CD-03 Progress Report - University at Albany - SUNY by wpr1947


									Progress Report CD-03, Fitzjarrald/Moraes                                                                  1

Progress Reports should include the following on the first page:
Investigation Group: CD-03
 (1) Title of the grant.
Periodic, transient, and spatially inhomogeneous influences on C exchange in Amazônia
(2) Type of report.

(3) Name of the principal investigators.
  David R. Fitzjarrald - State University of New York, Albany
  Osvaldo Luiz Leal de Moraes - Universidade Federal de Santa Maria (UFSM)

(4) Period covered by the report.
1/1/03 – 12/31/03
(5) Name and address of the recipient's institution.
Atmospheric Sciences Research Center
University at Albany, SUNY
251 Fuller Road
Albany NY 12203
(6) Grant number.

1. Narrative of 2003 activities:
          Mesoclimate studies. Part of the CD-03 effort is to assess any special characteristics of the LBA
Santarém study area that might introduce a bias in the ecosystem productivity estimates. Claims are
already being made about the importance of carbon uptake in Amazonian forests to the global budget (or
lack of it) based on LBA-ECO measurements. The project‟s tall flux towers, for example, are located in a
thin area of forest sandwiched between cleared lands and near to large rivers that are known to influence
cloudiness in their proximity. Characterizing site biases has led CD-03 to deploy a widely dispersed
instrumentation net, including sensors at each flux tower. Our effort includes a detailed analysis of
historical data obtained at airports designed to assess how the LBA-ECO operation years compare with
those in the recent past. Preliminary results (Fig. 1) show how that temperature and precipitation are higher
and wind speed is lower during the LBA-ECO years compared to the recent past. Fig. 1 also illustrates the
precipitation modulation by the 1998-98 El Niño. While the annual passage of the intertropical
convergence zone is clear in the Belém surface pressure signal, it is surprising that the daily averaged wind
speed at Santarém correlates well with the observed Belém-Santarém surface pressure difference.
Progress Report CD-03, Fitzjarrald/Moraes                                                                         2

Fig. 1. Decadal record of climatic variables in eastern Amazonia. Top panel: Surface pressure observation at
Belém, Pará and (red) the Belém-Santarém surface pressure difference. Middle panel: Daily-averaged temperature
(red) and wind speed at the Santarém (STM) airport Bottom panel: Ten-day accumulated precipitation and the
Southern Oscillation index (red). Vertical lines during 2001 indicate the period of the USP Intensive mesoscale

         A complementary initiative to understand radiation, rainfall and temperature anomalies led to the
development of a small network of surface weather stations. These units were obtained from a variety of
sources, but are operated and analyzed by CD-03. Two U. S. Forest Service weather stations were installed
in 1998. LBA-ECO provided support for three more, which came on line in July 2000. Three more weather
were obtained from EMBRAPA. Each station was upgraded by CD-03 to include a GPS receiver that
keeps track of time, a barometer, and soil temperature and moisture sensors. Data are retrieved manually.
Progress Report CD-03, Fitzjarrald/Moraes                                                                   3

                        Fig. 2 LBA-ECO weather station network operated by CD-03.

         Flux studies. The Pasture site (km77) began operation in August 2000. CD-03 was responsible for
design, purchase, shipping, installation and operation of the photovoltaic power system at this site. In April
2001, the laser ceilometer and subcanopy flux measurements, and the radiation suite were installed at the
Old-Growth site at km67. The CD-03 data acquisition system was installed, and data from the above-
canopy sonic anemometers are now ingested on the same time base as are the radiation, ceilometer, and
subcanopy flux measurements. In August 2001, an acoustic sounder was installed first a Belterra, in
support of the Mesoscale experiment.
         Members of CD-03 made regular visits to the various field sites during 2003. In March and April,
2003, Fitzjarrald and Sakai installed repaired instruments and computers at the FLONA sites. The sodar
was moved to the pasture site, in support of UFSM balloon campaigns in 2001 and 2003. In July 2003,
CD-03 installed extensive instrumentation at the second tower at the km 83 Cut site. These instruments are
designed to track the microclimate in the gap near the 2 nd tower. Also, we aim to understand the extent to
which CO2 may accumulate and be swept out of the gap, as we have seen in model simulations.
         During the same month, the “Draino” set of sensors was installed at the km 67 “Old Growth” Site
(see Appendix 1). These instruments are deployed to determine the importance of lateral subcanopy flows
on the estimates of ecosystem respiration made using the eddy covariance system.
         Canopy structure studies. In July 2003, collaborator G. Parker from the Smithsonian
Environmental Research Center conducted extensive surveys of the vertical and horizontal structure of the
forest canopy in the vicinity of the flux towers at km67 and km83 (see Appendix 2). Parker‟s work
demonstrates that the 2 sites are very similar in average forest structure even after the selective logging.
Thus, findings by other groups that there is little difference in net uptake after the logging are not
         Instrument development. During an extended visit at UAlbany, SUNY, UFSM graduate student
Rodrigo da Silva characterized the properties of the Russian CO2 sensor. In September and October 2003,
a tethered balloon borne sonde was constructed at UAlbany and tested in New York. Details of the field
characterization of the instrument are given in Appendix 3. The new tethered balloon sonde, which
measures, CO2, T, q, and pressure, was taken to the Santarém site. Reports of this field campaign are not
yet available.
         Intensive field campaign. In November 2003, the UFSM team conducted another series of tethered
balloon field campaigns. The objective was to determine the thickness of the nocturnal stable layer at the
pasture site and to understand the CO2 buildup in the stable boundary layer at the km77 pasture site. The
CD-03 „CO2 sonde‟ was deployed in conjunction with the older AIR tethersonde, which provided
Progress Report CD-03, Fitzjarrald/Moraes                                                                                            4

complementary measurements of wind speed and direction. During a second 10-day period, these profile
observations were performed near the cut site tower at km83, inside the FLONA Tapajós.

2. Narrative of 2004 workplan:
        The aims of CD-03 during 2004 are:
a. Complete documentation of existing data sets, data assurance and submission.
b. Restore instruments damaged by acts of nature and power difficulties at km83 and km67 sites.
c. Conduct intensive field observation periods for the „DRAINO‟ deployment and at km83.
d. Submit papers that are in progress.

3. Description of any difficulties encountered or any issues to resolve (if needed):
          We suffered the loss of a large number of instruments at the km83 and km67 sites during some
electrical disturbances that occurred at the beginning of September 2003. In addition to the loss of 5 sonic
anemometers, two computers and five UPS units were put out of action. All optical interfaces between
dataloggers and serial instrument outputs and our computers were destroyed. As of this time (January
2004), we are still trying to get the instruments back on line. We continue to have difficulty with the power
quality at km67 and km83 sites. At our km77 site, there has been difficulty with the diesel generator that
serves as a backup to our solar panel power system. We are currently working with LBA staff and other
PI‟s to address these difficulties.

4. Description of training activities conducted, including lectures, public outreach, and short courses:
1) O. M. M. Moraes. Conferência "O LBA e as Torres de Fluxo", I Fórum das Instituições de Ensino
Superior de Santarém e o Projeto LBA, Junho de 2003, Santarém.

2) O. C. Acevedo et al., "Surface to atmosphere exchange in an Amazon pasture/agricultural site", LBA
workshop “Biogeochemical changes associated with the establishment and maintenance of
agroecosystems”., 2 - 4 November 2003, Brasilia.

3) O. M. M. Moraes. Participação na Mesa Redonda " A Amazônia e Mudanças Climáticas " no III
Congresso Interamericano de Qualidade de Ar, Julho de 2003, Porto Alegre.

