High resolution spatiotemporal patterns of flow at the landscape scale in montane non‐perennial streams

Intermittent and ephemeral streams in dryland environments support diverse assemblages of aquatic and terrestrial life. Understanding when and where water flows provide insights into the availability of water, its response to external controlling factors, and potential sensitivity to climate change and a host of human activities. Knowledge regarding the timing of drying/wetting cycles can also be useful to map critical habitats for species and ecosystems that rely on these temporary water sources. However, identifying the locations and monitoring the timing of streamflow and channel sediment moisture remains a challenging endeavor. In this paper, we analyzed daily conductivity from 37 sensors distributed along 10 streams across an arid mountain front in Arizona (United States) to assess spatiotemporal patterns in flow permanence, defined as the timing and extent of water in streams. Conductivity sensors provide information on surface flow and sediment moisture, supporting a stream classification based on seasonal flow dynamics. Our results provide insight into flow responses to seasonal rainfall, highlighting stream reaches very reactive to rainfall versus those demonstrating more stable streamflow. The strength of stream responses to precipitation are explored in the context of surficial geology. In summary, conductivity data can be used to map potential stream habitat for water‐dependent species in both space and time, while also providing the basis upon which sensitivity to ongoing climate change can be evaluated.

to map potential habitats in drylands, and assess how the distribution of these habitats might shift with climate change Zipper et al., 2021). In this paper, we leverage a conductivity dataset from a network of sensors deployed in streams along an arid mountain front to investigate the spatio-temporal distribution of flow, map flow condition, identify potential controls on surface flow and sediment moisture, and establish a seasonal classification of flow for dryland streams. The resulting spatial and temporal maps of temporary flow can provide useful information for assessment of habitat suitability for a wide range of species, and to support improved interpretations of the linkages between climate forcing and mountain front hydrology.
Dryland regions, defined as areas where plant productivity is limited by water availability, cover about 41% of the land surface (Millennium Ecosystem Assessment, 2005) and are dominated by temporary streams that dry at least once per year (Messager et al., 2021).
Significant drylands include the southwestern region of the United States, where $81% of streams are classified as non-perennial, a proportion which rises to 94% in Arizona (Levick et al., 2008;Nadeau & Rains, 2007). Non-perennial streams occasionally dry out (fully dry streambed), and can be classified as ephemeral or intermittent, with ephemeral reaches reaching surface flow only in response to rainfall, while intermittent reaches display cycles of drying and wetting (Busch et al., 2020;Gallo, Meixner, Lohse, & Nicholas, 2020;Levick et al., 2008;Levick et al., 2015). Perennial streams flow yearround, supplied by groundwater discharge to the stream bed. Streams that alternate between perennial, ephemeral and intermittent reaches are considered interrupted or spatially intermittent (Levick et al., 2008). Streamflow permanence is controlled by various environmental factors such as rainfall distribution, evaporative demand, topography, underlying geology, streambed composition, channel morphology and vegetation (Costigan, Jaeger, Goss, Fritz, & Goebel, 2016;Goodrich, Kepner, Levick, & Wigington Jr., 2018;Levick et al., 2018;Shanafield, Bourke, Zimmer, & Costigan, 2021;Singer & Michaelides, 2014), but climate-induced aridity (balance between rainfall and evapotranspiration) is considered an overarching key driver Sauquet et al., 2021). In areas with a seasonal distribution of precipitation, such as the region of the Southwest USA affected by the North American Monsoon, flow permanence can follow this highly uneven temporal distribution (Eng et al., 2016;Singer & Michaelides, 2017).
High variability in upstream-downstream arrangement of perennial and non-perennial streams support a mosaic of habitats for plant and animal life (Boulton, Rolls, Jaeger, & Datry, 2017;Datry et al., 2014;Larned et al., 2010). Spatial and temporal variations in habitat patch distribution and composition lead to high watershedscale species diversity (Burnett et al., 1998;Larned et al., 2010;Stromberg et al., 2015). In drylands specifically, the presence of these wet reaches contributes to the strong contrast in water availability between riparian areas and the surrounding arid landscape (Levick et al., 2015;Stromberg et al., 2015), leading to contrasts in flora and fauna both in terms of species composition and density (Goodrich et al., 2018;Levick et al., 2008;Sabathier, Singer, Stella, Roberts, & Caylor, 2021). The denser vegetation of riparian forests and wetlands is used for foraging, nesting or as migration corridors and stopovers, cool and humid refuges, and seed dispersal corridors (Datry et al., 2014;Levick et al., 2008).
More frequent and severe droughts linked to climate change are projected to significantly alter flow intermittence patterns and hydrologic connectivity in dryland streams by increasing the number of zero-flow days and the length and frequency of dry channel reaches (Jaeger, Olden, & Pelland, 2014;Sauquet et al., 2021;Zipper et al., 2021). In the United States, a general decline in surface-water availability and soil moisture is expected across the southwestern region (Seager et al., 2013), which would dramatically impact ephemeral and intermittent channels. This water-availability decline, added to other stressors such as water pumping and other flow diversions, lead to loss of wetlands and the species they host (Hendrickson & Minckley, 1985). Knowing precisely where and when there is surface flow is essential to map the distribution of potential streamside habitats, but also to anticipate habitat distribution shifts induced by climate change (Allen et al., 2019;Jaeger et al., 2014;Sauquet et al., 2021).
To understand how flow permanence varies along streams across a mountain front within a dry climatic region, we use electrical conductivity sensors to detect dryness and wetness of the streambed (Blasch, Ferré, Christensen, & Hoffmann, 2002;Chapin, Todd, & Zeigler, 2014;Jaeger & Olden, 2012). These sensors can be used in ephemeral headwaters to map perennial and intermittent flow (Adams, Monroe, Springer, Blasch, & Bills, 2006;Assendelft & van Meerveld, 2019). The fine spatial resolution and high temporal frequency of observations are capable of capturing flow variability (Arismendi et al., 2017, Larned et al., 2011 to support classification of ephemeral and intermittent streams and better understand the environmental factors governing water distribution (Jensen, McGuire, McLaughlin, & Scott, 2019). A similar method was used by Gallo et al. (2020) across the same mountain front with a limited number of sensors across three canyons, and focusing on rainfall and sediment hydraulic conductivity. We use daily conductivity from 37 sensors across 10 canyons to compare seasonal flow timing to precipitation and underlying geology. This high spatial and temporal resolution dataset, which provides daily information for all the main headwater streams on the north-eastern slopes of the mountain range, allows for an understanding of landscape-level flow patterns and helps decipher regional (rainfall) and local (geology) environmental controls on flow permanence.

