Approach

We have organized our development and experiments in four areas.  The first is the hardware-based Microclimate Sensor Array.  This includes sensor testing and deployment for above and below ground environmental sensors, avian nesting sensors, and plant ecophysiological sensors, as well as the electronics, communications hardware and system software to control and record the operation of the sensors in the field.

The second area is the Image Collection System. Our objective is to develop an automated image collection system that will support a variety of soil, plant, and animal studies. Currently we are capturing images of nest box activity, plant growth, moss hydration status, and scene observation at the James Reserve and at a newly deployed site at the Deep Canyon Desert Research Station. These images are used in biological investigations, and we also plan to use them as training images for automatic categorizing images by the activity in the image.  If we are successful with the automatic categorizing we hope to use the algorithm to on a processor co-located with the image capture hardware so that we can categorize the image in real time and avoid the requirement for high bandwidth connection to the image capture sites.

The third is the TEOS Data Management System (DMS), a database containing the Automated Image Collection library, data sets including the Micro Climate Sensor Arrays, and the associated data injection, manipulation and retrieval script libraries.  New quality control tools have been added to examine aspects such as data gaps and to correlate these among the many data formats stored in the database.

The fourth area is Initial Science Investigations.  Plants at the James Reserve present an excellent model system for the utilization of CENS and NIMS technologies for science-driven applications.  Data collection on plant ecophysiology (rhododendron, the star moss, and bracken ferns), insect abundance and diversity, and cold air drainage, and soil surface energy fluxes at JR using imaging systems as well as environmental and physiological measurements using fixed sensors as well as mobile NIMS sensors have been made to examine patterns of phenology and ecophysiology across environmental gradients.  Continued studies with this system in 2006, will utilize existing CENS and NIMS installations at JR, and provide scientifically significant use-cases.

Systems/Experiments

(1) Micro Climate Sensor Array (CMS/ESS/CAD).  Quadrupled the number of wireless microclimate sensor sites and continued to operate the original wireless sensor site, now in its fourth year of data collection.  Added power and Ethernet (wired and wireless) to study areas located throughout the James Reserve..  All of the collected data is stored in local databases and available to researchers throughout the world.  Many of the reports contained within this document provide excellent examples of how this data is both used and viewed.

Figure set 1. On the left is an original CMS weather station that is now being used as a reference to compare newer technology hardware and firmware developed for the ESS systems.  Each weather station consists of a number of standard sensors.  Pictured are: rain, wind speed, wind direction, temperature, humidity and leaf wetness.  On the left is part of the Cold Air Drainage (CAD) transect.  By placing 27 of these sensors through the James Reserve, we have dramatically increased our wireless sensor coverage.

Figure set 2. In spite of being struck by a falling tree, CMS Weather 4 has continued to produce data.  All of the original CMS motes (now in their fourth year of deployment) are beginning to experience more frequent failures, mostly in the form of lost connectivity.  With recent ESS successes, our hope is to replace CMS within the next six months.

(2) Automated Image Acquisition.  We installed several wildlife observation cameras at remote UCNRS sites.  Images are collected automatically, documenting raptor nesting locations and bighorn sheep activity. Purchased and installed a 5TB disk array for storage of images and other data.  Began using two additional tower cams to assist with phenology studies of local, native plant species. These images along with surrounding microclimate and soil data are being collected automatically and made available through a web accessible database. Began experimentation utilizing a mote-based camera (Cyclops) for capturing and wirelessly transmitting nest box images.

Figure set 3. Cameras at Deep Canyon, Agave Hill, and a remote watering hole are challenging to install and maintain.  Aside from environmental factors such as weather and animals, consideration has to be given to sources of power, network connectivity and even potential vandalism.)  Bighorn sheep are one example of why an unattended observation system is such a powerful aid to researchers.  Capturing an image like this might take weeks of sitting in 100F temps, whereas a networked camera allows easy access from the comforts of an office – with far less disturbance to the research object.  New technologies (such as the mote-based Cyclops pictured below in center) require less infrastructure for deployment and allow researchers to better protect cameras from external factors.)  If image quality and sample rates available from Cyclops can meet researcher needs, their low cost and flexibility of installation will allow images (like the test picture below right) to greatly increase sample sizes and coverage areas resulting more accurate statistical analysis in a variety of plant and animal habitats.

