
Gassman, P.W., Reyes, M.R., Green, C.H., Arnold, J.G., Gassman, P.W., 2007. The soil Longman, R.J., Frazier, A.G., Newman, A.J., Giambelluca, T.W., Schanzenbach, D.,
and water assessment tool: historical development, application, and future research Kagawa-Viviani, A., Needham, H., Arnold, J.R., Clarck, M.P., 2019. High-resolution
directions invited Review Series. Retrieved from Trans. ASABE 50 (4), 1211–1250. gridded daily rainfall and temperature for the Hawaiian Islands (1990–2014).
https://www.card.iastate.edu/research/resource-and-environmental/items/asabe J. Hydrometeorol. 20, 489–508.
_swat.pdf. Lorentis, D.G., Collischom, W., Olivera, F., Tucci, C.E.M., 2010. GIS-based procedures for
Giambelluca, T.W., Nullet, D., 1991. Influence of the trade-wind inversion on the climate hydropower potential spotting. Energy 35, 4237–4243.
of a leeward mountain slope in Hawaii. Climate Res. 1, 207–216. Ly, S., Charles, C., Degre, A., 2011. Geostatistical interpolation of daily rainfall at
Giambelluca, T.W., Chen, Q., Frazier, A.G., Price, J.P., Chen, Y.L., Chu, P.-S., Eischeid, J. catchment scale: the use of several variogram models in the Ourthe and Ambleve
K., Delparte, D.M., 2013. Online rainfall atlas of Hawai‘i. Bull. Am. Meteorol. Soc. catchments, Belgium. Hydrol. Earth Syst. Sci. 15, 2259–2274.
94, 313–316. Macdonald, G.A., Davis, D.A., Cox, D.C., 1960. Geology and ground-water resources of
Giambelluca, T.W., Shuai, X., Barnes, M.L., Alliss, R.J., Longman, R.J., Miura, T., the island of Kauai. Hawaii Division of Hydrography, Bulletin, Hawaii, p. 13.
Chen, Q., Frazier, A.G., Mudd, R.G., Cuo, L., Businger, A.D., 2014. Mair, A., Fares, A., 2010. Time series analysis of daily rainfall and streamflow in a
Evapotranspiration of Hawai‘i. Final Report Submitted to the u.s. Army Corps of volcanic dike-intruded aquifer system, O‘ahu, Hawai‘i, USA. Hydrgeol. J. 19 (4),
Engineers—Honolulu District, and the Commission on Water Resource Management. 929–944. https://doi.org/10.1007/s10040-011-0740-3.
State of Hawai‘i. Mair, A., Fares, A., 2011. Comparison of rainfall interpolation methods in a mountainous
Gingerich, S.B., Engott, J.A. 2012. Groundwater availability in the Lahaina District, West region of a tropical island. J. Hydrol. Eng. 15 (4) https://doi.org/10.1061/(ASCE)
Maui, Hawai‘i. U.S. Geological Survey Scientific Investigations Report, 2012-5010. HE.1943-5584.0000330.
Gingerich, S.B., Wolff, R.H. 2005. Effects of surface-water diversions on habitat March, J.G., Benstead, J.P., Pringle, C.M., Scatena, F.N., 2003. Damming tropical island
availability for native macrofauna, northeast Maui, Hawaii. U.S. Geological Survey streams: problems, solutions, and alternatives. Bioscience 53, 1069–1078.
Scientific Investigations Report, (2005-5213). Milly, P.C.D., Dunne, K.A., Vecchia, A.V., 2005. Global pattern of trends in streamflow
Gingerich, S. B. (1999). Estimating Transmissivity and Storage Properties from Aquifer and water availability in a changing climate. Nature 438, 347–350.
Tests in the Southern Lihue Basin, Kauai, Hawaii. U.S. Geological Survey Water- Mimura, N., Nurse, I., McLean, R., Agard, J., Brifuflio, I., Lefale, P., Payet, R., Sem, G.,
Resources Investigation Report, (99–4066). 2007. Small Islands. In: Parry, M.I., Canziani, O.F., Palutikol, J.P., van der Linden, P.
Guan, X., Zhang, J., Yang, Q., Tang, X., Liu, C., Jin, J., Liu, Y., Bao, Z., Wang, G., 2020. J., Hanson, C.E. (Eds.), Climate Change 2007: Impacts, Adaptation, and
Evaluation of precipitation products by using multiple hydrological models over the Vulnerability. University Press, pp. 688–716.
Upper Yellow River Basin. China. Remote Sensing 12 (24), 4023. https://doi.org/ Mora, C., Frazier, A.G., Longman, R.J., Dacks, R.S., Walton, M.M., Tong, E.J., Sanchez, J.