| METHODS
We investigated spatial and temporal variability of flow by mapping daily electrical conductivity (EC) values across our study site and compared these values to daily rainfall. We then sorted each sensor in a seasonal classification to link flow condition to seasonal rainfall. This response is evaluated further by comparing rainfall and EC values over several years of variable precipitation distribution. Lastly, we compared stream reaches and their seasonal classes to permeability of the underlying geology to examine the role of geology as a potential factor to flow patterns in nonperennial streams.

| Study site
Our study site spans 10 non-perennial streams spread across the eastern side of the Huachuca Mountains, a mountain range in southeastern Arizona that is part of the Madrean Sky Islands (Figure 1a Streamflow is fed by rainfall and to a lesser extent by snowmelt and the local water table. Short but strong monsoon storms that occur from July to September comprise $60% of annual rainfall, with less intensive winter precipitation providing the remainder. The driest season occurs before the monsoon, from May to June. Precipitation is greater at higher elevations ( Figure 1c). The monsoon brings intense thunderstorms that turn into runoff and floods, while milder winter rains and snowmelt more readily infiltrate and provide soil moisture (Loik, Breshears, Lauenroth, & Belnap, 2004;Vera et al., 2006).
The streams of the Huachuca Mountains cross over a diversity of geologic units (mudstone, limestone, quartzite, and granite), as well as several faults (Brown, Davidson, Kister, & Thomsen, 1966) before reaching the lowlands. Channels have cascade and step pool morphology at the upper extents typical of steep headwater streams and transition to pool riffle morphology in the downstream valley (Wohl & Pearthree, 1991). The valley surrounding the mountains is composed of permeable basin fill, terrace deposits and stream alluvium. Water crosses the valley underground within the basin fill (or in washes during the strongest monsoon events) to reach the two main intermittent rivers draining the area: the San Pedro River to the east, and the F I G U R E 1 Study area (National Agriculture Imagery Program image) with streams and location of sensors (BT, Blacktail; SL, Slaughter House; SPR, Split Rock; H, Huachuca; RS, Rock Spring; W, Woodcutters; G, Garden; T, Tinker; B, Brown; R, Ramsey Canyons) (a), location of the study site in Arizona, United States (b) and regional annual rainfall distribution (from PRISM, https://prism.oregonstate.edu/normals/) (c). White lines represent the main rivers and the mountain streams equipped with sensors [Color figure can be viewed at wileyonlinelibrary.com] Babocomari River to the north (Gungle, 2006;Levick et al., 2008) ( Figure 1a).

