Conducting Hydrologic Research with Hobbyist Electronics: Gaining New Insights
into Stemflow Processes with Low-Cost Custom Sensor Platforms
Brandon Turner, David J. Hill, Darryl E. Carlyle-Moses
Department of Geography and Environmental Studies, Thompson Rivers University

Abstract:
The need for greater spatiotemporal resolutions has been a key constraint in the study of the hydrology
of vegetated surfaces. Of particular importance to further understanding is the capability to link intrastorm dynamics of interception of precipitation to the meteorological conditions which produced it.
Within the study of interception, the dynamics of precipitation routed from the canopy down a plant’s
stem (stemflow) represents a significant and highly variable localized input of water. In this study, a
“maker” approach is applied to demonstrate a new sensing system which leverages the low-cost and
accessibility of hobbyist electronic components in the monitoring of stemflow production. A case study
highlighting the performance of the new sensing system serves to demonstrate the potential utility and
effectiveness of these low-cost platforms in linking the intra-storm variability of stemflow production to
the overall meteorological conditions.

1

Introduction
The increasing availability and affordability of low-cost hobbyist-grade electronics, such as the Arduino
microcontroller, Raspberry Pi microcomputer, and XBee wireless radio, that has accompanied the
popularity of the “maker movement” presents an opportunity to improve spatial and temporal resolution
of environmental sensing (Hill, et al., 2014). As such devices become increasingly powerful, smaller, and
cheaper, it will be possible to embed these devices more deeply within urban and natural systems—
enabling the measurement of key parameters of these system at unprecedented spatial and temporal
scales. These high spatiotemporal resolution data have the potential to transform our understanding of
hydrologic processes and ability to adaptively manage water resources systems (e.g., Hill 2015). Despite
the importance of how precipitation is intercepted or transmitted to the land surface by vegetative
canopies, however, there are few suitable measurement devices that can be deployed in spatially-dense
networks at the tree-scale. The interception of precipitation by plant canopies is an important component
in the hydrology of vegetated environments. Two primary phenomena of interest within the study of
interception are throughfall, (i.e., the precipitation that falls to the ground from an overhanging canopy)
and stemflow (i.e., the portion of precipitation routed down a plant’s stem). Recently, there has been a
push to investigate stemflow and the environmental factors that govern its initiation depth, total volume
and chemical characteristics, because it is now recognized that stemflow plays a significant role in spatially
localized ecohydrology (Levia & Germer, 2015). In particular, stemflow is important in the study of
interception, because it embodies an input of water and nutrients at a finite point in space (i.e., the plant
stem) (Levia Jr. & Frost, 2003), and thus, is a key factor in processes ranging from soil erosion to
groundwater recharge.
Despite having been investigated for more than a century, stemflow measurement technology has
changed little over the years. Typically, a collar is wrapped around a tree’s bole several times and this
collar collects stemflow and discharges it into a collection vessel (Horton, 1919; Levia Jr. & Frost, 2003).
The volume of collected stemflow is then measured and compared with the meteorological conditions
that produced it. When unassisted by some form of gauge or metering device, this method of
measurement of the collected stemflow is performed at the end of the storm. This is done to avoid
disrupting the flow of stemflow into the collection vessel. Thus, the measurement produced is a stormtotal stemflow volume or storm-scale average stemflow rate, which cannot be used to explore intra-storm
mechanics affecting the discharge rate of stemflow. Furthermore, measurements must occur soon
enough following the rain event so that loss of the collected stemflow to evaporation is minimized. Thus,
evaporation rates dictate the number of trees that can be accurately measured by a single person, limiting
the spatial resolution of stemflow measurements within a stand of trees.
In recent years, new forms of technology have been introduced into the study of stemflow. When
attempting to link specific environmental factors, such as rainfall inclination, to the generation of
stemflow, some researchers have leveraged tipping bucket gauges (Van Stan II, Sieger, Levia Jr., & Scheick,
2011). These gauges offer the ability to collect valuable data regarding stemflow generation throughout
an event. While a significant leap forward, these devices still suffer from mechanical limitations (i.e. they
require a bucket to fill before a data event can be logged), high expense (limiting wide spatial deployment)
and weak or average temporal resolutions (typically 5 minutes) (Iida, et al., 2012).