4) Orientation of undergraduate student, I. Cibelle G. Sampaio, FIT, Santarém PA. Ms. Sampaio presented
a poster at the recent LBA-ECO Meeting:
         Variation of photosynthetic radiation in a primary forest in western Pará,
        I. Cibelle G. Sampaio, O. L. L. Moraes, D. R. Fitzjarrald, and R. K. Sakai. Poster presented at the
        7th LBA-ECO Science Team Business Meeting, 5-8 November 2003, Fortaleza.
       The principal goal of this study is to characterize the solar radiation regime in the subcanopy environment for a rain forest.
       The experimental site is located at Tapajós national forest (2,51180 S, 54,57290 W) in a primary tropical forest. At this site a
       65m micrometeorological tower was installed with a set of pyranameters (Kipp a Zonen, model CNR1) and PAR sensors
       (LICOR model LI191AS) photosynthetically active radiation. These take solar radiation measurements of global incident
       and reflected radiation. A network of 16 pyranameters (LICOR Li-190 model) was installed around tower to measure total
       radiation incident upon the forest floor. For comparison between total radiation and PAR, one PAR sensor was included in
       this network. The measurements discussed here were made 07 to 14 July of 2003, at 10:00 to 14. The ratio of PAR
       (photosynthetically active radiation) to incident solar radiation the canopy was 36% at canopy top, while in the subcanopy it
       was 4%. In the canopy, the PAR-Albedo (ratio of PAR radiation reflected and radiation incident PAR) is 0.02 .The average
       of PAR incident radiation is 1226 mol/m2/s. In the subcanopy, the maximum average of PAR radiation was of 40.285
       mol/m2/s, and minimum average was of 1.148 mol/m2/s. The PAR incident radiation on the canopy is approximately 8 times
       greater than in the subcanopy. The average PAR radiation absorption of 97.457% is explained by high-density of the

5) Graduate Student Rodrigo da Silva, UFSM, obtained a „sandwich‟ scholarship to be in residence at
UAlbany during the period January-October, 2003. During this time he worked with the computer
simulation model (LES) and with instrument development (see report in Appendix 3).

6) We have trained INPA researcher Julio Tóta in the use of the subcanopy advection system. R. Staebler
produced a detailed manual about the SUNY data acquisition system. Since it is based on open-source
linux code, it can be emulated in Brazil at relatively small expense. Details are given in Appendix 1.
Progress Report CD-03, Fitzjarrald/Moraes                                                                                           5

5. CD-03 LBA DIS Contact Person: Please designate a participant in your group to be the contact
person for LBA DIS questions about data, metadata, etc. If no person is identified, then it is assumed
that the PI is the contact person for LBA DIS questions.

  Name: Ricardo K. Sakai
  Address: ASRC, University at Albany, SUNY, 251 Fuller Road, Albany, NY 12203, USA
  Phone #:     518-437-8766

6 . Data set descriptions, including status of metadata registration and online data availability. LBA DIS
policy encourages metadata registration for data sets in preparation.

Data Set Title: Mesoscale Meteorological Station Network at Santarem (PA) region
Data Availability with URL: Data Not Available on a Web site, but available through ftp connection.

Data Set Title: Turbulent fluxes from an Eddy Correlation System at a Pasture in the Santarém (PA) region
Data Availability with URL: Data Not Available on a Web site, but available through ftp connection.

Additional data sets dealing with canopy structure, subcanopy flows, cloud base information are being
prepared for submission to the archive.

7. List of publications (in print, in press, and in review) funded by NASA's LBA component. Please
submit a PDF file of any publications in print.

1) Evolução Temporal do Perfil de CO2, para a estação sêca, em uma área de pastagem na Amazônia, A. C.
Siqueira, O. L. L. Moraes, O. C. Acevedo, R. Silva, D. R. Fitzjarrald, R. K. Sakai, XIII Congresso
Brasileiro de Agrometeorologia, Agosto de 2003, Anais, vol 1, 99-100.

2) Perfil de CO2 e vapor d´água, para uma região de pastagem na Amazônia, A. C. Siqueira, O. L. L.
Moraes, O. C. Acevedo, R. Silva, D. R. Fitzjarrald, R. K. Sakai, III Workshop Brasileiro de
Micrometeorologia, Novembro 2003, Publicado em: Ciência e Natura, Volume Especial, 285-290.

3) Variação Rítmica dos Sistemas Convectivos na Região Amazônica, V. Anabor, O. C. Acevedo, O. L. L.
Moraes, III Workshop Brasileiro de Micrometeorologia, Novembro 2003, Publicado em: Ciência e Natura,
Volume Especial, 237-242.

4) Land-use change effects on local energy, water and carbon balances in an Amazonian agricultural field,
Ricardo K. Sakai, David R. Fitzjarrald, Osvaldo L.L. Moraes , Ralf M. Staebler, Otávio C. Acevedo,
Matthew J. Czikowsky , Rodrigo da Silva, Eleazar Brait , and Valdelirio Miranda, LBA-ECO Special Issue,
Global Change Biology, 2004, in press.
             To study how changing agricultural practices in the eastern Amazon affect carbon, heat and water exchanges, a 20 m
       tower was installed in a field in August 2000. Measurements include turbulent fluxes (momentum, heat, water vapor, and
       CO2) using the eddy covariance approach (EC), soil heat flux, wind and scalar profiles (T, q, and CO2), soil moisture
       content, terrestrial, total solar radiation, and PAR (photosynthetically active radiation, 400-700 nm). At the beginning of the
       measurements, in September 2000, the field was a pasture. On November 2001, the pasture was burned, plowed and planted
       in upland (non-irrigated) rice.
             Calm nights were the norm in this site. Anomalously low values of net ecosystem exchange were found using the EC
       method, even when the common criterion u* < 0.2 m s–1 was used to identify and exclude poor performance nights. We
       observed more plausible values of NEE using criterion u* < 0.08 ms-1, indicating that the criterion must be revised downward
       for flow over surfaces smoother than forests. However, even using the lower threshold, u* was lower than this limit for 82%
       of nights, and this led to nocturnal respiration underestimates. We compensate for this difficulty by estimating the
       respiration rate using the nocturnal boundary layer budget (NBLb) method.
       Land use change from pasture to rice cultivation strongly affected both diurnal rates of turbulent exchange but also the
       pattern of seasonal variation. Seasonal wet and dry season differences in vegetation state were clearly detected in the albedo
       and PAR-albedo. These reflectivity changes were accompanied by modified net radiative flux, turbulent heat flux and
       evaporation rates. The highest evaporation rate was observed during the rice crop, when the field had total evaporation
       approximately half the precipitation input, less than that of the surrounding forest. Effects of the land cover changes were
       also detected in the carbon budget. For the pasture, the maximum CO2 uptake occurred in May, appreciably delayed from
       the start of the rainy season. After the field was plowed and the soil was exposed and there was efflux of CO 2 to the
Progress Report CD-03, Fitzjarrald/Moraes                                                                                            6

       atmosphere day and night for an extended period. Highest values of carbon uptake occurred during the rice plantation.
       Though the upland rice took up carbon at double the rate of the pasture that it replaced, the field was left fallow for much of
       the year, during the dry season

5) Inferring Nocturnal Surface Fluxes from Vertical Profiles of Scalars in an Amazon Pasture , Otávio C.
Acevedo, Osvaldo L. L. Moraes, Rodrigo da Silva, David R. Fitzjarrald, Ricardo K. Sakai, Ralf M.
Staebler, Matthew J. Czikowsky, LBA-ECO Special Issue, Global Change Biology, 2004, in press.
       Ecosystem carbon budgets depend on there being good representative surface flux observations for all land use types during
       the entire diurnal cycle. In calm conditions that often occur at night, especially in areas of small roughness (such as
       pastures), ecosystem respiration rate is poorly measured using the eddy covariance (EC) technique. Nocturnal vertical
       profiles of temperature, humidity and winds were observed using tethered balloon soundings in a pasture in the eastern
       Amazon during two campaigns in 2001. The site is characterized by very weak winds at night, so that there is insufficient
       turbulence for the eddy covariance technique to determine fluxes accurately. To compensate, the time evolution of the
       profiles is used to determine surface fluxes at early morning and these are compared to those observed by eddy covariance at
       a nearby micrometeorological tower. The nocturnal boundary layer thickness h is determined as the height to which the
       surface fluxes must converge so that energy budget closure is achieved. The estimated values range from 30 m, around 2200
       LST, to more than 100 m just before dawn. These are in good agreement with the observed thickness of a frequently-
       observed fog layer during the middle of the night. During the early portion of the night, when the accumulation layer is
       shallow, there is appreciable decrease of dCO2/dt with height. On calm nights, CO2 accumulation rate is larger near the
       surface than at higher levels. On windier nights, this accumulation rate is vertically uniform. Hence, extrapolation of tower
       profiles for estimating fluxes must be done carefully. Although uncertainties remain large, an alternate approach to the eddy
       covariance method is described for measuring nighttime surface CO2 fluxes under stable atmospheric conditions.