| Sensor array
A total of 37 electrical conductivity (EC) sensors were installed along 10 streams of the Huachuca Mountains, and operated between 2010 and 2014 (Sabathier & Jaeger, 2022) (Figure 1a). Originally, 44 sensors were installed, but seven were omitted from this study because of short recording periods or quality issues. These sensors were initially installed to quantify flow condition (flow, wet or dry) through both time and space, including longitudinal flow connectivity (Jaeger & Olden, 2012). Their high spatial and temporal resolution is useful for capturing responses to local and short-term climatic events over wide areas (Adams et al., 2006;Assendelft & van Meerveld, 2019;Jensen et al., 2019). These EC sensors recorded relative conductivity every 15 minutes, with large values reflecting surface water presence, and smaller values reflecting dry channel conditions. Conductivity values are considered relative to each other consistent with other studies (Jensen et al., 2019;Warix, Godsey, Lohse, & Hale, 2021) as sensor values were not calibrated with a solution of known conductivity. The data collected can be used to detect onset and end of flow in non-perennial streams that are too small or too dry to be equipped with streamflow gauges (Blasch et al., 2002;Chapin et al., 2014;Goulsbra, Lindsay, & Evans, 2009;Stromberg et al., 2015).
We used daily average values of relative conductivity from the June 1, 2010 to May 31, 2011 to analyze flow permanence. This time frame was chosen because it covers a full year during which all 37 sensors operated without gaps. To investigate inter-annual variability of flow, we used four sensors (G2, H7, T1 and T2) that recorded EC for 3 years. Electrical conductivity records a low and constant value in dry sediment and progressively increases in wet sediments, finally exhibiting an abrupt increase at the onset of streamflow (Blasch et al., 2002;Goulsbra et al., 2009). Because the sensors are buried to a depth of <10 cm in the channel bed, sediment type or grain size distribution can affect the recorded values (Blasch et al., 2002), and may also cause a delay between the onset or cessation of flow and observed electrical conductivity (Adams et al., 2006;Blasch et al., 2002).
Sensors G4 (Garden Canyon) and H3 (Huachuca Canyon) were located close to U.S. Geological Survey (USGS) stream gauges 9470800 and 9471310, respectively. Daily streamflow data were downloaded from the USGS database (U.S. Geological Survey, 2022).
The co-location of conductivity and flow data allowed us to directly classify conductivity in terms of flow permanence. We compared gauged stream discharge with adjacent EC sensor values for the same time-step and period (Adams et al., 2006;Blasch et al., 2002;Stromberg et al., 2015) (Figure 2a,b). Some discrepancies between these datasets are expected, due to mismatches in precise location and measurement resolution, but their comparison provides an indication of how EC sensors react to flow conditions. Acknowledging the uncertainty of this method and the potential influence of spatial variability in stream bed substrate, we focus on general categories covering a range of values. Thus, we built a scale between relative conductivity values and flow state (dry, wet sediment and flow) ( Table 1). A relative conductivity value of À90 is considered to represent dry sediment, as it is the lowest values reached by the sensors, and it is the only value that remains constant with no variations for days or weeks at a time. The threshold for water in the stream is 0 as this is the value reached during the sharpest conductivity peaks, following the strongest rainfall events ( Figure 2).