2
This study develops a stemflow sensor system
that can provide high spatial and temporal
resolution measurements of stemflow rate. This
sensor system consists of nodes that can be
incorporated into an ad-hoc network via wireless
communication, and which can wirelessly
stream collected data in near-real time to an
Internet repository.
Each sensor node is
individually powered and capable of measuring
the volumetric flow rate of stemflow at a
frequency of 0.1 Hz – a significant increase in
temporal resolution compared to storm-scale
averages. Furthermore, due to the relatively low
cost of the nodes and their ability to stream data
wirelessly, this sensing system can be scaled to
large deployments without requiring a
substantial number of personnel to be deployed Figure 1: Illustration of sensor network functionality
in the field. The next section describes the
sensing system components and the design of each sensing node. A case study will then be presented, in
which the platform was deployed to monitor a rainfall event. Finally, the performance of the sensor node
will be evaluated for measuring stemflow and conclusions given.

Methods
The project began with a distinct “maker” attitude and perspective. The goal was to leverage low-cost,
readily available hobbyist electronics in a manner that would produce a robust research capable platform.
To this end, we made deliberate selections for hardware and software which ensured that all portions of
the methodology could be reproduced, adapted or enhanced for other environmental monitoring
applications. The stemflow sensor system developed in this work is a network of sensor and gateway
nodes that connect with each other to backhaul data to the Internet as illustrated in figure 1.

Sensor Node
The sensor nodes are autonomous devices that can be deployed at the base of trees to measure the
stemflow generated by the trees during a rain event based on the rate of change of water depth in a
collection vessel over time. As illustrated in Figure 2, the sensor nodes are composed of a stemflow
collection vessel, stilling well, and sensor housing. Stemflow is routed from the tree into the collection
vessel, and a stilling well is used to dampen waves caused by turbulence within the vessel during filling.
At the top of the stilling well is a sensor housing that serves as a weatherproof enclosure for electronics
and a mounting point for sensors. Each sensor node also contains a 12 V, 20 Amp deep cycle battery
charged by a 20 W solar panel, a sufficient power supply to power the sensor indefinitely.
The primary occupant of the sensor housing is an Arduino Uno microcontroller, a programmable device
designed to handle electronic signals from the sensors with minimal processing overhead. Prior to the
maker movement, to design a platform such as this would require the use of a bare microcontroller
integrated circuit (colloquially known as a “chip”) around which a circuit board would be engineered to
handle everything from power delivery to signal inputs, a non-trivial task requiring a very strong

