An important current problem in micrometeorology is the characterization of
turbulence in the roughness sublayer (RSL), where most of the measurements
above tall forests are made. There, scalar turbulent fluctuations display
significant departures from the predictions of Monin–Obukhov similarity
theory (MOST). In this work, we analyze turbulence data of virtual
temperature, carbon dioxide, and water vapor in the RSL above an Amazonian
forest (with a canopy height of 40 m), measured at 39.4 and 81.6 m above
the ground under unstable conditions. We found that dimensionless statistics
related to the rate of dissipation of turbulence kinetic energy (TKE) and the
scalar variance display significant departures from MOST as expected, whereas
the vertical velocity variance follows MOST much more closely. Much better
agreement between the dimensionless statistics with the Obukhov similarity
variable, however, was found for the subset of measurements made at a low
zenith angle

In the atmospheric surface layer above the roughness sublayer (RSL) height

In principle, the availability of mean concentration data would make flux-gradient methods a natural choice to estimate scalar fluxes above the forest. Unfortunately, the difficulty of applying MOST in the RSL adds considerable uncertainty to this approach.

Maybe one of the earliest reports of the failure of flux-gradient methods
when measurements are performed too close to the roughness elements was made
by

In the roughness sublayer, scalar and velocity gradients are weaker than
above, leading to higher values of the corresponding turbulent diffusivities

Other attempts to organize roughness sublayer data include the use of

In this paper, we analyze roughness sublayer data collected as part of the ATTO project (Amazon Tall Tower Observatory), a German–Brazilian project undertaken under the leadership of the Max Planck Institute (Germany), Instituto Nacional de Pesquisas da Amazônia (INPA), and Universidade Estadual do Amazonas (UEA) (Brazil). A 325 m tall tower has been erected as a forest site 150 km NE of Manaus and is currently undergoing instrumentation. Preliminary measurements have been made at an 82 m tall tower built at the site, and some analyses from the measured micrometeorological data are described here.

The main purpose of ATTO is to better understand the role of the
Amazonian biome in the context of global climatic changes. Specifically, the
project aims at a better understanding and modeling of gaseous exchanges between
the forest and the atmosphere

Given the importance of correctly estimating trace gas fluxes over the Amazon forest, the lack of a theory for the roughness sublayer is clearly a major obstacle in the understanding of surface–atmosphere interactions with far-ranging implications for the regional and global hydrology, ecology, and climate.

Moreover, given the always present need to take into account site-specific features in any micrometeorological study, we here attempt to provide a general analysis of roughness-sublayer-related questions at the ATTO site prior to the construction of the main tower. We expect that once measurements at the main tower become available, a better understanding of the questions preliminarily assessed here will be possible.

The variables analyzed in this study were chosen on the basis of data availability (the preliminary campaigns had to be restricted to fewer variables than those that will be available once ATTO is fully implemented) as well as their usefulness to assess two main questions: (i) how does the Amazonian RSL change the canonical (i.e., measured above the RSL and reported as “classical”) MOST similarity functions, and (ii) how useful or promising would flux-gradient and related methods applied within the Amazonian RSL be for the estimate of the fluxes of VOCs and other chemicals whose high-frequency measurement may be difficult to perform on a long-term basis?

In this work, the sections are organized as follows. In
Sect.

The study area is located at Reserva de Desenvolvimento Sustentável
Uatumã (RDSU) (Uatumã sustainable development reservation), in the counties
of São Sebastião do Uatumã and Itapiranga, in the northeast of Amazonas
State, Brazil. The site is 150 km northeast of the state capital Manaus,
between the coordinates 59

In the forest, between 200 and 250 tree species per hectare can be found, with a
mean height of 40 m and with some individuals reaching 50 m. The site
itself is located on a plateau (

The micrometeorological data were measured at an 82 m tower with a
rectangular cross section of

In this work, we analyze pilot data from the 39.4 and 81.6 m heights,
measured during April 2012. The data analyzed are the three wind components

Both the CSAT3 and the Gill sonic anemometers report sonic virtual temperatures. The
instantaneous values of

In the same vein, the instantaneous fluctuations of

Recently, it has been found that
the CSAT3 and Gill R3 sonic anemometers may require flow distortion corrections. For the CSAT3,
these corrections may change

The 10 Hz data were analyzed in 30 min data blocks (“runs”). Incomplete
runs were excluded, and spikes were removed following

As a result, 21.5 % of the 81.6 m level and 50.2 % of the 39.4 m
level runs were left. However, after this test, some strongly nonstationary
time series remained, mainly in scalar data, even when linear detrending was
applied. For this reason, these remaining data were further checked with two
tests (both after the removal of the linear trend). The first was the reverse
arrangement test (

As a percentage of the unstable runs only (which are the ones actually analyzed in this work), the figures are 9.4 % (81.6 m level) and 24.1 % (39.4 m level). Although typical of micrometeorological studies, this somewhat low number of usable runs is bound to limit, for example, the percentage of reliable fluxes that can be retrieved in long-term studies. It is also likely that a larger number of runs fail quality control checks in the RSL in comparison with standard applications of MOST in the surface layer. In this work, we needed to concentrate on the analysis of good-quality data at the expense of time coverage. Parallel efforts will be required to increase data availability and representativeness.

In this section, we briefly review some results, which are used in the next section to analyze the data.