We are also working with the CENS Systems Group to contribute test and training imagery in support of the CYCLOPS Imaging sensor node development project. CYCLOPS, which is described under the Systems Research section of this report, is intended to provide a technology that will greatly improve our ability to collect and process in-situ images relevant to our biological investigations. 

Utilities and Networks: The James Reserve's local and wide-area network system (wired and wireless), data storage systems, and our locally generated electrical power distribution systems were maintained and upgraded.  An increase in photovoltaic output and efficiency was affected by upgrading the PV panels that supply power to the off-the-grid James Reserve.

Figure set 4. About two-thirds completed when the picture was taken, the new PV array at JR increased our energy harvesting capacity by about 50% using the same footprint as the previous 10 to 30 year-old panels.  By adding a high-gain wireless access point to the AMARSS tower, we were able to better support the NAMOS research at nearby Lake Fulmor.)

We began a yearlong experiment to evaluate the effects of microclimate and foliage on RF communication reliability at frequencies used by our wireless sensors.  Also, a six-month effort was made to evaluate 24 hour aspirated, solar aspirated and non-aspirated temperature and humidity housings.

Figure set 5. We were able to utilize already existing mote housings to setup an RF communications transect to study the effects of environment and foliage on signal propagation.  By determining what variables affect such propagation, we may be able to make future networks adaptable to their environment and therefore more energy efficient, allowing researches more flexibility in their deployments.  On the right is an experiment to determine the differences between aspirated and non-aspirated temperature and humidity enclosures.  The solar panels and battery enclosure pictured were originally deployed as part of the prototype measurement system (PMS) in late 2002.  Reusing existing infrastructure for new experiments helps extend our research dollars.

Installed two microphones and a live, audio streaming device to the acoustic array tower and began streaming 24/7 stereo audio to augment the 24/7 video of the acorn woodpecker habitat.

Figure set 6. The top of the 30’ tower had to be modified and a special bracket built to accommodate the two microphones.  Each microphone was placed inside a wire housing (blimp) using foam and then surrounded by Gore-Tex bag to protect it from moisture.  To reduced wind noise, an additional bag made from two-inch long fake fur was then placed over the Gore-Tex.  A Plexiglas panel covered with neoprene separates the microphones to provide better stereo separation.  The microphones are so sensitive that they occasional pick-up vehicle noise from the highway more than 1km away and behind them.  By combining audio and video information, our long-term goal is to allow researchers to more accurately both localize and identify wildlife activity within a given habitat without their own presence influencing the activity under observation.  Currently the video portion is being used to collect both plant and avian activity while the audio is being streamed via the Internet for individual scientists to store locally for analysis.

Installed over 1,000’ feet of low voltage wire and Ethernet cable to support various experiments within the NIMS2/AMARSS transect.  Also installed a wireless access point on top of the video observation tower to support experiments at Lake Fulmor.  Performed field repair on the NIMS1 cable robot to allow image capture experiments to continue throughout the season.

Figure set 7. To support soil respiration experiments power was installed in the NIMS2/AMARSS transect.  The photo on the left shows the NIMS2 robot above one of the soil study areas.  As with all equipment deployed for an extended period, the NIMS1 robot needed, on-site repair to perform its task of collecting images on a regular basis.  On the far right is the NIMS1 node parked during a winter storm.

(3) Data Management System (DMS).  The TEOS Data Management System (DMS) is a database containing the Automated Image Collection library, several data sets including the Micro Climate Sensor Arrays and the associated data injection, manipulation and retrieval script libraries. Previously named the 'CMS' database, as it formerly contained only data from the CMS-MICA wireless motes, the new 'DMS' was expanded this year to store additional sensor data and image captures. Newly-added datasets include the Onset-brand HOBO micro-climate data loggers, the Cold Air Drainage (CAD) and additional image captures from in and outside the James Reserve.

Given the impending addition of more streaming data from several other system (ESS, CAD, NAMOS, NIMS) Michael Wimbrow created the 'AdminGUI' tool (Figure 1) for the DMS database. This web application walks users through the process of adding new site and sensor installation to the database, as well as a system for loading data directly from files, as in the case of the HOBO data loggers.

Figure 8 'AdminGUI' CENS Database Administration Interface

Quality analysis and quality control for sensor data streams:

With the addition of new data streams, and the occasional human and systems errors causing data loss, the need for a quality analysis and control became evident. Sean Askay created a number of scripts that traversed each data stream in the DMS to look for gaps. Though we knew that our CMS and HOBO systems have occasionally fallen victim to lightning strikes, inclement weather, hardware malfunctions and human error, we had little idea of the extent of data loss. The exact times and cause of such gaps in data are important to identify because data loss during certain periods can be fatal to certain scientific studies. Figure 9 shows the data availability and absence in our primary CMS micro climate system as an example of this tool.