3
understanding of electronic design. Arduino is an open-source hardware project which offers a number
of ready to use microcontroller units and a simplified software programming interface that significantly
lowers the barriers to entry when designing electronics.
The Arduino Uno in the sensor node manages the data collection and processing from three sensors: a
wetness sensor connected to the analog interface, an ultrasonic rangefinder connected to the PWM
interface, and a temperature/relative humidity sensor connected to the digital interface. In addition, we
added an XBee interface shield and radio module attached to the UART interface of the Arduino to enable
low-power radio communications for near-real-time data streaming. All three sensors and the XBee radio
modules were wired to a simple custom interface board equipped with reliable connectors and soldered
joints. Using a board, rather than wiring the components directly to the Arduino reduced strain on the
connecting wires and made it simpler to manage power distribution to all modules as they all rely on the
same 5 V output of the microcontroller.
The ultrasonic rangefinder (Model: Maxbotix HR-LV EZ3) integrated into the sensor node measures the
distance from the sensor to the surface of liquid in the stemflow collection vessel. The ultrasonic sensor
used in this study has a minimum measurement distance of 300 mm and a measurement resolution of 1
mm. As shown in Figure 2, the sensor was suspended from the base of the sensor housing which was
mounted onto an ABS pipe perpendicular to the bottom of the collection vessel. In early prototypes, the
ultrasonic device was simply pressure fitted into an opening at the bottom of the housing. We found that
this design was heavily influenced by environmental ultrasonic noise (primarily from passing vehicular
traffic), because the housing acted as an amplifier transferring the ultrasonic noise directly to the sensor.
To overcome this issue, the design was modified to allow the ultrasonic device to be mounted beneath
the sensor housing within the confines of the stilling well. Nylon spacers were used to distance the
ultrasonic rangefinder from the housing and minimize interference.
We chose a remote sensing solution (i.e., ultrasonic depth measurements) for stemflow volume
measurement because such a device would be less prone to mechanical failures than a device that
required physical interaction with the collected stemflow. An issue inherent with using ultrasonic devices,
however, is their tendency to report erroneous data due to sonic interference or random internal
electrical noise. To mitigate these sources of noise, a mode-filter was implemented as part of the program
deployed on the Arduino. The filter takes a set of 121 raw depth measurement samples every 10 seconds
and reports the mode of the distribution of sampled depth values as the characteristic depth during this
10-second period. This filter, however, is not able to overcome sustained external noise.
The ultrasonic sensor was located inside a stilling well so that the ultrasonic signal has a relatively flat
surface to reflect off. The stilling well, however, creates a high humidity environment (during rain events)
in which temperature may fluctuate widely. These changes in the temperature and humidity of the air
within the stilling well will impact the speed of the ultrasonic signal as it travels from the sensor to the
surface of the liquid in the collection vessel and back. Because the range between the sensor and the
liquid surface is derived from the travel time of the signal as it reflects off the liquid surface, it is necessary
to adjust the measurement to account for variations in the speed of sound.
The ultrasonic sensor utilizes an in-built temperature compensation to overcome variations in the speed
of sound in the recommended operating range of -15 °C to +65 °C. Temperature and relative humidity
measurements were acquired by a separate temperature/relative humidity sensor (Model: DHT22) as
well. This sensor is utilized to monitor humidity in the event that a post-hoc temperature or humidity

4
correction needed to be applied to the data. As shown in Figure 1, this sensor is mounted on the bottom
of the housing inside the stilling well near to the ultrasonic sensor.
Finally, a wetness sensor was used to acquire accurate measurements of the timing of stemflow initiation
and cessation, which are necessary to accurately correlate meteorological conditions with the dynamics
of the monitored tree. The ultrasonic device, while effective at monitoring changes in the depth of the
liquid in the collection vessel, lacks the measurement precision to identify small changes in the collection
vessel’s depth due to the entrance or tapering off of small volumes of collected stemflow associated with
the initiation and cessation of stemflow, respectively. For this reason, a wetness sensor (Model: FC-37)
was attached to the outlet of the stemflow collar to mark the point in time when water begins/ends
flowing into the collection vessel.. The wetness sensor is a variable resistor created by a circuit board with
interlaced, non-connecting copper traces. One trace is supplied a constant voltage while the other is
attached to ground. When water passes over the surface of the circuit board, it completes the contact
between the two traces and causes an instant and measurable change in the electrical resistance. In the
interest of reduced complexity, we recorded the changes in a binary manner, either wet or not wet,
however the sensor is capable of estimating how much of its surface is covered by water.
The final component incorporated into the sensor platform is the XBee radio. XBee is a low-power,
moderate range radio created by Digi International. These devices follow IEEE 802.15.4 Wireless
standards, which is the guideline set for low-rate wireless personal area networks. Though a proprietary
technology, the XBee platform shares the accessibility and community support which makes the Arduino
environment attractive. The model of radio chosen for this study (XBee-PRO 900HP) uses a proprietary
implementation of the ZigBee network protocol, known as DigiMesh, that simplifies use and expands
functionality. The radios can transmit at distances up to 1,600 m and draw minimal power making them
ideal for remote, battery dependent deployments. Data collected by each sensor node can be streamed
by the XBee radio to a gateway node.