| Flow condition classification
Relative conductivity was further classified into two classes: dry and wet, with the wet class including wet sediment and flow (or standing water) ( Table 1). The seasonal classification of each sensor was estab- To understand the impact of rainfall on flow permanence, we used daily rainfall from PRISM (https://prism.oregonstate.edu/recent/ ), a gridded dataset at 4-km resolution, modeled by interpolation from ground stations, climate data and elevation (Daly et al., 2008). Rainfall from June 2010 to May 2011 was only 351 mm, which is a characteristically dry year compared to the 30-years (1991-2020) average precipitation from PRISM of 409 mm. Rainfall distribution across the year was also slightly unusual, with a stronger monsoon in 2010-2011 (308 mm vs. 235 mm for the 30-year average), but a drier winter and spring.
We defined a classification based on temporal distribution of flow condition throughout the year, a common way to classify ephemeral streams (Costigan et al., 2016;Eng et al., 2016;Sauquet et al., 2021).
Daily rainfall was used to define the seasons based on precipitation distribution. We divided the year into four seasons, based on rainfall temporal distribution: dry spring (May-June), the summer monsoon (July-September), dry autumn (October-November) and wet winter (December-April). For each sensor, we counted the number of "wet" days (wet sediment or flow, relative conductivity above À90, table 1) in each season. If the sensor measured flow or wet sediment for more than 50% of the season, then the whole season is considered "wet" for this sensor. All 37 sensors could then be assigned to one of six classes depending on when the stream reach is wet (always dry, wet during monsoon, wet during monsoon and autumn, wet during monsoon and winter, wet from monsoon to winter, always wet).
Underlying geology was also investigated for its association with local flow permanence; this was made possible based on the location of units with different permeability and fracturing (Goodrich et al., 2018;Larned et al., 2011;Levick et al., 2008). We used the hydrogeologic map and report from Brown et al. (1966), which provides information on geologic units, springs and faults across the Huachuca Mountains, to conduct a qualitative interpretation of the links between geology and flow condition. The hydrogeologic map, covering the north-east section of the Huachuca Mountains and the plain between the mountain front, the San Pedro River and the Babocomari River, was digitized by hand in QGIS and augmented with information on lithologic unit permeability (Sabathier & Jaeger, 2022).
T A B L E 1 Relative conductivity measured by the sensors and its translation to flow condition and simplified state used for seasonal classification. Flow condition is represented by a set color ramp through this paper [Color    drying/re-wetting, before going back to remaining wet from monsoon to winter in 2012 (Figure 7c).

| DISCUSSION
Non-perennial streams in drylands are important sources of moisture and hotspots of biodiversity Datry et al., 2014;Larned et al., 2010). As such, understanding the timing and distribution of flow is critical for mapping habitats and their potential climate change vulnerability . In this paper, we demonstrate how electrical conductivity sensor data can be used to map distribution of surface water and channel sediment moisture at high spatiotemporal resolution in small non-perennial streams. This information can then be used to classify stream reaches by seasonal patterns, a useful metric to summarize the temporal variability of flow in a way that can be compared to climate patterns and related to wildlife and vegetation dynamics. The Most reaches are very sensitive to rainfall and only flow during the monsoon and/or the winter rain season, while others remain constant (always dry or always wet) no matter the precipitation input (Gallo et al., 2020). The light winter precipitation and melting snow (low intensity and long duration) travels more slowly and has greater potential to infiltrate into the ground and feed the many springs that supply the perennial reaches (Stromberg et al., 2015), while the intense monsoon storms (high intensity, high frequency, and short duration) are more likely to initiate overland flow in the canyons (Levick et al., 2008;Stromberg et al., 2015). Non-perennial reaches can be more or less responsive to rainfall. Some reaches get wet both during monsoon and winter, responding to the smallest precipitation events, and others that need significant rain falling in a short period only flow during the monsoon.
Despite the small sample size and the limited number of parameters investigated in this study, we can still combine our findings and the literature to identify the potential controls on flow permanence in these canyons. The reaches that exhibit wet sediment or flow during the monsoon and stay wet through the dry autumn and to the winter are likely fed by local aquifers that manage to fill up during the monsoon. For example, flatter areas can allow for seepage into the local aquifer to feed the stream downstream, and faults form preferential paths for groundwater drainage to springs (Lovill, Hahm, & Dietrich, 2018;Martin, Kampf, Hammond, Wilson, & Anderson, 2021). Areas sheltered by vegetation or the surrounding topography might also stay wet longer, as evaporation is reduced. As for the reaches that remain dry, we noticed that they were either on top of permeable sand and gravel layers or colluvium.