6) Silva Dias, M. A. F., P. L. Silva Dias, Marcos Longo, D. R. Fitzjarrald, and A. S. Denning, 2003. River
breeze circulation in Eastern Amazon: observations and modeling results. Theoretical and Applied
Climatology, LBA Special Edition, in press.
       The observations during the CIRSAN/LBA field campaign conducted close to two major rivers of the Amazon Basin, the
       Tapajós and the Amazon, indicate that during weak trade wind episodes the Tapajós river breeze actually induces a westerly
       flow at the eastern margin associated to a line of shallow cumulus. The river circulation is interpreted with the help of a high
       resolution numerical simulation. A single cell forms during late morning and evolves into the afternoon with ascending
       motion in the eastern margin and a descending branch in the western margin suppressing cloud formation. During night
       convergence is seen at the Tapajós river center.

8. (Optional) Any other publications that you would like to include (e.g. commentaries, letters to the
editor, articles in popular magazines): Nothing to report

9. Participants: Please include participant names, university affiliation, nationality, and for students,
include degree sought and provisional thesis title.

Name: Otávio C Acevedo
  Citizenship: BRAZIL

  Name: Vagner Anabor
  Citizenship: BRAZIL

  Name: Matthew J Czikowsky
  Citizenship: USA

  Name: Rodrigo da Silva
  Citizenship: BRAZIL

  Name: David R. Fitzjarrald
  Citizenship: USA
Progress Report CD-03, Fitzjarrald/Moraes                        7

  Name: Osvaldo Luiz Leal de Moraes
  Citizenship: BRAZIL

  Name: Geoffrey Parker
  Citizenship: USA

  Name: Ricardo Sakai
  Citizenship: BRAZIL

  Name: Irene Cibelle Gonçalves Sampaio
  Citizenship: BRAZIL

  Name: Valdelirio Miranda Silva
  Citizenship: BRAZIL

  Name: Ralf Manfred Staebler
  Citizenship: GERMANY

  Name: Julio Tóta da Silva
  Citizenship: BRAZIL

  Name: Alexander E. Tsoyref
  Citizenship: USA

People to Add to CD-03:
    (contact via Osvaldo Moraes email.)
 Name: Adriano Siqueira
  Role: Participant
  University/Organization: Universidade Federal de Santa Maria
  Nationality: BRAZIL
  Degree Sought (if student): Ph. D.
  Provisional Thesis Title (if student): Not yet available

 Name: Cintya Martins
 Role: Participant
 University/Organization: Universidade Federal de Santa Maria
 Nationality: BRAZIL
 Degree Sought (if student): M.S.
 Provisional Thesis Title (if student): Not yet available

Name: Hans Rogério Zimermann
  Role: Participant
  University/Organization: Universidade Federal de Santa Maria
  Nationality: BRAZIL
  Degree Sought (if student): Ph. D.
  Provisional Thesis Title (if student): Not yet available
Progress Report CD-03, Fitzjarrald/Moraes                                                                         8


Appendix 1. Subcanopy sensor networks installed at the km67 “Old Growth” LBA-ECO site.
         A. CD-03 Subcanopy light sensor array
{See I. C. Sampaio abstract for details of instrumentation. Light sensors are multiplexed such that data are
obtained every 10 seconds. These data are recorded on the same time base as are the other CD-03 sensors
at km67, including the radiation measurement suite at tower top.}

Figure A1. CD-03 light sensor array near the tall tower at the km67 site.

          B. Draino LBA {this is an edited section from R. M. Staebler, 2003: Forest subcanopy flows and micro-
scale advection of CO2, Ph. D. dissertation, University at Albany, SUNY.}
                   We deployed a system of horizontal wind and CO2 field measurements at the “Old
Growth” forest site at km67 in the LBA-ECO study area. Early reports of net ecosystem uptake, based on a
few weeks of data at a single site, suggested a small but positive sink of CO 2, which, when scaled by the
area of the Amazon region, could account for a significant fraction of the missing global carbon [e.g.,
Grace et al., 1995]. More recent reports, based on longer observations, indicate smaller uptakes [Miller et
al., 2003], and that results are strongly dependent on how nocturnal fluxes are treated. Horizontal
subcanopy motions are thought to advect CO2 laterally at night. If true, the might explain the necessity to
exclude large amounts (20-60% or nights in some cases) of nocturnal eddy flux data from consideration.
This is accomplished using an arbitrary „cut off‟ based on the observed nocturnal momentum flux. At one
midlatitude site, Staebler and Fitzjarrald (2004, Ag. & Forest Meteor., in press) became among the first to
document this process and to discuss its physical causes. Preliminary studies indicate that keeping track of
CO2 concentrations and the subcanopy wind near the surface will be adequate to describe horizontal flows
at the km67 site in the FLONA Tapajós.
          Since results from LBA will likely be used to represent the Amazon in its entirety in global change
models, it is of paramount importance to determine whether there are any systematic problems with the
location of the flux sites. Their proximity to the Tapajos Escarpment leaves open the possibility of
systematic drainage flows towards the west (see Figure A1).
Progress Report CD-03, Fitzjarrald/Moraes                                                                      9

          Figure A2. Wind roses at the LBA “Old Growth” site (km 67), for the period DOY 140-230, 2002. The
central wind rose (heavier lines) is for the sonic anemometer at 58m, the other 4 are at 1.8m.

          Several months of subcanopy wind data are already available based on our pilot study in 2002
indicate flows that are generally aligned with flows aloft, even at night (Figure A1). The north and west
subcanopy anemometers may suggest higher frequencies of easterly and north-easterly winds at night,
which would agree with drainage flows towards the Tapajos. The problem is that any potential drainage
flows are in the same direction as the synoptic trade winds, as well as any nocturnal land breeze. A number
of cases of low wind speeds aloft coupled with strong subcanopy temperature inversions near the ground
will have to be collected to determine whether drainage flows do occur there. Unfortunately, wind speeds
are low and environmental conditions are harsh for sensitive equipment, and the currently employed 2D
sonic anemometers (CATI/2) suffer from shifting offsets that are not easily correctable. They were replaced
by the superior SPAS/2Y sensors when the horizontal CO2 gradient system was installed in July 2003.
          From profile measurements of CO2 and the wind speed [Saleska, pers. comm.], shape factor
profiles can be constructed to guide the installation of the horizontal network. The measurements should be
made at a height that is likely to see a significant fraction of the horizontal transport uc, so that the process
of vertical integration (by applying the shape factor) does not hinge on the horizontal transport estimate at a
level that contributes insignificantly. The mean profiles obtained (Figure A2) suggest that the transport
layer is significantly thicker, largely due to build-up of respired CO2 in a much thicker layer. However, the
transport estimate at 2 m level is close to its maximum most of the night, and thus this level should be
sufficient (in addition to being technically feasible!) to make representative estimates of the horizontal
transport. During daytime hours, a secondary peak in the transport profile becomes prominent, which
would prohibit using this kind of approach to estimate the horizontal advection in the whole layer up to
58m using measurements only at 2 m.
Progress Report CD-03, Fitzjarrald/Moraes                                                                             10

           Figure A3. Normalized profiles of wind speed (dotted line), CO2 (solid line) and their product (i.e. the
transport term) at the Old Growth Site, LBA.

In July 2003, the more DRAINO system was installed at the km67 site, as indicated below.