Gateway Node
The gateway node is designed as an Internet access point for the sensor nodes. It is based on a Linux
microcomputer (Raspberry Pi) which receives data from a USB connected XBee module and transmits
collated data to the Internet via 802.11 wireless networking. The gateway node implements a messaging
protocol (RabbitMQ) which allows it to queue data for transmission to a variety of data services. Using a
locally installed messaging protocol on the gateway node allows for a degree of fault tolerance in the
event of internet connection failures. The gateway node also serves as a command and control point for
wireless nodes, allowing remote access to enable or disable specific nodes should the need arise.

5

Case Study
Study Site
To test the sensor system, we deployed a single sensor node at the base of a tree on the Thompson Rivers
University (TRU) campus in Kamloops, British Columbia, Canada. Our criteria for site selection were: (1)
the tree must be standing alone with no other canopy overhanging or obstacle present; (2) the tree must
produce minimal to average volumes of stemflow; (3) a meteorological station could be sited near the
deployed sensor while at the same time observing appropriate setbacks from nearby obstructions. The
tree selected is a juvenile green ash (Fraxinus pennsylvanica) tree near the main road which
circumnavigates the TRU campus. This tree fit our selection criteria well, the species was known to
produce weak stemflow volumes (Schooling & Carlyle-Moses, 2015), it is isolated from other trees and it
was near a suitable location for weather station deployment. The location of the tree and meteorological
station are illustrated in Figure 3. The tree is maintained by the university grounds staff and is regularly
irrigated by means of a low-spray sprinkler located at its base. This tree has a diameter at breast height
(DBH) of 10.8 cm and a height of 7.4 m.

Figure 2: Study location on the TRU Campus, Kamloops BC, Canada; 50°40'23.44"N, 120°21'53.08"W.

6
The deployed sensor node is shown in Figure 4. Stemflow
from the tree was collected by a stemflow collar
wrapped twice around the tree, starting at
approximately 1.5 m above ground level and continuing
down the tree to a final height of approximately 0.5 m
above ground level. The stemflow collar was constructed
from a length of flexible PVC tubing that was cut in half
prior to being fastened to the tree. Once attached, the
portion of the collar abutting the tree was covered with
silicone to prevent leaking as well as ensure a water-tight
contact. The end of the collar was attached to an
unmodified, short length of the same tubing which was
then directed into the collection vessel. The collection
vessel used in this study was a 19 L plastic construction
bucket of standard dimensionality (truncated conical
shape). This type of container was selected because it is
inexpensive and widely available.

Figure 4: Image of Deployed Sensor Platform

A large high density polyethylene (HDPE) storage tote
was used to hold the collection vessel as well as a box containing the lead-acid battery. A wooden crossbrace was mounted across the narrow middle span of the tote to increase rigidity and provide a fastening
point for the lid. The lid was made from a piece of plywood cut to fit the size of the tote. The solar panel
was bolted to this lid. A hole was drilled through the plywood directly over where the center of the bucket
rests inside the tote. Through this hole the pipe acting as the stilling well was inserted and the sensor
housing mounted on top. All connections between the stilling well and the lid were sealed using electrical
duct seal putty.
A Hobo U30 weather station equipped with a number of leaf wetness sensors, anemometer,
pyranometer, wind vane, hygrometer, thermometer and barometer was also deployed. A tipping bucket
rain gauge with a resolution of 0.1 mm / per tip was also used. The data gathered from these devices
was not necessary to the operation of the sensor platform developed in this project but were used to
support the analysis of the data collected by the sensor node.

Calibration of Sensor Node
Because the 19 L bucket used as a collection vessel has a truncated conical shape with a slightly convex
base it was necessary to establish an empirical calibration that relates liquid volume and depth within the
bucket. Data was collected from a series of 10 repetitions of an incremental filling of the container,
wherein 100 mL of water was added each minute to a total volume of 5 L. After each addition, the
ultrasonic distance measurement was noted. As shown in Figure 5, the depth and volume data possess a
strong linear correlation, and linear regression was used to model the calibration curve. This calibration
indicates that each 1mm of water depth corresponded to approximately 53 ml, with a standard error of
23 mL, of water in the collection vessel. The calibration also revealed that the sensor was not capturing
the first approximately 100 mL of incoming water, a result of the convex bottom.

7

Figure 5: Linear Regression of depth vs. volume for calibration.

