Pearlstone-run Report
2026-11-06
Site Description
Catchment: Pearlstone is a retreat center and educational property in Baltimore County that combines working agricultural land with conservation and outdoor learning programs.
(39°31’11”N, 76°50’51”W), draining 0.71 square miles of agricultural land cover with a mixed riparian corridor bordered by open fields.
Sensor Deployment: Three in-stream sensors were deployed at Downstream, Midstream, and Upstream positions, with a fourth Air sensor mounted as a dry atmospheric reference. The Downstream sensor was first placed in a small ravine and was moved after a tree fell across that position. The new position showed active flow over a streambed of large stones. Midstream was fully submerged in an open reach with moderate flow. Upstream was placed within a well-developed riparian zone with trickling flow.
Goal: To document flow patterns in this watershed and assess whether low-cost thermal sensing is a viable approach for detecting subsurface flow activity and flow permanence.
Data Note: Each sensor records a temperature reading every 15 minutes, and its memory holds a maximum of 8,192 readings, which is about 85 days. Once the memory is full, the sensor stops recording until it is retrieved and restarted. The sensors deployed in September 2025 reached this limit in early December 2025, and they were not retrieved until January 2026. This period appears as blank space in the heatmap and as a break in the temperature plots.
Study Area Map
Figure 1: Study area map of Pearlstone Run, Baltimore County, Maryland. Sensor points indicate the locations of the Upstream, Midstream, and Downstream in-stream temperature loggers deployed along the longitudinal transect. The inset map shows the location of Pearlstone Run relative to all study sites within the Baltimore County area.
Figure 2: Daily precipitation recorded at the Mount Washington 0.5 N rain gauge (NOAA station US1MDBC0003) from July 2025 through October 2026. Each point shows the total rainfall for one day in inches.
Figure 3: Stacked bar chart showing the proportion of days classified as wet (blue), dry (red), and uncertain (gray) for each sensor position at Pearlstone Run across the full study period.
Figure 4: Daily wetness state classifications for all sensor positions at Pearlstone Run from July 2025 through May 2026. Each row represents a sensor position (Air, Downstream, Midstream, Upstream) and each column represents a single day. Blue indicates wet, red indicates dry, and gray indicates uncertain classifications.
Sensor Images
07/01/25
Figure 5: Initial deployment in July 2025. Vegetation
at the Upstream position extends down to the stream. Midstream is a
slow-moving reach with an eroded bank. Downstream sits at the bottom of
a hillside in a ravine, also with slow flow.
09/15/25
Figure 6: September 2025 deployment. Upstream has
trickling flow surrounded by dense vegetation. Midstream has riparian
cover and slow-moving water. Downstream sits in a pond-like area created
by a tree that fell across the site.
01/20/26
Figure 7: January deployment shows how mid-winter
effects the water surface flow. By this time, plant life has gone
dormant for the season. Upstream and Downstream locations were frozen
over, while Midstream had extremely slow waters.
03/09/26
Figure 8: March 2026 deployment. All positions have
consistent flow. The Downstream sensor was moved to an adjacent
position. Because of the fallen tree, the pond where it previously sat
may have produced inaccurate results.
05/25/26
Figure 9: May 2026 deployment. Vegetation has regrown
since the previous deployment. All positions show flow, consistent with
groundwater recharge, and Downstream shows the most movement.
07/22/26
Figure 10: July 2026 deployment, one year after the
first. Midstream and Downstream have consistent flow. Upstream has begun
to dry above the sensor, visible in the upstream-facing image.
10/10/26
Figure 11: October 2026 retrieval. Flow is slower at
all stream sensors. Riparian cover is beginning to drop leaves but
remains abundant.
Thermal Overview Plots
Figure 12: Temperature recorded every 15 minutes by all four sensors from July 2025 to October 2026. The Air sensor reflects atmospheric conditions and the stream sensors reflect channel conditions at each position. The break from December 2025 to January 2026 is the data gap. In-stream water temperatures begin in the low 70s°Fin summer before declining toward freezing by January, while the Air sensor exhibits considerably larger diurnal swings throughout the entire period.
Figure 13: Daily mean temperature for all four sensors. Averaging hourly readings to a daily mean smooths short-term fluctuations and reveals seasonal temperature trends across the monitoring period. Convergence between Air and stream sensor temperatures may reflect overcast, cool, or rainy conditions suppressing diurnal contrast rather than reduced surface water presence. After July 2026, the Upstream sensor departs from the other stream sensors and moves closer to the Air sensor.
Case studies
Purpose: To illustrate sensor response during different atmospheric and seasonal conditions. Each case study presents a four-panel diagnostic plot followed by a dryness score table for a single selected day.
4-Panel Plots: A four-panel diagnostic figure showing each sensor’s thermal behavior over a 24-hour period.
Panel 1 shows raw temperature readings.
Panel 2 shows temperature rise above each sensor’s morning minimum.
Panel 3 shows the rate of temperature change per minute.
Panel 4 shows the computed dryness score for each sensor. Scores near 0 indicate wet conditions, scores near 1 indicate dry conditions.
Dryness Score table: A summary of each sensor’s thermal classification for the selected day.
The thermal classifier scores each sensor-day from 0 to 1, where 0 indicates wet conditions and 1 indicates dry conditions. The score increases as conditions become drier.
Warming score and cooling score reflect thermal response during the morning and afternoon phases respectively.
Daily score is the average of both.
A conflict flag indicates the two phases produced contradictory results.