Figure A4. Left: DRAINO installation currently at km67 site. Dots indicate the CO2 inlets at the satellite locations,
collocated with 2D sonic anemometers. Right: Locations of aspirated T/RH probes (“FS”) and CO2 inlets on the 5 m
tower at the center of the array.
Progress Report CD-03, Fitzjarrald/Moraes                                                                   11

{Reproduced below are excerpts from the Operating Instructions for the DRAINO array instrument suite.}
Operational Instructions
Ralf Staebler, ASRC, UAlbany {Now at the Canadian Meteorological Service.}
     1. Introduction
     2. Air Sampling System
     3. Data Acquisition
     4. Data Processing
     5. Operating Instructions:
              a. Bi-weekly checks
              b. Licor calibration
              c. Sonic anemometer calibration
              d. Data processing
1. Introduction
This system is based on a design for similar measurements at Harvard Forest from 2000-2002. For more
details on the design considerations and the physics involved, please refer to Ralf Staebler‟s PhD
dissertation (obtain an electronic version by emailing
The general idea is to measure the horizontal advection of CO 2, which may explain the missing vertical
CO2 fluxes on calm nights, when the eddy covariance technique fails to properly detect nocturnal
respiration. A horizontal gradient in CO2, coupled with a persistent flow in a certain direction, can create
non-zero horizontal advection terms of the form (u)(dc/dx), which are commonly assumed to be zero. The
aim is to actually measure this term to determine its significance.

2. Air sampling
          A schematic of the air flow system is given in Figure 1. The flow out of the main pump at km67
was measured to be about 45 L/min. There is continuous flow through all 10 lines at all times. A rotating
valve diverts an additional amount of air from the main lines through the Licor, for 20 seconds for each of
the 10 lines. During this “active” period, the flow through the line is increased by typically 0.5 L/min, as
measured by rotameter at the inlet.
          The flow rate through the 10 sample lines is best determined by timing the arrival of a CO 2 spike,
generated by breathing into the inlet, at the Licor. If this is not possible, e.g. on solo trips, a flow meter
(rotameter) can be attached to the inlet. Note that the rotameter itself represents a major restriction
(resistance) to the flow, and the flow rate read does NOT actually represent the unrestricted air flow
through the line. But the reading can still be used to check for consistency, and that there is actually flow
through the line. The flow rates read with a rotameter are around 3 L/min for the six short lines (profile)
and about 1.0 L/min for the long (horizontal) lines, and will increase by about 0.5 L/min while the line is
active (i.e. during the 20 seconds while the rotating valve is sampling the line).
Progress Report CD-03, Fitzjarrald/Moraes                                                                                    12

                                                        Rotary Valve



                                                                                5 L/min
Inlets        1                                                                        Nitrogen
                  2                                                                           A      B(in) B(out)

                                                             Pump                         Licor 7000
                                                            50 L/min
                                                                                       CO2/H2O Analyzer


                                                                           Serial (RS232) to computer

Figure A5: The Draino air flow system.

3. Data Acquisition
The data acquisition directory for Draino is /home/lba/draino67/data. This directory contains the scripts
which call the “robot”, which collects the data. These scripts are:                         ttyC5            (Cyclades Port # 6)                     ttyC8            (Cyclades Port # 9)                    S.N. 010905                      ttyC9            (Cyclades Port # 10)                   S.N. 010901                     ttyC10           (Cyclades Port # 11)                   S.N. 010903                      ttyC11           (Cyclades Port # 12)                   S.N. 010904                        ttyC12           (Cyclades Port # 13)                   S.N. 010902              ttyC7             (Cyclades Port # 8)              ttyC13            (Cyclades Port # 14)

[Note: 2D5m is currently actually at 2.0m on the central 5m tower, until a replacement ATI 3D can be
found for the defective instrument we started out with]

Remember that to run programs or scripts on linux that are not in the general path, you have to preface the
command with ./ (e.g., ./
All the above scripts are combined in, which is called by /etc/rc.d/rc.local on start-up.

The robot program (written in C++) collects any serial stream from either the cyclades serial board (ttyCx)
or the on-board serial COM ports (ttySx). The options (obtained from the help screen, by typing ./robot –h)

Usage: robot [-2] [-eo] [-ahiknpuvw] [-b n] [-B n] [-c f] [-d p] [-f s]
Progress Report CD-03, Fitzjarrald/Moraes                                                                 13

             [-g s] [-l p] [-L p] [-q Nh | Nm] [-r n] [-s Nr | Ns] [-t n]
             [-U p] [-z s]
         -2 : 2 stop bits
         -e : even parity
         -o : odd parity
         -a : scroll data on screen
         -h : print this message
         -i : ignore CR in incoming data stream
         -k : make fifo recognize hex numbers (e.g., use with GILL)
         -n : do not poll device on start up (e.g., use with SODAR)
         -p : print internal defaults
         -u : unlock serial port before locking it
         -v : print version
         -w : terminal emulation mode (writable device)
         -b n : set bits to n [-b 8]
         -B n : set baud rate to n [-B 9600]
         -c f : use file f for configuration [-c robot.cfg]
         -d p : use serial port p [-d /dev/ttyC0]
         -f s : use file s for RTM fifo output [-f .FIFO]
         -g s : tag string s to log file and data files [-g sonic]
         -l p : recycle power to power port p (1-8) [-l 1]
         -L p : use parallel port p (0-2) to recycle power [-L 1]
         -q Nh : pack N hours of data per file [-q 24h]
         -q Nm : pack N mega bytes of data per file [-q 200m]
         -r n : stop after n data files (0: nonstop) [-r 0]
         -s Nr : time stamp after N records (0r: no stamp) [-s 100r]
         -s Ns : time stamp after N sec (0s: no stamp) [-s 10s]
         -t n : set timeout to n sec (0: disable timeout) [-t 60]
         -U p : unlock serial port p and exit [-U /dev/ttyC0]
         -z s : start on 13 Jan 2000 @ 5pm LT: [-z 13-01-2000__17:00:00]

The script calls “rtm”, which is the program that merges all the above serial streams and puts
them into one file, with the extension *.rwd.
./rtm – h gives the following options:

Usage: rtm [-ahpuv] [-f s] [-l s] [-q Ns] [-s Nr] [-t n]
         -a : scroll output on screen
         -h : print this message
         -p : print default settings
         -u : unlock fifo before locking it
         -v : print version
         -f s : read from fifo s [-f .FIFO]
         -l s : use file s as lookup table [-l rtm.tbl]
         -q Nh : pack N hours of data per file [-q 12h]
         -s Nr : time stamp after N records (0r: no stamp) [-s 600r]
         -t n : set timeout to n sec (0: disable timeout) [-t 60]
         -e s : use extension s [-e raw]                                    rere2003d
Progress Report CD-03, Fitzjarrald/Moraes                                                                 14

       Serial stream
       scripts (*.sh),
       calling “robot”


Fig. A5, Data acquisition flowchart.

          For real-time monitoring purposes, a tabular display program exists in /home/lba/draino67/data,
called ./disp. Also, gnu-plots have been customized for display of the CO2 concentration (./gnuCO2), the
Flying saucers (./gnuFST and ./gnuFSRH), and the first sonic (./gnusonic1). To customize one of this for
other parameters, all that needs to be done is to change the parameter number near the bottom of the
gnuplot script.

Appendix 2.
Small-scale structure of the canopy of a tropical rainforest from ground-based lidar measurements
Geoffrey G. Parker and CD-03 collaborators
Smithsonian Environmental Research Center

The structure of the forest canopy is important to a variety of critical forest functions. However, obtaining
the measurements of canopy structure that are necessary to understand these forest functions has been
difficult. Most available techniques provide poor spatial resolution and are at scales not linked to the
footprints of canopy functions. Consequently, progress in understanding canopy structure-function
relationships has been slow. Lidar remote sensing shows promise for yielding the desired resolution,
particularly in the vertical dimension (Lefsky et al. 2002). However, most remote sensing altimetry sensors
are either very expensive or are currently in a development phase and are not available to support field
research. In this project we used a portable laser rangefinding system to make dense measurements of the
location of canopy elements and assembled these into high-resolution views of structure in three
Progress Report CD-03, Fitzjarrald/Moraes                                                                  15

dimensions. We focused especially on the change in the structure of a tropical rainforest canopy caused by
selective logging.