May 4, 2016 Event
To illustrate the performance of the stemflow sensor system developed in this work, data collected during
a rainfall event on May 4, 2016 will be analyzed. As shown in the hyetograph in Figure 5, the event
produced a total of 5 mm of rainfall over 4.93 hours with a peak intensity of 13.5 mm/h. The start of the
event was determined using the leaf wetness sensors attached to the meteorological station. The clocks
on both the sensor platform and MET station were synchronized the internet clock.

8
16.00

Intensity (mm/h)

14.00
12.00
10.00
8.00
6.00
4.00
2.00
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
2.20
2.40
2.60
2.80
3.00
3.20
3.40
3.60
3.80
4.00
4.20
4.40
4.60
4.80

0.00

Hours (Since Storm Start)
Figure 6: Hyetograph of May 4, 2016 Rain Event

Based on measurements from the wetness sensor on-board the sensor node, illustrated in Figure 7,
stemflow initiation was estimated to have occurred approximately 50 minutes into the event after a total
rainfall depth of 0.9 mm had fallen. This time of initiation was estimated as corresponding with the first
drop in voltage recorded by the sensor. Further fluctuations in voltage likely correspond with changes in
stemflow rate and/or splashing of stemflow as it is routed into the collection vessel.
Stemflow measurements from the sensor node were processed using Kalman Smoothing prior to further
analysis. The smoothing was performed to reduce the moderate, environmental noise contamination in
the ultrasonic depth measurements. Once smoothed, all depth measurements were converted to
volumes using the calibration curve developed for the sensor. The smoothed cumulative stemflow
volume data derived from the sensor node are plotted along with the cumulative rainfall volume
measurements from the weather station in Figure 8. As illustrated by this figure, the stemflow appears
to follow the same general pattern of intensity as rainfall, albeit lagged by around 1 hour. Notable in the
stemflow data is the dips where there appears to be negative stemflow. These dips are a result of
variation in the measurements by the ultrasonic sensor and are not indicative of some net loss of
stemflow.
Final stemflow volume derived by the sensor was compared to a physical measurement of the same
quantity made using a graduated cylinder following the event. The sensor registered a total final volume
of 1.6 L which was very close to the physical measurement of 1.65 L. This total volume of stemflow was
also used to calculate the stemflow funneling ratio (Equation 1) of this tree during this event:
𝑉𝑉
(1)
𝐵𝐵 ∗ 𝐺𝐺
where, 𝑉𝑉 is the volume of collected stemflow, 𝐵𝐵 is the basal area of the tree and 𝐺𝐺 is the gross
precipitation. The product of 𝐵𝐵 ∗ 𝐺𝐺 represents the equivalent volume of water one would expect in a rain
gauge with an equivalent area given the same rainfall depth. The funneling ratio approximates the
amount of concentration of rainfall that has occurred given the canopy of the tree (as well as other factors)
(Levia Jr. & Frost, 2003; Herwitz, 1986) and was found to be 40.1.
𝐹𝐹 =

9

Figure 7: Wetness Sensor Voltage Measurements, Lower values represent greater coverage.

Figure 8: Cumulative Stemflow Volume and Cumulative Rainfall Depth.

