3. Data description
3.1 Data overview
The CHOSEN database contains up to 13 different hydrometeorological
variables, including streamflow, precipitation, air temperature, solar
radiation, relative humidity, wind direction, wind speed, SWE, snow
depth, vapor pressure, soil moisture, soil temperature, and isotope
values, with availability varying from site to site (Figure 4). The HJ
Andrews and Bonanza LTERs have measurements of all 13 variables, with
most of the other watersheds having data of around ten
hydrometeorological variables. Discharge record lengths range from three
years at Calhoun to 78 years at the San Diego River (California Current
Ecosystem LTER), with a median of 19 years.
Figure 4. Span of time series availability and duration across
watersheds
Among all the 13 hydrometeorology variables included in the dataset,
discharge, precipitation, snow depth, soil moisture, and isotope data
are particularly important for hydrologic process studies. Discharge and
precipitation time series are available in all CHOSEN watersheds (Figure
5), and seven catchments have soil moisture and snow measurements with
records exceeding five years. Although publicly available isotope data
are limited, we identified six watersheds with isotope time series
longer than one year (Figure 4).
Figure 5. Distributions of record spans for different variables
3.1.1 Precipitation
In the CHOSEN dataset, 27 watersheds have more than five years of
precipitation data, and 20 watersheds have more than ten years (Figure
5). Twenty-five watersheds have less than 10% missing precipitation
values, increasing to 29 watersheds after applying gap-filling methods
(Figure S1). The sparsest precipitation raw data are from the Bonanza
site, where 24% of the missing values were filled by regression. More
precipitation gap-filling information is available in the supplementary
material.
3.1.2 Soil moisture
Soil moisture is essential for investigating hydrologic connectivity and
runoff processes, especially where vertical flow dominates (Bracken et
al., 2013). Soil moisture measurements are available in 18 watersheds
(Figure 5), usually including multiple stations and depths. Seventeen of
these catchments have less than 10% missing soil moisture data after
gap-filling (Figure S2). The longest soil moisture records on average
are in the HJ Andrews watershed, including multiple stations dispersed
in several sub-watersheds monitoring at different depths. Like the HJ
Andrews watershed, other sites commonly measure soil moisture data at
multiple stations, facilitating gap-filling through spatial regression.
3.1.3 Snow depth / SWE
At high latitudes and altitudes, snowmelt can play an important role in
streamflow generation and nutrient export, and snow accumulation and
melt may be particularly sensitive to climate change. Eight of the
CHOSEN watersheds have snow depth data with less than 10% missing
values after gap-filling (Figure S3). Sagehen watershed has the longest
snow depth record (61 years), with 39 years of SWE data (Table S2).
3.1.4 Isotope data
Isotope tracers (e.g., 18O and deuterium) are
important for estimating catchment transit time distributions, which,
along with hydrologic response timescales, can be used to characterize
the temporal dynamics of the water cycle. Though publicly available
isotope measurements are less abundant than hydrometeorological data,
six of the CHOSEN watersheds have publically available isotope time
series. Among those watersheds, Shale Hills has the longest isotope time
series, consisting of 1103 days of isotope measurements between
2008-03-28 and 2011-12-31. Most of the sites have sub-weekly
δ18O and δ2H measurements in
precipitation and streamflow (Table S4).
3.2 Example data from Dry Creek
watershed
This section presents example data from the Dry Creek Experimental
Watershed (DCEW), located in the semi-arid southwestern region of Idaho,
USA, 16 km northeast of the city of Boise. Raw data were downloaded from
the Boise State University research pagehttps://www.boisestate.edu/drycreek/dry-creek-data/ . Daily
measurements of discharge, precipitation, soil moisture, snow depth, and
six other hydrometeorological variables were collected starting in 1999
at multiple streamflow gauges, weather stations, and soil moisture
sensors distributed in this area (Figure 6).
Figure 6. Dry creek experimental watershed
(Source:https://www.boisestate.edu/drycreek/dry-creek-data/)
Over half of the variables at Dry Creek have less than 10% missing
values at daily time steps. After applying gap-filling methods, all
hydrometeorological variables except snow depth have less than 10%
missing values (Figure 7). The sparsity of snow depth data is due to the
ephemeral nature of the region’s snowpack.
Figure 7. Data filling methods applied to Dry Creek data
The intensively monitored data included in CHOSEN allow for detailed
analyses of hydrometeorology variables at both the seasonal and
interannual time-scale. Here, we briefly describe some of the gap-filled
data from the 2011-2012 hydrological year at Dry Creek (Figure 8). For
streamflow, the highest discharge values were monitored at Lower Gauge
(LG) which is located downstream of the watershed. The lowest discharge
values are from two tributaries: Treeline (TL) and Bogus South Gauge
(BSG). Patterns of precipitation match the responses in streamflow,
especially in January and April. Springtime snowmelt is reflected in
both a decrease in snow depth and persistent high flows during March and
April. The soil moisture time series vary greatly from station to
station, but generally reflect seasonal patterns of precipitation,
snowmelt, and evaporative demand, with shorter-term fluctuations in
shallower sensors showing the influence of individual precipitation
events. This example highlights how CHOSEN data can be instrumental in
understanding the responses of soil moisture and discharge to
hydrometeorological drivers.
Figure 8. Cleaned Dry Creek daily data from 2011-10-01 to
2012-09-30