Study Area
In July of 2003 we took measurements of canopy structure in two study areas in the FLONA Tapajós, Para,
Brazil, as part of the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA). One site at km
67 along road BR 163 was the intact site and the other at km 83 was the selectively logged site. Sampling
of canopy structure was conducted along existing transects at each site, trails of variable openness. The
layout of these lines differed between sites. At km 67 we used three straight 1000m transects radiating
away from the flux tower in the direction of the prevailing easterly wind of four established by the Harvard
team to study biomass and litterfall. The northeast, east and southeast transects are here called transects 1,
2, and 3, respectively. At km 83 we used three parallel transects upwind of the old flux tower, two were
litter collector transects (lines E and I) and another that connected the two towers (line G) - the lengths
varied, from 300 to 350 m and totaled 975 m. Also at the km 67 sites we sampled additional transects in
the vicinity of the “DRANO” project, totaling 407 m.

Lidar measurements
We used a Riegl LD90-3100HS first-return laser rangefinder (operating at 890 nm and 1 kHz, laser safety
class I) mounted to the front of a frame at 1 m above the ground and pointed upward. The frame was
carried by a person, attached with a hip-belt and shoulder straps. Power for the laser was supplied from a
12V motorcycle battery. The laser made 1000 measurements per second that were automatically averaged
in groups of five to yield about 200 data points per second. The data were transferred through a serial cable
to a small notebook computer (Toshiba Libretto 50 CT), also mounted on the frame. The location of each
range measurement was estimated from its sequence in the data file, assuming a constant walking speed.
Distances between measurements were typically less than 1 cm - the spot size of the laser beam is 4-6 cm at
the ranges measured.

The data files were edited to identify out-of-range values (such as when penetrating through canopy
openings to the sky) and to remove spurious values. The edited files were processed through a customized
program to group the ranges horizontally, calculate the vertical profiles using the method of MacArthur and
Horn (1969), estimate the surface area density using the overlap transformation (Parker and Lefsky, in
preparation), and assign coordinates to each estimate. Here we used bins that were 2 m in the horizontal
and 1 m in the vertical. The resulting estimates refer to parallelepiped-shaped voxels of 1x2x1 m in the x, y
and z dimensions, respectively.

To view the resulting three-dimensional estimates, we made height sections from the transects. These are
presented as contour plots of surface area density, where the abscissa is horizontal distance along the
transect and the ordinate is height above ground. The raw estimates were not interpolated for these
presentations, however, the contours were smoothed with a cubic spline. To understand the overall vertical
structure in each sampled area we extracted the mean and standard deviation for each height and present
these as bar graphs.

Additional statistics
The maximum canopy height is defined by the highest observed surface in any transect. The local outer
canopy height is the maximum surface height in a column – together these define the outer canopy surface.
The average value of this height is called the “mean outer canopy height” and its standard deviation is the
“rugosity.” The mean height of all surfaces, weighted by the surface area density, is the “mean weighted
height.” Much of the canopy is open space and two sorts of porosity are recognized. The “enclosed
porosity” is the fraction of space under the outer canopy not occupied by surfaces; the “total porosity” is
analogous to the maximum canopy height. Cover is the fraction of horizontal locations with any canopy
surface area above (one minus the “gap fraction”). The mean Canopy Area Index (CAI) is the average area
of canopy surfaces per ground area (m2m-2). The cumulative distribution of local top heights, called the
hypsograph, is another useful summary of structure. The widths of canopy holes were estimated and
accumulated in a gap-size distribution, using 0.1 m - wide bins.
Progress Report CD-03, Fitzjarrald/Moraes                                                                   16

Results and Discussion
Forest Interior Transects
Figure 1 shows the height sections of surface area density from the six 500m sections of the three transects
at km 67. Figure 2 shows similar height sections several shorter lines around the DRANO study area near
the km 67 flux tower. Height sections for the selectively logged forest between the flux towers at km 83
are in Figure 3.

Mean Vertical Structure
The vertical pattern of average surface area density differs little among the three areas (Figure 4). All have
a bottom-heavy structure with density maxima at 7-10 m above ground and a diminishing mean density up
to 48 m. The maximum density appears to be slightly lower in the km 83 transects than for those at km 67.
The smoothness of the vertical patterns varies among the three areas, but likely reflects the differences in
sample size. The pattern for the DRANO area is similar to that of the full sample of the km 67 area.

The intact and selectively logged sites differ somewhat in their distributions of top heights (Figure 5). Both
canopies are tall but there is more material concentrated in the topmost zones in the intact forest than in the
selectively logged site. Also the selectively logged site has more (10%) of its outer surface within 5-6 m of
the ground compared to the intact site (almost none). The differences between the sites in the hypsographs
could be interpreted as material missing from the km 83 canopy, especially in the uppermost and lower
heights, which is consistent with the removal of some of the larger crown and the creation of gap areas at
the ground.

Gap size distributions
Both sites have similar, very low gap fractions - km 83 site is somewhat more open overall (0.041) than at
km 67 (0.014). There are some differences in the size distributions of these holes (Figure 6). The
selectively logged site has more openings at all sizes, only a few of intermediate size (1.35 and 5.35 m),
and no gaps of the size usually associated with tree-falls. No gaps wider than 1.0 m were seen in the intact

Summary statistics
Structural statistics tabulated below show few differences between the sites.

Characteristic                            Km 67 – intact forest             Km 83 - selectively logged
Number of points                          1500                             487
Cover, fraction                           0.98 + 0.0526                    0.98 + 0.1201
CAI, m2 m-2                               7.39 + 1.118                     6.96 + 1.746
Maximum height, m                         49.5                             46.5
Mean weighted height, m                   13.36                            13.40
Mean outer canopy height, m               20.14                            17.98
Rugosity, m                               10.03                            8.42
Total porosity, %                         73.51                            73.16
Included porosity, %                      32.36 + 15.83                    26.23 + 16.25


We thank Cibelle Sampaio and Julio Tóta for help in the field and Sr. Miranda for help in logistics. This
project was supported by the US National Aeronautics and Space Administration (NASA) and the
Smithsonian Environmental Research Center (SERC).


Harding D, Lefsky MA, Parker GG and Blair B (2001). Laser altimetry canopy height profiles: methods
and validation for closed-canopy, broadleaved forests. Remote Sensing of Environment 76:283-297.
Progress Report CD-03, Fitzjarrald/Moraes                                                               17

Lefsky MA, Cohen WB, Harding DJ and Parker GG (2002). Lidar remote sensing for ecosystem studies.
Bioscience 52:19-30.

MacArthur RH and Horn HS (1969). Foliage profiles by vertical measurements. Ecology 50:802-804

Parker GG, Chapotin SM, and Davis M. (2002). Canopy light transmittance in an age sequence of
Douglas-fir/western hemlock stands. Tree Physiology 22:147-157.

Parker GG and Lefsky MA (in prep.) Leaf overlap in some deciduous broadleaf forests: implications for
canopy structure and radiation environment.
Progress Report CD-03, Fitzjarrald/Moraes                                                                                                       18

                       km67 Harvard transects July 2003
                                                                                     0   1          2       3    4     5     6    7
                                                                                                        estimated surface area density
                                transect 1 first half




                            0          50        100     150   200             250            300          350       400      450         500

                                transect 1 second half



                        500            550       600     650   700             750            800          850       900      950        1000

                                transect 2 first half
           height, m



                            0          50        100     150   200             250            300          350       400      450        500

                                transect 2 second half


                        500            550       600     650   700             750            800          850       900      950        1000

                                transect 3 first half


                            0          50        100     150   200             250            300          350       400      450        500

                                transect 3 second half


                        500            550       600     650   700             750            800          850       900      950        1000

                                                                     horizontal distance, m
 Figure 1. Height sections of canopy surface area density along six 500m m transects at the intact forest
                                              site, km 67.
Progress Report CD-03, Fitzjarrald/Moraes                                                                                                                                                                                                        19

             vertical slice of canopy surface area
             at Drano transects km67 Tapajos

                   "A"                                                                                                               "B+C"




                    10         20                     30                 40             50                 60                        10          20            30              40                   50              60         70           80

                   "D"                                          "E"                                                                                        "F"                                                                "G"








                     5   10        15       20                  5         10       15       20        25        30        35    40    45    50             5        10    15         20        25        30        35              5   10   15


                                                      "H"                                                                                  "I"                                                           "J"








                                        5        10        15       20        25                                                                                    5    10     15        20        25        30    35   40   45
                                                                                        5        10        15        20    25   30    35   40    45   50

Figure 2. Height sections of canopy surface area density along short transects in the DRANO study area at
                                       the intact forest site, km 67.
Progress Report CD-03, Fitzjarrald/Moraes                                                                                               20

   km83 transects July 2003
                                                                             LT2 [200,175]->[200,525]



                        200             250           300              350              400                450           500

                                                           LT3 [150,150]->[150,550]



     150                200             250           300              350              400                450           500          550

                                                                             LT1 [100,175]->[100,475]



                         200            250            300             350              400                 450
  Figure 3. Height sections of canopy surface area density along three transects at the selectively logged
                                            forest site, km 83.

