Flood Prediction Using Watershed Delineation and a Unit Hydrograph: Cienega Creek, Arizona 

Objective

Cienega Creek is a perennial (year-round) creek and an important water resource in arid southeastern Arizona. However, the watershed is susceptible to flash flooding, particularly during the monsoon seasons. The objective of this project is to delineate the Cienega Creek watershed in ArcGIS Pro using a Digital Elevation Model (DEM) and a pour point located immediately downstream of the confluence between Cienega Creek and Davidson Canyon (the outlet). After delineating the watershed, a unit hydrograph will be developed to characterize runoff response and estimate the timing of peak flow. Flow velocity and travel time will be used to generate an isochrone map and a corresponding unit hydrograph plot. Because the study area lies within a natural preserve that includes tourism, ranching, and nearby residences, these results can support flood prediction and inform planning for future flood events. 

Data

Study Area

Cienega Creek begins within the Las Cienegas National Conservation Area in southeastern Arizona, approximately 25 miles southeast of downtown Tucson near Vail. The watershed is situated between the Santa Rita and Empire Mountains to the south, the Rincon Mountains to the north, and the Whetstone Mountains to the east (PAG, 2005). Seasonal monsoon storms can deliver short-duration, high-intensity rainfall that can trigger flash flooding in mountain-influenced basins. Rocky terrain and relatively limited infiltration promote rapid runoff and efficient routing toward valley channels. Because many homes are located along the Davidson Canyon Wash corridor and several major trails are near the Davidson Canyon–Cienega Creek junction, understanding watershed response timing is important for evaluating how quickly floodwaters may reach exposed areas. 

Watershed Outlet (Pour Point)
A pour point was defined using field GPS coordinates of (32.01995, −110.644). This location is positioned immediately downstream of the Davidson Canyon confluence with Cienega Creek, just south of Vail. The pour point was used as the watershed outlet for delineation. 

DEMs

Elevation data were acquired from USGS EarthExplorer as Shuttle Radar Topography Mission (SRTM) 1 arc-second (~30 m resolution) DEM tiles (radar-derived elevation). Two tiles were downloaded to provide complete coverage for the study area and support hydrologic preprocessing and watershed delineation in ArcGIS Pro (Table 1). 

A map displaying the DEM elevation and pour point location is shown in Figure 1.

Data Quality Considerations

SRTM DEMs provide moderate-resolution topography suitable for watershed-scale hydrologic analysis but may not capture small channels, engineered drainage features, or fine-scale terrain that can influence flow routing in flash-flood settings. DEMs can also contain artificial depressions (“sinks”) that interrupt flow direction calculations; therefore, depression identification and filling were performed prior to watershed delineation.

Methods

GIS Software and Data Preparation:

All analysis was completed in ArcGIS Pro 3.6.2 using the Spatial Analyst extension (Hydrology toolset and raster analysis tools). All datasets were processed in a consistent projected coordinate system, NAD_1983_StatePlane_Arizona_Central_FIPS_0202_Feet, to ensure accurate distance-based operations and consistent raster alignment. 

Due to my area of interest spanning between two DEMs, both rasters were combined into a single continuous DEM using the MosaicToNewRaster tool. This new layer was projected to the new coordinate system (NAD_1983_StatePlane_Arizona_Central_FIPS_0202_Feet) to minimize spatial misalignment and ensure distance-based tools (e.g., snapping and travel-time calculations) operated in feet . 

The outlet (pour point) was stored as a CSV and imported using XY Table To Point. Because the coordinates were provided in latitude/longitude, the point was first created in WGS 1984 coordinates and then projected to NAD_1983_StatePlane_Arizona_Central_FIPS_0202_Feet to match the DEM and analysis layers. 

Watershed Delineation

(1) Determine Sinks

The Sink tool was used to locate depressions in the DEM, since these are areas where flow direction cannot be calculated. Because this study area is not known for having natural sinks, these sinks are likely results from inaccuracies introduced during data collection from the sensor.

(2) Fill DEM

The DEM was corrected using the Fill tool, raising sink cells elevations to match that of the lowest neighboring cell, ensuring flow can be routed continuously across the surface.