10.3390/rs12244023. J., Kaiser, L.R., Stender, Y.O., Anderson, J.M., Ambrosino, C.M., Fernandez-Silva, I.,
Guo, B., Zhang, J., Xu, T., Croke, B., Jakeman, A., Song, Y., Yang, Q., Lei, X., Liao, W., Giuseffi, L.M., Giambelluca, T.W., 2015. The projected timing of climate departure
2018. Applicability assessment and uncertainty analysis of multi-precipitation from recent variability. Nature 502, 183–187.
datasets for the simulation of hydrologic models. Water 10 (11), 1611. https://doi.
org/10.3390/w10111611.

11
A.M. Strauch et al. Journal of Hydrology 636 (2024) 131287

Moriasi, D.N., Arnold, J., Van Liew, M., Bingner, R., Harmel, R., Veith, T., 2007. Model Soil Survey Staff. 2017. Natural Resources Conservation Service, United States
evaluation guidelines for systematic quantification of accuracy in watershed Department of Agriculture. https://websoilsurvey.sc.egov.usda.gov/. Accessed [03/
simulations. Trans. ASABE 50 (3), 885–900. 15/2019].
Moriasi, D.N., Gitau, M.W., Pai, N., Daggupati, P., 2015. Hydrologic and water quality SSFM International, Inc. 2019. Waiahi Hydropower Long-Term Water Lease Draft
models: Performance measures and evaluation criteria. Trans. ASABE 58 (6), Environment Assessment.
1763–1785. Strauch, A.M., 2020. Evaluating the management of a tropical reservoir: Implications of
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I climate change for water availability. Pac. Sci. 74, 115–135.
— A discussion of principles. J. Hydrol. 10 (3), 282–290. https://doi.org/10.1016/ Strauch, A.M., MacKenzie, R.A., Bruland, G.L., Giardina, C.P., 2015. Climate driven
0022-1694(70)90255-6. changes to rainfall and streamflow patterns in a model tropical island hydrological
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2011. Soil and Water Assessment system. J. Hydrol. 523, 160–169.
Tool: Theoretical Documentation, Version 2009. Texas Water Resources Institute Strauch, A.M., Giardina, C.P., MacKenzie, R.A., Heider, C., Giambelluca, T.W.,
Technical Report No. 406. Salminen, E., Bruland, G.L., 2017. Modeled effects of climate change and plant
Newman, A.J., Clark, M.P., Longman, R.J., Gilleland, E., Giambelluca, T.W., Arnold, J.R., invasion on watershed function across a steep tropical rainfall gradient. Ecosystems
2019. Use of daily station observations to produce high-resolution gridded 20, 583–600.
probabilistic precipitation and temperature time series for the Hawaiian Islands. Strauch, A.M., MacKenzie, R.A., Giardina, C.P., Bruland, G.L., 2018. Influence of
J. Hydrometeorol. 20, 509–529. declining mean annual rainfall on the behavior and yield of sediment and particulate
NOAA, National Oceanic Atmospheric Administration, 2010. Coastal Change Analysis organic carbon from tropical watersheds. Geomorphology 306, 28–39.
Program. https://coast.noaa.gov/digitalcoast/tools/lca.html. Strauch, A.M., Tingley III, R.W., Hsiao, J., Foulk, P.B., Frauendorf, T.C., MacKenzie, R.A.,
Nullet, D., Juvik, J.O., Wall, A., 1995. A Hawaiian mountain climate cross-section. Infante, D.M., 2022. Population response to connectivity restoration of high
Climate Res. 5, 131–137. elevation tropical stream reaches in Hawaii. Conservat. Sci. Practice. https://doi.
Olden, J.D., Poff, N.L., 2003. Redundancy and the choice of hydrologic indices for org/10.1111/csp2.12836.
characterizing streamflow regimes. River Res. Appl. 19 (2), 101–121. Timm, O.E., Diaz, H.F., 2009. Synoptic-statistical approach to regional downscaling of
Poff, N.L., Allan, J.D., Bain, M.B., Karr, J.R., Prestegaard, K.L., Richter, B.D., Sparks, R.E., IPCC twenty-first-centry climate projections: seasonal rainfall over the Hawaiian
Stromberg, J.C., 1997. The Natural Flow Regime: A paradigm for river conservation Islands. J. Clim. 22, 4261–4280.
and restoration. Bioscience 47 (11), 769–784. Timm, O.E., Giambelluca, T.W., Diaz, H.F., 2015. Statistical downscaling of rainfall
Prairie, Y.T., Alm, J., Beaulieu, J., Barros, N., Battin, T., Cole, J., del Giorgio, P., -changes in Hawai‘i based on the CMIO5 global model projections. Journal of
DelSontro, T., Guérin, F., Harby, A., Harrison, J., Mercier-Blais, S., Serça, D., Geophysical Research: Atmospheres 120.