Figure 9. Data gaps in excess of 12 hours throughout the life of the Campbell Scientific Weather Station (1999-present). Green signifies data available, black indicated missing data.

Most of the data gaps that occur simultaneously across all CMS motes were due to system-wide problems, i.e. network problems, database/server issues, etc.  Gaps that were specific to individual motes are likely caused by wireless connectivity problems, condensation on circuit boards, a few known lightning strike events (and subsequent equipment replacement). It is interesting that 'Weather 4' was frequently plagued with wireless communication problems, and in one case, damage from a falling tree; it has the most data gaps.

Gaps in the HOBO system are not all the result of hardware issues or human error. Some of them were planned breaks in service, or were the result of configuration changes. These different types of gaps  are not distinguished from hardware problems (lightning strikes, sensor failure) or human error (in the downloading of dataset from the data loggers) in the data analysis tool. Our next step in the QA/QC process is to accurately identify the causes of all gaps to ensure future problems can be avoided.

The tool has allowed us to examine the frequency of data gaps in the CMS system, and compare them to the unscheduled gaps in the HOBO system; losses were much smaller and usually identified and rectified much sooner in the CMS system. The time commitment involved in harvesting data from the HOBO loggers limits collections to about once a month. In the earlier days of the HOBO system, data was sometimes only downloaded every 3-9 months. Thus, unfortunately, we did not discover problems until large amounts of data were already lost. This clearly demonstrates the advantages of a streaming sensor network system in that we are able to identify problems quickly by monitoring the real-time data streams.

(4) Initial Science Investigations. Plant and animal ecophysiological studies. Phenology of native azaleas, Rhododendron occidentalis, and bracken fern, Pteridium aquilinum, and carbon modeling of the star moss, Tortula princeps, were recorded using NIMS mobile imaging, robotic fixed video on the Trailfinder micrometeorological tower, a fixed camera installation (“MossCam”), and temporary arrays of temperature, photosynthetically active radiation (PAR), and soil moisture sensors.

Moss cam: The moss Tortula princeps undergoes changes in reflected visible light during cycles of drying and hydrating in the field and the MossCam Project has collected digital images of T. princeps at least daily since 2003.  Laboratory studies can be used to calibrate these images to indicate field physiological conditions.  Drying the moss 6 days in the laboratory resulted in a decrease of net CO2 uptake to near zero; recovery after rewetting occurred within 10 minutes

Drying the moss 6 days in the laboratory resulted The difference in reflectance between hydrated and dry T. princeps was maximal around 550 nm and maximal net CO2 uptake was linearly related to the green:red ratio of laboratory images when net CO2 uptake was positive.  Using the green:red ratio of field images and otherwise assuming ideal conditions, the total carbon gain for a six day period around a 1.3 mm rain event was about 208 mmol CO2 m-2, equivalent to 69 days of respiration under dry conditions

Figure 10. The MossCam Project installation, showing the white environmental housing that covers the networked digital video camera that is pointed at a stand of Tortula princeps on a boulder at the James Reserve in the San Jacinto Mountains of southern California, USA.  (A) November 10, 2005, 1800 h after about 20 days of no precipitation and (B) the same day, 1900 h after 0.5 mm recorded precipitation.

Figure set 11. The percentage of maximal net CO2 uptake related to the green:red ratio calculated from images taken in the laboratory simultaneously with gas exchange and the resulting predicted net CO2 uptake for T. princeps at the James Reserve before and after a rain event based on this green:red ratio.  Images were captured in the field between 18 – 23 August 2003 by the MossCam.

Bracken ferns: Ferns at the James Reserve present an excellent model system for the utilization of CENS and NIMS technologies for science-driven applications.  Data collection on bracken ferns at JR was begun in 2004 using imaging systems as well as environmental and physiological measurements to examine patterns of phenology and ecophysiology across light gradients at the forest edge.  Continued studies with this system in 2005 involved measuring their responses to sunflecks and in the coming year will utilize existing CENS and NIMS installations at JR.