10

Discussion
The May 4th event presented in this paper represents the first of many of events that were measured by
a sensor node during two 3-month summertime deployments on several green-ash trees including the
one described in the case study. Through this initial deployment, care was taken to validate the data
gathered from the sensor node with hand measurements taken immediately following the rain event
Based on this validation of the data being produced by the sensing system, we found that the stemflow
sensing system performed well except in rare cases where the sensor was compromised in some way by
an external force (e.g., water leaking into boxes). Based on lessons learned from these sensor node
failures, the design has evolved to improve reliability and resiliency. Evolutions have included switching
from a custom 3D printed enclosure for the sensor to a mass-market weather sealed electrical box. Also,
the model of ultrasonic sensor was changed from one with a wide beam angle to a narrow beam model
in the same line. This change was undertaken as the wider beam device was more prone to interference
caused from reflection off the walls of the stilling well. Overall, however, the sensing system has proved
to be sufficiently robust for long-term deployment and capable of accurately measuring stemflow at high
temporal resolution. Finally, in semi-arid environments, the temperature within the stilling well may
exceed 40 °C with a relative humidity over 70%, conditions for which the onboard compensation of the
ultrasonic device may struggle to overcome, thus, future generations of the sensor node may include a
custom humidity and temperature correction.
The May 4th 2016 event was chosen to demonstrate the science questions that can be explored using data
from the stemflow sensing system developed in this work. The data from this event exhibited fewer noise
artifacts in ultrasonic measurements than following storms. In addition, the May 4th event had the lowest
stemflow initiation depth of all recorded events thus far with a total rainfall depth of 0.9 mm initiating the
first volume of stemflow. Given the high temporal resolution, the sensor has offered a window into
stemflow dynamics that is much clearer than we anticipated at the outset. With precise stemflow
initiation and cessation times as well as near real-time discharge monitoring there is greater potential to
link stemflow dynamics to specific meteorological conditions.
Understanding how storm dynamics affect stemflow is valuable when modelling how a given plant canopy
will affect the water budget of a localized region (i.e. the base of a tree). Factors such as rainfall
inclination, drop size and velocity, wind direction and velocity, and rainfall intensity all affect initiation,
cessation and volume of discharge of stemflow. Constant stemflow requires a pathway to be established
between the canopy and the base of a plant along the stem. The previously outlined factors all affect how
that pathway forms and its durability. Currently, many of these factors are correlated using datasets
gathered at temporal intervals of 5 minutes or more which offer insights but fail to capture subtle changes
which may occur during a given event. Monitoring these subtle changes has the potential to improve
models attempting to forecast stemflow dynamics.
Increases in spatial resolutions are also needed to better model the influence of storm dynamics upon
stemflow. Common practices such as manual measurements are labor intensive making them difficult to
scale across a larger sample size while maintaining capability to collect data in a timely manner (an issue
of particular importance in arid regions). Electromechanical sensors currently used to monitor stemflow
assist with the labour issue of scalability but at a significant financial cost. The sensing system in this work
aimed to create an electronic platform that reduced necessary labour while still economically scaling
across a larger sample. Increasing spatial resolutions could potentially open further avenues of inquiry

11
into such questions as the effects of topography, canopy density influences or physiological variations
across a species.
The maker approach used in this project was critical to the success of the sensing system. Hobbyist grade
electronics are widely available and affordable, allowing for the overall cost of the platform to remain low.
Lower costs allow for more sensor nodes to be deployed which increases the potential spatial resolution.
Additionally, the relative simplicity of design allows for any interested individual to construct a sensor
node and collect data. This accessibility may open opportunities for citizen participation in future studies.
Finally, the maker approach led the project to use tools which are flexible and adaptive. At the conclusion
of a given study, it is possible to repurpose or adapt the sensing nodes for a new application, thus reducing
investment in a new suite of instrumentation.

Conclusion
The stemflow sensing system developed and validated in this work is capable of providing the high spatial
and temporal resolution data necessary to explore the role of tree- and storm-structure in partitioning
rainfall into overland flow and infiltration into the subsurface. These relationships are key in
understanding the role of forests in regional hydrology and urban plantings in urban hydrology, among
other things.
Approaching design and application from the point of view of a “maker” permitted us to assess the
potential of open source hobbyist grade electronics for environmental monitoring. Using off-the shelf
hobbyist electronic hardware and open source software (e.g. RabbitMQ) leveraged the economy of scale
provided by the large number of consumers of this equipment. For this reason, a large system consisting
of many sensor nodes and gateway nodes can be affordably constructed. Furthermore, due to the
marketing of these electronics components to people with little or no formal knowledge of electronic
design, the components are relatively easy to assemble and adapt to new data collection needs., thus, the
sensing system developed in this work could be adapted to monitor other hydrological processes. In
particular, the gateway node can receive data from a heterogeneous system of sensor nodes, all
communicating via wireless transmissions (such as the Digimesh network used in this work), and stream
these data into the Internet. Where a traditional specialized research instrument may remain unused for
many years between studies, an Arduino based platform can be repurposed with relative ease to service
a new problem. Within hydrology, Arduino microcontroller platforms have become popular as
inexpensive custom dataloggers (Hund, Johnson, & Keddie, 2016; Fisher & Gould, 2012; Pallugna, Cultura,
Gozon, & Estoperez, 2013) while their potential as a nexus for innovation within the field has largely
remained untapped.
The sensing system overcomes challenges of common stemflow research techniques. It reduces
necessary labour investment by constantly monitoring the volume of collected stemflow, reporting it to
an internet repository. Additionally, the remote sensing technique used to monitor stemflow discharge
rate benefits from not being constrained by mechanical limitations which allows for more frequent
measurements to be taken. The case study offered a clear image of the dynamics of stemflow discharge
during a storm event owing in part to its high temporal resolution. With this sensing system it will be
possible to explore existing scientific questions in the field more deeply as well as open new questions.