| Geology and additional controls
Streams in the Huachuca Mountains, as is true in other ephemeral streams of the Southwest USA, show abrupt longitudinal changes in flow permanence influenced by geomorphological processes and discontinuities (Goodrich et al., 2018;Larned et al., 2011;Lovill et al., 2018). An example of how water moves downstream along Huachuca Canyon is shown in Figure 8. Geology can alter surface hydrology through permeability of underlying formation, spring location, perched aquifers, faults, fractures or sediment deposits (Levick et al., 2008). The headwaters of the streams studied here are mainly located on top of mudstone and sandstone, before meeting limestone. Water (blue arrows) seeps in fractured mudstone and limestone before reaching the surface when encountering impervious units (granite) and faults. At the bottom of the mountain front, water travels down in permeable sediment layers to reach the regional water table [Color figure can be viewed at wileyonlinelibrary.com] collect water, which is then released to springs and streams, while the very high permeability of limestone, due to a high density of fractures and solution channels, is interrupted by impervious siltstone beds (Brown et al., 1966). This upper area of the mountains is also dissected by faults that form preferential flow paths for water. The diversity of structures, each with their own permeability to water, in part, leads to the diversity of flow permanence patterns we see along the canyons. The lower half of the mountain range is underlain by quartzite and granite and it is on top of these impermeable bedrock units that we observed an increase in flow permanence in our canyons and where most of the perennial flow occurs. Down in the San Pedro River basin, water travels over the low permeability conglomerate before reaching the sand and gravel of the sedimentary basin fill that form a highly permeable fan around the Huachuca Mountains (Brown et al., 1966). This is the area with some of the driest reaches in our study. Rainfall distribution is also highly dependent on elevation, with higher areas receiving more rainfall and lowland stream reaches receiving lower precipitation. Monsoon storms can also cover small extents and might cover only one watershed, bringing water to one canyon while its neighbors remain dry.
Local channel conditions and human activity can override expected geologic response at local scale. Channel geometry and stream channel density, itself dependent on grain size and sediment composition, are important reach-scale controls on flow permanence and streambed sediment moisture (Gallo et al., 2020;Larned et al., 2010;Pate, Segura, & Bladon, 2020;Whiting & Godsey, 2016).
Some sensors, such as BT2, were dry no matter the underlying geology; a result likely due to a thick and very permeable sediment layer in the streambed. There are also anthropogenic controls on flow permanence in the Huachuca Mountains. The streams of Garden and Huachuca Canyons have historically been used as a water source for the U.S. Army Installation Fort Huachuca. Spring boxes and pipes are still redirecting water down to the fort (Brown et al., 1966). Some downstream reaches are in urban areas, which can also affect flow regimes. Artificial impervious surfaces prevent rainwater from seeping through the sediment and redirect it instead to the non-perennial washes, which leads to flow being present more often and for longer periods (Gungle, 2006).

| Interannual variability
Response to inter-annual variations in rainfall was also variable, with