Sobek, S., Bachon, D., 2017. Greenhouse gas emission from freshwater reservoirs: Timm, O.E., Takahashi, M., Giambelluca, T.W., 2013. On the relation between large-
What does the atmosphere see? Ecosystems 21, 1058–1071. scale circulation pattern and heavy rain events over the Hawaiian Islands: Recent
Rosero-Lópex, D., Knighton, J., Lloret, P., Encalada, A.C., 2020. Invertebrate response to trends and future changes. J. Geophys. Res. Atmos. 118, 4129–4141.
impacts of water diversion and flow regulation in high-altitude tropical streams. Tomasella, J., Hodnett, M.G., Cuartas, L.A., Nobre, A.D., Waterloo, M.J., Oliveira, S.M.,
River Res. Appl. 36 (2), 223–233. 2008. The water balance of an Amazonian micro-catchment: The effect of
Ruosteenoja, K., Carter, T.R., Jylhä, K., Tuomevirta, H., 2003. Future climate in world interannual variability of rainfall on hydrological behavior. Hydrol. Process. 22,
regions: an intercomparision of model-based projections. for the new IPCC emissions 2133–2147.
scenarios. 644, Finnish Environment Institute, Helsinki. Tomlinson, M.S., De Carlo, E.H., 2003. The need for high resolution time series data to
Sakoda, E.T. 2007. Setting instream flow standards for Hawaiian Streams—the role of characterize Hawaiian streams. J. Am. Water Resour. Assoc. 39, 113–123.
Science. In: Biology of Hawaiian Streams and Estuaries (N.L. Evenhuis, J.M. USGS, U.S. Geological Survey. 2019. USGS National Elevation Dataset (NED). https://
Fitzsimons, eds). Bishop Museum Bulletin in Cultural and Environmental Studies, 3: catalog.data.gov/dataset/usgs-national-elevation-dataset-ned.
293-304. Vogel, R.M., Kroll, C.N., 1991. The value of streamflow record augmentation procedures
Scholl, M.A., Gingerich, S.B., Tribble, G.W., 2002. The influence of microclimates and fog in low-flow and flood-flow frequency analysis. J. Hydrol. 125 (3–4), 259–276.
on stable isotope signatures used in interpretation of regional hydrology: East Maui. Willmott, C.J., Robeson, K., 1995. Smart interpolation of annually averaged air
Hawaii. Journal of Hydrology 264, 170–184. temperature in the United States. J. Appl. Meteorol. 34, 2577–2586.
Sharpley, A.N., and Williams, J.R. 1990. EPIC Erosion/Productivity Impact Calculator: 1. Wohl, E., Barros, A., Brunsell, N., Chappell, N.A., Coe, M., Giambelluca, T., Goldsmith, S.,
Model Documentation. US Department of Agriculture Technical Bulletin No. 1768, Harmon, R., Hendrickx, J.M.H., Jubik, J., McDonnell, J., Ogden, F., 2012. The
235 pp. hydrology of the humid tropics. Nat. Clim. Chang. 2, 655–662.
Shen, S.S.P., 2001. Interpolation of 1961–97 daily temperature and precipitation data Xie, P., Arkin, P.A., 1997. Global precipitation: A 17-year monthly analysis based on
onto Alberta polygons of ecodistrict and soil landscapes of Canada. J. Appl. gauge observations, satellite estimates, and numerical model outputs. Bull. Amer.
Meteorol. 40, 2162–2177. Meteor. Soc. 78, 2539–2558.
Shukrullah, S., Naz, M.Y., 2021. Hydropower generation in tropical countries. In: Yi, T.J., Mahmud, M.R., Reba, M.N.M., Hashim, M., Norman, M., Jaafar, W.S.W.M.,
Sulaiman, S.A. (Ed.), Clean Energy Opportunities in Tropical Countries. Green 2021. Forecasting the near future of rainfall in humid tropics using the high-
Energy and Technology, 10.1007/978-981-15-9140-2_2. resolution satellite precipitation data. Phys. Chem. Earth, Parts A/B/C 124 (2),
Spangler, S., Heskett, M., Uyeno, D.D., Strauch, A.M., 2017. Balancing in- and off-stream 103069.
uses of water in the Hawaiian Islands. WIT Trans. Ecol. Environ. 216, 15–26.

12

You might also like