                                mean canopy height profiles, FLONA Tapajos July 2003
               50                                     50                                          50
                        drano lines at km 67                   fetch lines at km 67                        transects at km 83
               45                                     45                                          45

               40                                     40                                          40

               35                                     35                                          35

               30                                     30                                          30

               25                                     25                                          25

               20                                     20                                          20

               15                                     15                                          15

               10                                     10                                          10

                5                                      5                                           5

                0                                      0                                           0
                    0     0.2     0.4     0.6   0.8        0     0.2    0.4       0.6       0.8        0    0.2    0.4    0.6   0.8
                                                                                        2    -3
                                                           surface area density, m m

Figure 4. Mean height profiles of canopy surface area density in the intact site (km 67) and DRANO study area and in
the selectively logged site (km 83). The error bars are standard errors.
Progress Report CD-03, Fitzjarrald/Moraes                                                                                       21

                       Hypsography of the outer canopy at the Tapajos sites






                                                         drano km67


                       0          10      20      30       40     50      60          70          80       90      100
                                                           cumulative percent

 Figure 5. Hypsograph of the local outer canopy height in the intact site (km 67) and DRANO study area and in the
                                         selectively logged site (km 83).
                                   km 67                gap fraction = 0.0142 (sd 0.0048)



                  0.05     0.15    0.25   0.35   0.45   0.55   0.65   0.75   0.85   0.95   1.05    1.15    1.25   1.35


                                   km 83                 gap fraction = 0.0408 (sd 0.0110)



                  0.05     0.15    0.25   0.35   0.45   0.55   0.65   0.75   0.85   0.95   1.05    1.15    1.25   1.35   5.35
                                                        gap size, m
   Figure 6. Gap-size distributions in the intact (km 67) and selectively logged sites (km 83). The error bars are
                            standard errors. Note the logarithmic scale on the y-axis.
Progress Report CD-03, Fitzjarrald/Moraes                                                                         22

Appendix 3. Field testing and characterization of the miniature CO2 sensor unit.
Rodrigo da Silva,
Universidade Federal de Santa Maria, RS, Brasil
                              Intercomparisons between DX6100 vs Licor 700
           Using a Licor 7000 CO2 gas analyzer sensor as standard we accomplished the inter-comparison
between two RTM DX6100 sensors of Russian manufacture, R1 and R2. The aim is to know if the sensors
R1 and R2 can be used in measures of CO2 profile and in long-term field measurements.
           Initially, we show the result of the first run in the lab, where we saw one influence of the
temperature on the CO2 concentration measured by the Russian sensors, i.e., does the Russian sensor have
temperature compensation? After that, we discuss the set up of the Russian devices, the output data
structure and to show one way to make the temperature compensation using the raw data. Finally, we show
the time series and results for the runs.
           Also, a comparison of digital and analog output signal was made for both Licor and Russian
sensor. In addition, some comments about mistakes found in the RTM Manual are made.
           All emails exchanged with RTM people are be attached in the end of this report.

The first running:

          The figure 1 shows the time series of the first laboratory run of the new Russian sensors and
Licor 7000. The dataset was collected using a CR10X data logger, taking the analog signal of the Licor and
Russian sensors at 1Hz for continuous 24 hours. Using the Robot program the digital and analog signals
were merged into the same file.
          The structure of the data files looks like:

108 2003 139 1908 21 512.9 481.5 379.9 23.08 98 22.09 21159 32579 29503 1124 3002 1.0842
512.6113 24095 29582 21504 2583 3012 1.0832 480.1413 382.03 23.06 98.07 0.1591 0.7221 0.6729 -
108 2003 139 1908 22 512.9 481.5 379.9 23.06 98.1 22.09 21145 32601 29505 1127 3002 1.0844
514.1769 24094 29589 21502 2579 3012 1.0835 481.7971 382.08 23.05 98.06 0.1594 0.7221 0.6724 -
108 2003 139 1908 23 514.9 482.5 379.9 23.07 98.1 22.09 21143 32610 29502 1121 3002 1.0842
512.7480 24071 29579 21501 2577 3012 1.0835 482.1083 381.96 23.05 98.06 0.1592 0.7222 0.6723 -
-------------------------------- --------------------------- --------------------------- ---------------------------
             Part I                   Part II                     Part III                     Part IV

           In part I are data from CR10X datalogger (cod, year, doy, h:m, ss.s, R1, R2, Licor, Tlicor,
Pressure and Tcr10x.). Part II and III are data from digital signal of R1 and R2 as described later. In part IV
is data from digital signal of Licor (Licor, Tlicor, Pressure, DAC 1, DAC2, DAC3)
Progress Report CD-03, Fitzjarrald/Moraes                                                                   23

Fig. 1: The first run of the Russian sensors. In both graphics red line is the Licor signal of temperature and
CO2 concentration, black and blue lines are R1 and R2 signals of temperature and CO2 concentration

           We can note a strong influence of temperature in the measure of CO 2 concentration in the
Russian sensors.
           The first thought was that the analog signal sensor does not have temperature compensation;
however, page 2-9 of the manual state that the device has this feature. Also, that feature could be switched
on/off with the default state being on. But, we still do not know if the analog and digital signals are
different and we cannot assume that feature was off.

Digital .vs. Analog:

           In this particular analysis, we could see that the digital and analog Russian outputs do not have
significant difference, except by the noise present in the signal. There is an offset between analog and
digital output in the Licor sensor, it is about 3.2 ppm, (figure 2). However, the error is minor when
compared to the Russian sensors. It is because the Licor output signal has a filter of 0.5s while the Russian
sensors just take the mean over 10Hz each second. The Licor noise is 0.40 ppm. The noise for sensors R1
and R2 are 2.6ppm and 1.2 ppm respectively.

Fig. 2: Auto-comparison of the Digital and Analog signal output for R1, R2 and Licor sensors. They all
show no significant difference between digital and analog signal output.

         Using an S-function to pass a filter in R1 and R2 we could see a better result in the auto
comparison between digital and analog signal. The S-function is like:

                                             > n <- 5
                                             > filtro <- rep(1, n)/n
                                             > var <- matrix(scan(datafile,...)...)
                                             > var <- filter(var,filtro)

          Figure 3 shows the result after passings that filter on the dataset. Note that filter was passed only
in the Russian sensors. The noise in R1 and R2 drops to 0.67 ppm and 0.45 ppm respectively.
Progress Report CD-03, Fitzjarrald/Moraes                                                                24

Fig. 3: Auto-comparison of analog and digital signals of the Russian sensors, R1 and R2 after passing a
running mean filter of 5s.

         The conclusion we could take from this analysis is that doesn't have difference between analog
and digital signal from Russian sensor, and according information's from RTM the methods of noise
reduction and repeating rate for both output signals are the same. Then, can we infer that both signals did
not make the temperature compensation? Also, may we assume that feature “temperature compensation” is
set up off?