12
With improved spatio-temporal data, insights into topics such as the effects of storm dynamics will be
much more substantial.

References
Fisher, D. K., & Gould, P. J. (2012). Open-source hardware is a low-cost alternative for scientific
instrumentation and research. Modern Instrumentation, 1(2), 8-20.
Herwitz, S. (1986). Infiltration-excess caused by Stemflow in a cyclone-prone tropical rainforest. Earth
Surface Processes and Landforms, 11(4), 401-412.
Hill, D. J. (2015). Assimilation of Weather Radar and Binary Ubiquitous Sensor Measurements for
Quantitative Precipitation Estimation. Journal of Hydroinformatics, 17(4), 589-613.
doi:10.2166/hydro.2015.072
Hill, D. J., Kerkez, B., Rasekh, A., Ostfeld, A., Minsker, B., & Banks, M. K. (2014). Sensing and
Cyberinfrastructure for Smarter Water Management: The Promise and Challenge of Ubiquity.
Journal of Water Resources Planning and Management, 140(7). doi:10.1061/(ASCE)WR.19435452.0000449
Horton, R. E. (1919). Rainfall Interception. Monthly Weather Review, 47(9), 603-623.
Hund, S. V., Johnson, M. S., & Keddie, T. (2016). Developing a Hydrologic Monitoring Network in DataScarce Regions Using Open-Source Arduino Dataloggers. Agricultural & Environmental Letters,
1(1). doi:10.2134/ael2016.02.0011
Iida, S., Shimizu, T., Kabeya, N., Nobuhiro, T., Tami, K., Shimizu, A., . . . Keth, N. (2012). Calibration of
tipping‐bucket flow meters and rain gauges to measure gross rainfall, throughfall, and stemflow
applied to data from a Japanese temperate coniferous forest and a Cambodian tropical deciduous
forest. Hydrological Processes, 26, 2445-2454. doi:10.1002/hyp.9462
Levia Jr., D. F., & Frost, E. E. (2003). A review and evaluation of stemflow literature in the hydrologic and
biogeochemical cycles of forested and agricultural ecosystems. Journal of Hydrology, 274, 1-29.
Levia, D. F., & Germer, S. (2015). A review of stemflow generation dynamics and stemflow-environment
interactions in forests and shrublands. Reviews of Geophysics, 53, 673-714.
doi:10.1002/2015RG000479
Pallugna, R. C., Cultura, A. B., Gozon, C. M., & Estoperez, N. R. (2013). Low-cost remote rain and stream
data acquisition system for mapping of potential micro-hydro sites. Energy and Power
Engineering, 5(04), 56-62.
Schooling, J., & Carlyle-Moses, D. (2015). The influence of rainfall depth class and deciduous tree traits on
stemflow production in an urban park. Urban Ecosystems, 18(4), 1261-1284. doi:10.1007/s11252015-0441-0

13
Van Stan II, J. T., Sieger, C. M., Levia Jr., D. F., & Scheick, C. E. (2011). Effects of wind-driven rainfall on
stemflow generation between codiminant tree species with differing crown characteristics.
Agricultural and Forest Meteorology, 151, 1277-1286.