The working hypothesis is that the fern as a whole has competing goals of maximizing its carbon gain and minimizing its water loss through acclimation of individual fronds to their local environments.  Exposed fronds (sun-fronds) may be acclimated to conserve water as a priority over carbon gain whereas fronds occurring more in the understory (shade-fronds) may be acclimated for carbon gain at the expense of more water lost. Thus, for sunfleck use, water conservation takes priority over carbon gain for sun-fronds whereas for shade-fronds light flecks will contribute substantially to their carbon budget.

Figure 12 indicates that stomatal closure happens relatively quickly after even only 5 minutes of shade to conserve water in sun fronds.

Modeling sunflecks by multiscale techniques is now underway and should result in usable models applicable to ferns in the field at the James Reserve. Recent work at La Selva biological station in Costa Rica also concentrated on the effects of sunflecks on different forms of Philodendron (terrestrial, hemiepiphytic, and epiphytic species) and was directly related to questions of sunfleck use efficiency.

Figure 13 is a combination of two curves, one smooth curve indicating that photosynthesis increases with carbon dioxide levels.  That curve is intersected by a shorter one measured during inductance.  It’s low intersection point indicates that photosynthesis during induction is limited by stomatal opening and not by enzymatic induction, a sign of water conservation over carbon gain.

Insects: Insect abundance and diversity is based on infrequent trapping at limited locations and is labor intensive.  One solution to this limitation is to use articulated NIMS imaging and automatic analysis to count insects “trapped” by various collection techniques.

Figure 14 indicates the resolution capabilities of the NIMS imager.

Sampling of insects can occur over 24-hours at multiple locations throughout the year.  The NIMS camera was tasked every night to take images of a sheet illuminated by a black light for a summer intern.  The questions that were addressed in the interns pilot study included (1) How temperature and humidity affect the relative abundance of Coleoptera (beetles) and Lepidoptera (moths) and (2) What was the average residence time of an insect visiting the light trap.

Based on direct observation of the light trap the average residence time of Coleoptera and Lepidoptera is 34 minutes and 36 minutes respectively. This data indicates that capturing an image of the light trap every five minutes is a sufficient sample to determine relative abundance of insects.

Cold air drainage down Hall Canyon measured with wireless sensors.  A wireless network of sensors was used to characterize a cold-air drainage event in the canyon surrounding the James Reserve. Cold air masses have been shown to affect the distribution of freezing-intolerant species as well as species that have differential growth due to ambient conditions including relative humidity.  The flow of cold air at night and the first hours of sunrise have major ecological consequences by limiting the vegetation types to those tolerant of freeze and thaw cycles. A network of wireless sensors provides the opportunity to track this event in real time and fully characterize the cold air flow down the canyon, which may last 1.5 hours, and the pooling of cold air in low lying areas. Cold air drainage generally occurs in cycles; oscillations of 1.5 hours with flow often between sunset and midnight, and coldest during early mourning.  Cold-air flow is only generated on clear nights with low wind speeds.  A thermal belt and transitional zone with higher temperatures will occur above the cold-air pool.

Figure 15. Multiple mote data from August 7 (noon) to 8 (noon), 2005. Cold-air event most notably occurred at 1 am and lasted 1.5 hours.

For the Cold Air Drainage deployment, we recommend spacing mote sensors 10 m apart, based on semivariogram results.  Data acquisition every 5 minutes can still be used to characterize a cold-air drainage event.  Mote sensors deployed further up in elevation would be necessary to capture all the characteristics of the cold-air pool.  Sensors along the cold air flow path would be necessary to model a cold-air drainage event as it happens.

Figure 16. A model of the cold air lake that forms within the James Reserve following Yoshino (1984). The two transition zones (indicated in light blue) form just below a thermal belt of warm air, where cold air has drained away from the slopes. Areas above the thermal belt are progressively cooler with the increase in altitude, indicated as the lapse rate zone.

Soil surface energy flux.  NIMS 2 is now equipped with an infrared thermometer that can measure surface temperatures remotely.  The first scan of surface temperatures occurred on February 10th 2006 for approximately 25 meters of terrain.  The sensor was moved outward and then back along the transect and recorded position, images, and surface temperatures  Differences between the outgoing and returning temperature values indicate that large variations over short time scales will most likely require adaptive sampling techniques in development with other projects at CENS.

Figure 17.  Shading greatly influenced the surface temperature, as was expected.  Work with mapping soil temperature variations over 24 hours can now begin, which is the first step towards obtaining a complete energy  balance description of the soil along the AMARSS transect.