Device Setup and Data File Structure:

           The manual and calibration information from RTM are included at the end of this report. On the
page 4-1, section 4 called Analyzer Commands are the commands set to remote control using RS-232 port.
           Following the classification of the manual the commands can be divided into the following
              othe commands for preset parameters of Analyzer, which allow to do adjustment operations
                  (hw, sy, pr);
                       hw: view/edit the table of preset parameters of hardware;
                       sy: view/edit parameters of hardware synchronization;
                       pr: view/edit parameters of digital temperature regulation;
              othe command for preset calibration parameters of Analyzer (fn): view/edit calibration table;
              othe commands for preset dynamical parameters, which establish the volume and repeating
                  rate of telemetry and some parameters of statistic data treatments (jb, di);
                       jb: view/edit parameters of measuring cycle;
                       di: control the structure of output telemetry;
              othe commands which allow to start and stop measurements (go, gc, gt, st);
                       go: start of measuring mode;
                       gc: start of calibration mode;
                       gt: start of test mode;
                       st: stop any operation mode;
              othe command ZE: sensor zero correction.

          The standard configurations for the Russian sensors are described in the Calibration specification
sheets in annex at the end of this report. For now we just comment some of them, which we've been
modifying for tests.
          The commands JB, SF and DI for R1 and R2 were:
                    jb 1000 2000 100 0 1 0
                    sf 0 200
                    di 215F
          The numbers means:
                    jb command:
                            o1000 – the threshold “Danger” level is equal to 1000;
                            o2000 – the threshold “Alarm” level is equal to 2000;
                            o100 – the repeating rate of telemetry data output is equal to 1 sec;
                            o0 – the number of measurement cycles is unlimited;
                            o1 – the factor of normalization of an analog signal is equal to 1;
                            o0 – the auto-start is disabled.
                    sf command:
                            o0 – the data is averaged over telemetry output period (10 measure);
                            o200 – the number of cycles of measurements is equal 200.
                    di command:
                            oOn page 4-6 has the description this important command. The parameter
                                215F is in HEX format and binary format is (010 0001 0101 1111). Each
                                output parameter of telemetry is enabled/disabled (1/0) by the appropriate
                                bit of the parameter of command di, in this case 215F. Page 4-32 has a
                                description of the format of output telemetry.
Progress Report CD-03, Fitzjarrald/Moraes                                                                  25

          According to the description in the table on page 4-8 the bit B12 represents the units of the gas
concentration measurement. In this case, using di 215F, this bit is 0 and according to the table the units are
represented in [mmol/g] and the measuring units recalculation is off.
           As an example, take the di command as 217F (010 0001 0111 1111), to see the structure of
digital output in the Russian sensors. This command gives to us the following parameters:

                                       < Usign, Uref, Tc, Vc, Tamb, Y, X >

           Where, Usign is the output signal of the measuring channel in ADC units; Uref is the output signal
of the reference channel in ADC units; Tc is the temperature of the optopair in ADC units; Vc is the cooler
supply voltage in DAC units; Tamb is the ambient temperature with 0.1K unit, Y is the not normalized value
of the measuring channel response in ADC units and X is the measuring gas concentration in ppm or
mmol/m3. The only difference in to use 215F is that parameter Y is disabled. The manual shows the
parameter D on the contrary Y. However, the right parameter is Y as will be showed later.
           Below, we could see the examples of the di command for three different configurations: in that
example we used 360ppm standard gas calibration at 100 kPa and 298 0K:

>di 215F
{ 25036 29993 23004 2439 2984 444.9531}
{ 25046 29999 23000 2431 2984 438.1994}
{ 25033 30003 23003 2437 2984 436.7321}
{ 25043 29984 23003 2437 2984 435.6398}
{ 25042 29986 23008 2448 2984 434.4421}
{ 25057 29983 23002 2436 2984 434.8454}
>di 217F
{ 25056 29968 23002 2448 2987 1.0829 442.5871}
{ 25052 29977 23002 2447 2987 1.0818 435.9780}
{ 25047 29995 23003 2448 2987 1.0815 434.3290}
{ 25057 29992 23002 2446 2987 1.0814 433.3481}
{ 25059 29987 23003 2448 2987 1.0814 433.3351}
>di 317F
{ 25056 30014 23000 2454 2987 1.0825 10924.8110}
{ 25052 30004 23002 2453 2987 1.0818 10831.4370}
{ 25038 29996 22999 2448 2987 1.0820 10858.6674}
{ 25063 30021 23004 2459 2987 1.0819 10840.3746}
{ 25066 29993 23002 2455 2987 1.0815 10786.0170}
{ 25049 30002 23004 2458 2987 1.0819 10836.1119}

          Note, in commands di215F and di217F the bit B12 is 0 and 1 in di317F. It means that to di215F
and di217F the units should be [mmol/g] according to page 4-8 or [mmol/m3]. According to page 4-33 and
to command di317F the units should be in [ppm].
         According to recent information from RTM, they said that the information in the manual has
changed. The B12 bit equal zero means concentration units in [ppm] and it equal 1 means [mmol/g]. As
well, they said that current device version the temperature compensation is unused because the temperature
sensor is placed at electronic board, not inside the gas-sampling cell.

Gas Concentration Calculation & Temperature Compensation:
Progress Report CD-03, Fitzjarrald/Moraes                                                                26

          Both, measuring Um and reference Ur channels are used to calculate the Y, as describe on page 2-
6 in the manual:

                                                     D = Um/Ur                                           (1)

                                                     Y = D0/D                                            (2)

          Where, Y is the not normalized value of the measurement, D0 is the zero rate at zero gas
          Using know polynomial coefficients A0, A1, A2 and A4 the gas concentration is calculated.

                                           X = A0 + A1*Y + A2*Y2 + A3*Y3                                 (3)

            Page 2-7 says that X is calculated in absolute mmol/m3 units; however, for R1 and R2 sensors the
result is in ppm units. We think the reason for that is because the polynomial coefficients were taken
straight in ppm units when the time of the calibration, like the documentation showed.
            The temperature compensation is made comparing the temperature during measurement process
Tm with the temperature of calibration Tc, following the equation as describe in page 2-9:

                                                   XT = X * (Tm/Tc)                                      (4)

where Tm and Tc are used in Kelvin.
           Because X was previously calculated in ppm units, this equation does not work (figure 4). For
that equation to work we have to find the right coefficients in absolute mmol/g units and then use the state
equation to transform to ppm units.

Fig. 4: Temperature compensation following instruction in the manual has no effect.

           Another way to make the temperature compensation is to use the raw data Y on contrary to X in
the equation. In other words, we make the temperature compensation one step before calculating X.

                                                   YT = Y * (Tm/Tc)                                      (5)
Progress Report CD-03, Fitzjarrald/Moraes                                                                 27

            The problem with using this equation is that the Tm is not placed into the measure cell. We can
solve this problem by putting temperature sensor into the tube line of the airflow. However, for now we are
going to use the temperature from R1 and R2 any way. Figure 1 the temperature behavior of R1 and R2 are
very similar to the Licor temperature with correlation of 0.98 and residual standard error of 0.2 oK.
            Figure 5 shows the result of the correction using the raw data Y and the signal after linear
regression. The errors for R1 and R2 are 6.3 and 7.27 ppm respectively with data correlation of 0.98. This
result is much better than the original data where the errors are 14.98 and 15.33 ppm respectively with data
correlation of 0.8. Using 10 seconds averaging in the data the error verified was 5.91 and 6.87 to R1 and R2
respectively. Using 60 seconds averaging the error is about 5.25 to R1 and 6.46 to R2.
            We can explain that errors either for to take the temperature not measured into the cell or the
temperature is not the only parameter to be considered. Also, the polynomial coefficient could be not right.
            To test the first supposition we can use the Licor temperature to make the correction. However,
the results showed that the errors increase to 9.9 ppm in both R1 and R2.

Fig. 5: Top: the Licor, R1 and R2 signals with temperature compensation using the raw data Y (equation 5).
Bottom: the signals after making the linear regression.

          Looking at figure 6 we can see that pressure measured by Licor drops just in the same period
where the major difference between Licor and Russian sensors occurs, after to make linear regression. This
could be the answer for the second guess. However, here we have to be careful because we do not know if
the pressure into the Russian device is the same as the pressure into the Licor cell. The cells have different
dimensions and the volume of the air passing into the cell is not the same either.
Progress Report CD-03, Fitzjarrald/Moraes                                                                  28

Fig. 6: The time series of temperature and pressure.