(3) Calculate Flow Direction

A flow direction raster was created from the filled DEM to represent the direction of steepest descent from each cell.

(4) Calculate Flow Accumulation

A flow accumulation raster was generated from the flow direction raster to estimate contributing area, where higher values indicate likely stream lines.

(5) Con tool to determine streams

A stream raster was created using the Con tool, with a flow accumulation threshold of ≥125,000 cells, identifying the primary stream channel for snapping the pour point (outlet).

(6) Snap Pour Point

The pour point was snapped to the nearest stream cell using Snap Pour Point. The distance from the original pour point location to the nearest stream cell was approximately 180 ft, which was used as the snap distance to ensure the outlet stayed near the intended location.

(7) Delineate Watershed

The Watershed was generated using the Watershed tool with the flow direction raster and the snapped pour point. The resulting watershed represents the contributing area draining to the Davidson Canyon-Cienega Creek outlet (Figure 2).

Unit Hydrograph

The unit hydrograph was developed by estimating spatial travel time to the outlet and summarizing contributing area by time interval. 

(1) Velocity field

For this study, a velocity field that is spatially variant, but time and discharge invariant was used. Velocity was assumed to vary with slope and contributing area, but remain constant with time and discharge. 

To create the velocity field, the Maidment et al. (1996) method was used:

V = Vm * (sbAc) / (sbAcm) 

• V = velocity of a single cell

• Vm  = the average velocity of all cells in the watershed

• sbAcm = average slope area term throughout the watershed

• sbAc = slope area term at the site

Slope area term (sbAc) was calculated: √(slope) * √(flow accumulation) for each cell. Average slope area term (sbAcm ) is the mean of the slope area term, which was roughly 7.71. The average velocity (Vm ) of this region was assumed to be 0.5 ft/s. Velocity was calculated in Raster Calculator.

The initial velocity field contained unrealistic low/high values, ranging from 0 to 518 ft/s, so velocities were constrained to a more reasonable range with a minimum of 0.05 ft/s and maximum velocity of 13 ft/s, with the higher value chosen as an average upper-end flow during monsoon season (National Weather Service).

A weight (impedance) raster was then calculated as: 1 / V. This weight raster represents resistance to flow (time per unit distance), which is required to calculate travel-time accumulation.

(2) Isochrone map

To simplify calculations to the contributing area only, the flow direction raster was clipped to the watershed using Extract by Mask. Travel time to the outlet was computed using the clipped flow direction raster and the weight raster, producing a travel-time surface ranging from 0 to ~77,000 seconds. 

Travel time was grouped into 30-minute (1,800-second) intervals using a reclassification table that assigned each cell to the nearest upper 1,800-second bin (e.g., 0-1,800→1,800; 1,800-3,600→3,600). These grouped travel-time zones form the isochrone map, where each band represents approximately equal travel time to the outlet. For readability, isochrones were displayed in 1-hour increments (Figure 3).

(3) Unit hydrograph

A table of isochrone zones was exported as a stand-alone table to summarize the relationship between time and area of water flowing into the outlet. Two new fields were added to the table using the Add Field tool: Area (ft2) and Unit Hydrograph Ordinate. Because area was initially summarized in cell counts, area was converted to square feet using: Area (ft2) = Count * (cell width) * (cell height), using the DEM cell size ~95.048158 ft. Values were updated in table using Calculate Field. The unit hydrograph ordinate was computed as area per time interval (1,800 seconds): Area (ft2) / 1,800.

The resulting unit hydrograph represents the amount of flow reaching the outlet over time, which is used to approximate when discharge at the outlet is expected to reach its peak following a unit input of rainfall excess (Figure 4).