Accomplishments

We have quadrupled the number of microclimate sensors at the James Reserve and maintained and upgraded our existing sensor and infrastructure systems.  We upgraded the wireless network infrastructure allowing researchers access to connectivity over larger environmental areas, in particular, at Lake Fulmor in support of NIMS and NAMOS field tests.  We’re evaluating in-house low-cost aspirated and non-aspirated housing to both better understand our data and to potentially improve future, sensor deployments.

We installed additional wildlife observation systems and microclimate sensors on towers at several locations throughout the James Reserve.  We also deployed networked cameras at two other, remote Natural Reserve System sites and installed a 5TB file server.  Field trials for the new CYCLOPS video-mote have begun, including a deployment with several avian nest boxes, and pit fall traps. Both of these applications will be expanded through the spring and summer of 2006, and algorithms for automated animal detection are being developed and tested.

New tools for automated data quality assurance and quality control of hundreds of sensor and image data streams were incorporated into the Data Management System. The debut of the TEOS Google Earth interface at the CENS annual research review generated Center-wide interest in the use of Google Earth for the visualization of sensor data from other CENS projects. Work has started towards the integration of statistical analysis tools into a combined database portal and data viewer as part of our Google Earth web browser.

Carbon balance modeling of star moss, T. princeps, has finished its first phase, resulting in a publication (Graham, E. A., M. P. Hamilton, B. D. Mishler, P. W. Rundel, M. Hansen.  Use of a networked digital camera to estimate net CO2 uptake of a desiccation tolerant moss.  International Journal of Plant Sciences, in press), using networked fixed-video and micrometeorological sensors as well as laboratory-based gas exchange measurements. Several remote sensing methods that do not rely on plant reflectivity in the infrared however, have been adapted for use by imagers at the James Reserve.  For instance, a NIMS node has been used to move a visible-light camera along a transect and capture images of branches of Rhododendron for the last two years.  This has resulted in an image chronosequence of leaf expansion. Recent spectral analysis of this image sequence has correlated well with hand-measured values of leaf production and year-to-year differences in the timing of leaf flushes.

Future Directions

CMS, ESS, and NIMS technologies are moving towards more rapidly and easily deployable systems.  Development of more robust and reliable equipment for use by field researchers as well as software tools to provide immediate feedback on validation of sensor operation and network topologies are already under study.  As our systems mature, the effort and special skills necessary to deploy them must be minimized for them to be useful research tools.

The original CMS system will be retired and replaced and then expanded using a much newer and more robust ESS system when the side-by-side evaluation is completed in late spring.  We are also exploring ways to minimize mote battery replacements (and associated costs) by means of highly efficient boost circuits and PV panels.  Our hope is that the communications/connectivity experiment will also yield some useful results for additional power savings that can incorporated into future network development.

The seamless integration of geospatial modeling and statistical analysis is one of our greatest priorities. This requires continued work with the 'R' statistical package, and possibility the use of traditional GIS server tools. Secondly, the effective navigation and representation of data through time—whether through Google Earth Enterprise's features or via the advancement of external web interfaces—is high on our list of desired developments. Both of these features will vastly improve the utility of our sensor data for biological research.

The addition of sensor data sources from new and existing CENS sensor systems is also a high priority. Data streams from several other sensor systems at the James Reserve, including AMARSS, NIMS, and NAMOS, are slated for immediate inclusion. In the case of NIMS, access to the device's web-based control application via Google Earth is planned shortly.

The long-standing leader of GIS software, ESRI, is soon to debut its own 3D world and data visualization program. Though Google Earth provides straightforward web-integration, it is not a complete GIS-capable tool. ESRI's impending 'ArcGIS Explorer' may be considered as an alternative, or additional tool for visualizing TEOS data and multimedia.

Work using the published model of photosynthesis vs. color of the star moss will continue, producing an annual carbon budget for the moss at the James Reserve.

Work on modeling bracken fern responses to light flecks will be conducted this Spring and Summer and will be incorporated into models of light fleck distribution conducted by NIMS graduate students to estimate total carbon gain.

Work on energy fluxes at the soil-atmosphere interface along the AMRASS transect will advance with additions of net radiation and wind speed sensors as well as extensive measurements of soil surface temperatures.

Work will continue on the cold air drainage modeling with the addition of a vertical transect of sensors that will result in terrain-independent measurements of air mass fluxes up and down Hall Canyon.