          The best result using the pressure and temperature together to correct the signal is the following:

                                                YT = Y * (Tm/Tc)* (Pc/P),                                  (6)

where Pc is the calibration pressure.
           Figure 7 shows the result of that correction. The error practically does not change, for R1 is
equal 6.9 ppm and 7.1 ppm to R2. The only difference is that the R1 and R2 signals after correction are
closer with other and the offset change too.

Fig. 7: Correction using the temperature and pressure.

          When we look at the signal of the difference or rate between Russian and Licor (figure 8) after
temperature compensation, we note that the signal still has one temperature influence. It could be because
we do not use the temperature into the cell to make the temperature compensation, so the signal still has
temperature influence.
          We are going to use second correction to remove the remainder temperature influence in the
Russian signal. We found one function of temperature for the rate of Russian and Licor:
Progress Report CD-03, Fitzjarrald/Moraes                                                                  29

                                 f(T) = RT / Licor                                                         (7)

                                 Rf = RT / f(T)                                                            (8)

where RT is the Russian signal with temperature compensation. The first guess for f(T) is linear equal:

                                  f(T) = -1.757 + 0.0092*(T) ; to R1                                       (9)
                                f(T) = -2.304 + 0.0110*(T) ; to R2                                        (10)

           After applied the equation 8 the error drops to 1.63 ppm to R1 and 1.89 ppm to R2, with
correlation of the 0.999 for both R1 and R2. The figure 9 shows the sequence of three plots for the Russian
signal from the original output with error about 15 ppm in the correlation, passing for the signal with
temperature compensation, with error about 5.6 ppm after remove the offset and finally the signal after
remove the remained temperature influence using a linear function of the temperature for the rate of
concentrations of the Russian and Licor which the error is about 1.7 ppm.

Fig. 8: The difference and rate of signal from Russian and Licor, after made the temperature compensation.
The signal after temperature compensation still has temperature influence.

Fig. 9: Sequence of the improvement of the Russian signal.

           We tried to improve this error using different coefficients for the equation 3 (calibration
equation), and also, change the value of the parameter D0 (zero calibration); however, that did not improve

the error verified, the major impact is over the value of the offset, decrease the value.
           The figure 10 shows the correlation of the Russian and Licor sensor for the original output
signal, for the signal with temperature compensation using the equation 5 and for the signal after to use the
equation 8.
      Progress Report CD-03, Fitzjarrald/Moraes                                                                30

      Fig. 10: Scatter plots for the original signal (top), after made the temperature compensation (middle) and
      after divide the signal by linear function of temperature.

      Soil Experiment (Large Range of CO2):

                 In this experiment we used a plastic box half filled with fresh soil. Eight runs were conducted;
      each run took about 20 minutes with the box close to see the build up of CO 2 concentration inside the box.
      We did this experiment to see what the response of the Russian sensors is when the CO 2 concentration
      changes suddenly.
                The figure 11 shows the time series for the temperature and concentration for the eight runs using
         a box with fresh soil and the table 1 shows the errors between Licor7000 and DX6100 sensors over all
           dataset for original output, after temperature compensation and after apply the equations 9 and 10.

                      Residual Standard Error (ppm) - R1              Residual Standard Error (ppm) - R2
1 Run                 3.61            2.78            2.61             2.26            1.78             1.5
2 Run                 3.02            3.08            2.91             2.33            2.30            2.24
3 Run                 3.21            2.71            2.41             1.67            1.48            1.33
4 Run                 2.53            2.79            2.33             2.11            2.14            1.64
5 Run                 2.40            2.70            2.09             1.93            1.71            1.38
6 Run                 2.31            2.16            1.99             1.44            1.57            1.38
7 Run                 2.52            2.35            2.29             1.78            1.50            1.44
8 Rum                 3.67            3.24            2.94             2.17            2.37            2.04
      Table 1: Shows the error for the original data (black), after temperature compensation (red) and after use
      equation 9-10 (blue).
Progress Report CD-03, Fitzjarrald/Moraes                                                                   31

Fig. 11: Entire dataset for the soil experiment. R1 is in black, R2 is in red and Licor is the blue line.

 Fig. 12: The scatters plots for original signal, after temperature compassion and after apply equation 9 and
           Because the temperature does not change too much in this runs we don‟t see one big difference
in the error showed in the table above.

Measurement in the field:

           For several days the sensors were run outside of the lab. We want to know how they work in
external environment over a large temperature range. We installed the Licor7000 and Russian sensors
inside a plastic box. The air sample was collected 5-meter far of the box at 1 meter high above the ground.
In the first three days the box was placed inside the room were the temperature in that period did not
change much. After, the box was placed outside without any shield to radiation. In that period the sensors
were exposed a high temperatures. Figure 13 shows the time sires of temperature and CO 2 concentration for
that period.
Progress Report CD-03, Fitzjarrald/Moraes                                                                32

Fig. 13: Time series for temperature and CO2 concentration from outdoors environment. Blue line is Licor
700, black line is R1 and red line is R2 measures.

            The spikes in the middle plot occurred when the temperature of the devices R1 and R2 goes up to
308 0K. When that occurs the device change the value of parameter D0 in the equation 2. In addition, a new
value for Tc must be considered to make the temperature compensation using equation 5, because the value
of D0 depends of the temperature of the gas in the moment of the zero calibration is made. The Russian
device cam saves 14 different parameters of calibration into the internal EPRROM. They came with 3
different tables from RTM Inc as we can see in the calibration specification manual.
          The bottom plot in the figure 13 shows the result after temperature compensation using equation 5.
      Easily we can see three different values in the off set, it occur because the device also change the
 coefficients in equation 6. Therefore, we cannot apply the same coefficients in equations 9 and 10 for the
  entire time series, although the method still may be applied with good results even the sensors had been
suffering over heat. Figure 14 shows the concentration after temperature compensation still has temperature
                                   dependence, although the weight varies.
Progress Report CD-03, Fitzjarrald/Moraes                                                                 33

Fig. 14: Relationship with temperature and concentration after temperature compensation along the time

           In figure 15 the dataset is separate in three different clusters or groups with same slope as we
expected once we can see at least three different offset values.
           Analyzing separately which cluster making the temperature compensation and removing the
offset observed an error mean of the 4.95ppm for the first part of the dataset (days 151-154, sensors inside
of the room), 5.13ppm between days 156-157 and among days 158-162 the error mean was about 6.01ppm.
And after to apply the right coefficients in the equation 9 and 10 we observed an error mean of the
2.13ppm, 2.73ppm and 2.51ppm for the same period respectively.
           To evict those differences of the offset in the data and that changes in the value of D0 we must to
protect the Russian device against the solar radiation. Alternatively, we could adjust the tables to work with
the same value of D0 for a large temperature interval.
           That experiment was very important to know how the Russian device works over extreme
environment conditions and to learn what we have to do to protect it against over heating. Also, is clear that
the temperature compensation using equation 5 works for different values of D 0 and the error after that is
about 5-6ppm and to apply the other correction using one function of temperature from equation 8 we shall
to know the right offset against Russian device and Licor7000 or another reference, so we can reduce the
error to more or less 2ppm.
Progress Report CD-03, Fitzjarrald/Moraes                                                                34

 Fig. 15: The scatter plots of concentration between Russian and Licor without temperature compensation
          (top), after to use equation 5 (center) and after to apply the equations 9 and 10 (bottom).


       The error is about 15 ppm without temperature compensation in the Russian sensor;
       Analog and digital outputs have the same signal;
       Because the temperature compensation is disabled, we cannot make the temperature compensation
        as the manual describe, we must use the raw data to make that correction (equation 5) and the
        error after that correction is about 5-6 ppm;
       Taking the average over 10 seconds after make the temperature compensation (equation 5) the
        error drop off about 6% and taking average over 1 minute the error drop off about 14%.
       Because the temperature used in equation (5) is not the temperature exactly into the cell the signal
        still has one temperature influence, that temperature influence is proportional of the rate of the
        concentrations; After removed that influence remainder, the error verified is about 1-2 ppm;
Progress Report CD-03, Fitzjarrald/Moraes     35

Appendix 4. Data Quality Assurance Officer:
Leon Zane Fitzjarrald, born 8/25/03.

To top