| Implications for conservation
Conductivity analysis demonstrated in this paper could be an important tool for mapping potential habitats for species of conservation interest. The key elements that make this work useful are its high temporal resolution (daily data in remote areas with complex topography, which makes fieldwork time-consuming), high spatial resolution (we are able to measure flow state at a precise location), and the fact that the high number of sensors spread across and along streams provide a landscape-scale overview of temporal and spatial distribution of moisture and flow across a whole mountain front. The dataset provides information on sediment moisture and surface flow, but also on which state is reached when, and how often. This record of spatiotemporal distribution of flow and soil moisture supports efforts to pinpoint reaches of perennial flow (such as H4 or G2) in an otherwise dry region. Reaches that manage to remain wet during the dry spring (all sensors in the "always wet" class) can play a critical role as moist and cool refuges. Once flow resumes in the drier reaches and the stream network connects, animals that had found shelter in the perennial reaches can re-colonize the whole network (Bunn, Thoms, Hamilton, & Capon, 2006;Larned et al., 2010). They can also be favorable habitats for species such as the Huachuca water umbel (Lilaeopsis schaffneriana ssp. Recurva), an herbaceous, semi-aquatic perennial plant which needs a permanently wet environment (Bagne & Finch, 2013). For non-perennial reaches, we are able to compare the limited periods of sediment moisture or surface water to the phenology of species of interest. The Chiricahua Leopard Frog (Lithobates chiricahuensis) can be found in temporary streams that dry periodically to discourage non-native predators and competitors while still staying moist enough for the frogs and with surface water for breeding, the breeding period depending on elevation (Bagne & Finch, 2013). A host of other amphibian species in the region are dependent on the patchwork of water availability , as are aquatic invertebrates . Being able to map the distribution of flow across the landscape could also be used to highlight potential wet corridors for allowing species dependent on sediment moisture and surface water to travel between favorable habitats and breeding locations. Data on flow permanence can be paired with wildlife and vegetation surveys (through camera traps, bioacoustics, or remote sensing for example) to study flow permanence as a parameter in habitat mapping and species distribution.
Riparian and semi-aquatic species in the study region are considered highly vulnerable to climate change (Bagne & Finch, 2013), so recording the flow condition in streams for several years could be a useful tool for detecting areas that are particularly sensitive to variations in rainfall and/or moisture. A shift in rainfall distribution, timing and intensity could, for example, change the distribution of seasonal water patterns across the landscape and create ecological shifts for communities depending on specific flow regimes Jaeger et al., 2014;Stromberg, 2013). Depending on the controls governing the presence of water in a reach, areas might be more or less sensitive to climate change. In reaches that are more sensitive to precipitation, a dry winter might lead to an earlier drying of a stream that usually flows until spring. A non-perennial stream that only responds to monsoon rains (wet during monsoon class, sensors H6 and G6 for example) might be very responsive to a stronger or weaker monsoon, while reaches sustained by groundwater inputs (always wet class, sensor BT1 or G2 for example) might be better buffered and could remain wet, affirming their critical status as refuge for drought-sensitive species (Gallo et al., 2020;Stromberg et al., 2015). Springs are the areas most likely to provide a steady water source to the surface, but they are reliant on sufficient water inputs in upstream locations that replenish the local aquifers. Thus, severe changes in rainfall regimes could also lead to shifts even in the wettest perennial reaches (Van Loon, 2015).
The vegetation and wildlife in these always wet reaches might also be more severely impacted, as they are adapted to a perennial water source, and a temporary dry-up could lead to changes in riparian forest extent and shift in species. Increases in dryness could also lead to a loss of connectivity, with flowing reaches becoming less frequent and more isolated (Jaeger & Olden, 2012;Seager et al., 2013).

| CONCLUSIONS
We documented the spatiotemporal variations in flow permanence and channel substrate moisture in the temporary streams of the Huachuca Mountains (southeastern Arizona, USA). We distinguished between reaches highly responsive to local climate and those with more stable flow patterns. Although climate is the first control on water distribution at the regional scale, we revealed that underlying geology, as well as other localized factors such as streambed composition and landscape topology, affect flow permanence locally. Our work shows how the high spatial and temporal resolution provided by electrical conductivity sensors can be used to build a local, reach-scale understanding of surface flow permanence and distribution by using a seasonal classification of flow patterns, and how the resulting. The resulting local, reach-scale understanding of surface water distribution can then provide critical information on potential habitat for riparian species and these habitats' sensitivity to climate change.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in ScienceBase at https://doi.org/10.5066/P90K3SIL. [Correction added on 26 November 2022, after first online publication: the sentence 'Daily electrical conductivity sensor data and the digitized geology layer following Brown et al. (1966) are found in Sabathier and Jaeger (2022).' has been deleted in this section] ORCID Romy Sabathier https://orcid.org/0000-0001-9401-7871