Final list of assumptions:

• Vm  (mean velocity) = 0.5 ft/s

• Velocity capped 0.05-13 ft/s

• Velocity is time/discharge invariant

• Stream threshold = 125,000 cells

• Isochrone Interval = 1800 seconds (30 minutes)

Results and Discussion

Watershed and Stream Network

Using a flow-accumulation threshold of ≥ 125,000 cells, the extracted stream raster produced a main-channel pattern consistent with expected drainage pathways. The outlet was snapped to the nearest high-accumulation stream cell using a snap distance of 180 ft, maintaining the pour point at the intended Davidson Canyon–Cienega Creek confluence. The resulting watershed delineation defines a contributing area of approximately 11.5 * 109 ft2 (~410 miles2) and extends from the surrounding mountain ranges to the outlet, capturing the primary tributary network draining into Cienega Creek. Overall, the watershed geometry and derived stream network indicate substantial headwater contributions and relatively efficient routing along the main channel. 

Velocity Field and Travel Time (Isochrones)

The initial velocity surface contained unrealistic extremes, so velocities were constrained to 0.05–13 ft/s. The adjusted velocity field shows higher velocities concentrated along stream corridors and lower velocities occur in upland areas, which is consistent with increasing flow concentration along high-accumulation paths. Travel time increases away from the outlet, ranging from approximately 1 hour near the outlet zone to a maximum of about 21 hours in the most distal southern portions of the watershed. Short travel times (1–4 hours) are clustered near the outlet and along the main stream network, whereas the longest travel times (18–21 hours) occur primarily in the southern headwater regions. A broad mid-range travel-time zone (~6–13 hours) covers much of the central and northeastern watershed, indicating that a large fraction of the basin contributes runoff within a relatively similar time window. This clustering of travel times helps explain the timing and shape of the basin response observed in the unit hydrograph. 

Unit Hydrograph

The unit hydrograph exhibits a steep rising limb followed by the dominant peak, a secondary peak, and a gradual recession. The primary peak discharge occurs at approximately 22,000 seconds (~6 hours), indicating that the largest concentration of contributing area reaches the outlet within this time window. This timing is consistent with rapid routing from near-channel and intermediate-distance portions of the watershed that connect efficiently to the main drainage network. A secondary rise occurs around 36,000 seconds (~10 hours), suggesting a second cluster of contributing area arriving later, likely associated with longer flow paths represented by the mid- to late-range isochrone zones. Following the secondary peak, the hydrograph recedes gradually and approaches near-zero values by approximately 77,000 seconds (~21 hours), consistent with the maximum travel times shown in the isochrone map. 

Model Assumptions and Limitations

Hydrograph timing and travel-time results are sensitive tot he velocity assumptions used. Although constraining velocities to 0.05-13ft produces more realistic values than the unconstrained output, the method still assumes velocity is time-and-discharge invariant, and depends on an assumed mean velocity, which simplifies real flood behavior. In addition, flow routing is derived from a moderate-resolution SRTM DEM, which may not capture small channels, engineered drainage features, or fine-scale topography that can influence runoff pathways. For these reasons, the unit hydrograph should be interpreted as a first-order estimate of response timing, rather than a calibrated discharge forecast. 

Conclusion

This project delineated the Cienega Creek watershed in ArcGIS Pro and applied a travel-time approach to generate an isochrone map and unit hydrograph at the Davidson Canyon–Cienega Creek outlet. The results suggest a rapid watershed response, with a dominant peak occurring at approximately 6 hours, a secondary rise near 10 hours, and maximum travel times approaching ~21 hours. These peak-timing estimates should be interpreted as first-order approximations because they depend on a simplified velocity field that is spatially variable but time-and-discharge invariant, an assumed mean velocity (0.5 ft/s), and constrained velocity limits (0.05–13 ft/s), as well as DEM-derived flow routing. Despite these limitations, the outputs provide a useful estimate of flood response timing that can support preparedness, warning communication, and response planning for visitors, ranches, and nearby residences in and around the preserve. 

References

National Weather Service. (n.d.). Flash floods and floods… the awesome power! A preparedness guide. Retrieved March 3, 2026, from https://www.weather.gov/pbz/floods 

Pima Association of Governments (PAG). 2005. Unique Water Nomination for Davidson Canyon, 63p.

U.S. Geological Survey. (n.d.). EarthExplorer. Retrieved February 28, 2026, from https://earthexplorer.usgs.gov/