Dryness state is the final classification — wet, dry, or uncertain.
September 18th, 2026 (Normal summer day)
Figure 14: Four-panel diagnostic plot for September 18th, 2026. Air temperatures ranged from 67 to 87°F with wind gusts up to 12 mph. Downstream and Midstream show dampened daily temperature movement compared with Air. Upstream shows the largest daily rise of the stream sensors, about 10°F above its morning minimum, compared with about 15°F for Air.
| Sensor | warming_score | cooling_score | daily_score | conflict_flag | dryness_state |
|---|---|---|---|---|---|
| Air | 1.0000 | 1.0000 | 1.0000 | FALSE | dry |
| Downstream | 0.0000 | 0.0006 | 0.0003 | FALSE | wet |
| Midstream | 0.1663 | 0.0000 | 0.0831 | FALSE | wet |
| Upstream | 0.5000 | 0.5001 | 0.5001 | FALSE | uncertain |
Table 1: A summary of September 18th sensor’s thermal classification. No conflict flags were detected. Upstream is classified as uncertain because its daily score of 0.5 falls between the wet and dry thresholds.
August 14th, 2025 (Summer rainfall day)
Figure 15: Four-panel diagnostic plot for August 14th,
2025. Temperatures ranged from 73–90°F with wind gusts up to 12 mph.
Streams are moving with expected flow patterns with dampened temperature
patterns compared to the air sensor. Air rises approximately 15 degrees
above its minimum.
| Sensor | warming_score | cooling_score | daily_score | conflict_flag | dryness_state |
|---|---|---|---|---|---|
| Air | 1e+00 | 1e+00 | 1e+00 | FALSE | dry |
| Downstream | 0e+00 | 0e+00 | 0e+00 | FALSE | wet |
| Midstream | 4e-04 | 1e-04 | 3e-04 | FALSE | wet |
| Upstream | 0e+00 | 0e+00 | 0e+00 | FALSE | wet |
Table 2: A summary of August 14th sensor’s thermal classification. No conflict flags were detected and all sensors are moving along as expected. No conflict flags are exhibited and all in-sensors are classified as wet.
October 10, 2025 (Normal winter day) Close-up
Figure 16: Four-panel diagnostic plot for December 10th, 2025. Temperatures ranged from 30–45°F with overcast skies and wind gusts of 13 mph. All three stream sensors were classified as wet, with no conflict flags. The Air sensor rose astonishingly, approximately 40°F above its morning minimum. At 6:00 AM and during the rise, the skies were clearer, so it can be assumed that the exposed sensor was hit by direct sunlight, and absorption of this isn’t a realistic atmospheric temperature.
| Sensor | warming_score | cooling_score | daily_score | conflict_flag | dryness_state |
|---|---|---|---|---|---|
| Air | 1.0000 | 1e+00 | 1.0000 | FALSE | dry |
| Downstream | 0.0003 | 1e-04 | 0.0002 | FALSE | wet |
| Midstream | 0.0000 | 0e+00 | 0.0000 | FALSE | wet |
| Upstream | 0.0912 | 1e-04 | 0.0456 | FALSE | wet |
Table 3: All three stream sensors were classified as wet with no conflict flags. Air scored 1.0 because its diurnal temperature swing remained responsive to the atmosphere while the stream sensors stayed thermally regulated by water.
January 25th, 2026 (Precipitated winter day)
Figure 17: Four-panel diagnostic plot for January 25th, 2026. Temperatures ranged from 15 to 25°F, with 3.37 inches of snow and sleet accumulating. The Air sensor stayed between roughly 14 and 21°F, representing the atmospheric temperatures and visibly the coldest sensor in the plot. While all the midstream and downstream positions sit comfortably at 33 degrees, the upstream sensor rises to approximately 6 degrees above its minimum.
| Sensor | warming_score | cooling_score | daily_score | conflict_flag | dryness_state |
|---|---|---|---|---|---|
| Air | 1.0000 | 1 | 1.0000 | FALSE | dry |
| Downstream | 0.0000 | 0 | 0.0000 | FALSE | wet |
| Midstream | 0.0000 | 0 | 0.0000 | FALSE | wet |
| Upstream | 0.9989 | 0 | 0.4994 | TRUE | uncertain |
Table 4: Dryness scores for January 25th, 2026. Air scored 1.0 and was classified as dry. Midstream and Downstream both scored 0 and were classified as wet. Because the upstream position rose in temperature and then stagnated, it created a conflicting warming score (0.9989) and cooling score (0), creating a conflict flag and classifying the sensor as uncertain.
Analysis
Over 418 days of record, Pearlstone maintained consistent surface flow at its Downstream and Midstream locations, with both sensors categorized as wet on more than 90% of days (94.7% at Downstream and 92.8% at Midstream). The upstream location was wet on 86.6% of days, the lowest of the three, and its dry and uncertain days were concentrated in the final months of the record, from about August 2026 through October 2026, as shown by the cluster of red and gray bands at the right edge of the Upstream row in the overall daily dryness heatmap. Before that period, Upstream was almost continuously wet, including through the first summer of record. This pattern is consistent with a near-perennial headwater stream in which the upstream position begins to dry during late summer baseflow recession, while the main channel maintains flow. Because this drying appears in only one of the two summers sampled, and only at the sampled points, it is not clear whether the upstream reach is seasonally intermittent or whether the late summer 2026 dryness reflects a drier than usual year or the placement of the sensor.