The dimensionless dissipation rate of turbulence kinetic energy (TKE) is given
as follows

In the surface
layer, the skewnesses of

In convective or near-neutral conditions,

Consider the temperature spectrum in the inertial subrange, in the form

Also on dimensional grounds, and when MOST is valid, a similarity function exists
that describes the temperature spectrum in the inertial subrange, viz.,

Histograms of

Again, we seek to determine to what degree

The original eddy accumulation method was proposed by

Although initially developed and tested for the measurement of CO

In the ATTO project, we will focus on the fluxes of several chemicals whose high-frequency measurement is still too laborious, too expensive, or downright impossible. Among these compounds are the VOCs released by the forests, the monoterpenes and isoprene being the most abundant followed by alcohols, carbonyls, acids, aldehydes, ketones, and esters. If applicable, the REA method will be an invaluable tool.

However, strictly speaking, to be valid the method requires that the same
value of

Therefore, before applying the REA method to measurements made close to the
canopy over a forest, it is important to assess both the validity of scalar
similarity and the equality of the

REA coefficients measured above the roughness sublayer, where

Horizontal (

The rates of dissipation of TKE for each run were calculated from the
longitudinal spectra on the basis of Kolmogorov's local isotropy theory

The results can be seen in Fig.

We find that, close to the
canopy top, at 39.4 m, the integral scale

In Fig.

REA coefficients measured within the roughness sublayer

Inertial subrange similarity for temperature spectra:

Similarly to the analysis of the longitudinal velocity spectra, we identified
for each run the inertial subrange of the temperature spectrum and fitted a
linear regression with a

This analysis suggests that inertial-subrange scales, approximately in the
range 0.02–0.8 Hz, are also influenced by roughness sublayer effects: in
other words, restricting the analysis to a range of smaller scales does not appreciably improve the predictions of MOST. Similar plots and results were
also obtained for the other scalars (water vapor and CO

Dimensionless standard deviation for vertical velocity

Dimensionless standard deviation for vertical velocity

The “variance method”, pioneered by

In the particular case of

The randomized

We also calculated the scalar fluxes for fairly to highly unstable conditions
(

Figures

Once more, the large scatter typical of roughness sublayer data is found:
note that the scatter is much larger in

This tendency of

The

Note that this RSL “excess variance” (in comparison with MOST predictions)
does not impact the fluxes

Dimensionless standard deviation with solar angle: 39.4 m. First
row shows the velocity similarity function, the second shows temperature
function, and the third and fourth show CO

It is known that the zenith angle (

In Fig.

Scalar flux similarity indices rte (above) and ste
(below) for 0

Scalar flux similarity indices rte (above) and ste
(below) for 20

These results are encouraging: at least in the hours around noon, similarity
relationships as good as those observed in the surface layer over low
vegetation can be obtained. This opens up the possibility of retrieving
fluxes, by means of a host of standard MOST methods for these hours of the
day. The results also require explanation. It is not immediately clear why
these low zenith angles produce best results, but at least two (entirely
phenomenological) explanations seem possible. One is vertical heterogeneity
of sources and sinks, such as highlighted by

Scalar flux similarity indices rte (above) and ste
(below) for 60

Still with respect to our results
for the zenith angle, it is important to mention that similar effects were
found by

Not surprisingly, it turns out that success or failure of MOST predictions in
the roughness sublayer is also related to the degree of scalar similarity,
although it seems that this aspect of RSL turbulence has not yet been fully
explored. Two simple indices that are able to describe similarity between the
fluxes of two scalars

Relaxation coefficient from eddy accumulation method, where

Both

Relaxation coefficient variation with solar angle –
0

Relaxation coefficient variation with solar angle –
20

For most of the runs in the range
0

The similarity indices for the other zenith angle intervals are shown in
Figs.

Statistics from REA at 39.4 and 81.6 m, where

Relaxation coefficient variation with solar angle – 60

As reviewed in Sect.

Since there was no REA system installed at the tower, we have simulated the
method using the eddy covariance data. The fast scalar data were used to obtain
updraft and downdraft REA samples by conditional sampling of

Given the results found in the previous subsection for rte and
ste, it is natural to expect the REA method to perform better,
again, in the range 0

The overall results, not classified according to zenith angle, are shown in
Fig.

In Table

Our mean values of

Statistics from REA for each zenith angle class.

The behavior of

We classify the

For 0

A version of the REA for momentum is possible. Although it is seldom used to
estimate

Statistics for

The plots of

An experimental study of the behavior of scalars in the roughness sublayer has been made, with the objective of assessing their departure from the predictions of MOST.

The TKE dissipation rate departures are larger at the 39.4 m
level and smaller at the 81.6 m level, suggesting a gradual transition out of
the roughness sublayer. This is not confirmed, however, by all turbulence
statistics that we analyzed. For example, the dimensionless scalar standard
deviations (

Moreover, an analysis of the scalar dissipation rates did not reveal any improvement in scalar behavior on smaller (i.e., inertial-subrange) scales, indicating that the observed departures from MOST are occurring on all scales.

A significant finding in this work is that the degree of departure from MOST
predictions is related to the zenith angle. This was found to impact several
MOST functions, like

Fairly good adherence of

The same pattern observed for the dimensionless scalar standard deviations
appears with regard to the flux similarity indices rte and
ste. Again, they are closer to the theoretical values of

Finally, the

All data used in this study are kept in the ATTO Databases at Instituto de
Pesquisas da Amazônia and Max Planck Institut Für Chemie. The overall
project description can be found at

We thank the Max Planck Society and the Instituto Nacional de Pesquisas da Amazonia for continuous support. We acknowledge the support by the German Federal Ministry of Education and Research (BMBF contract 01LB1001A) and the Brazilian Ministério da Ciência, Tecnologia e Inovação (MCTI/FINEP contract 01.11.01248.00) as well as the Amazon State University (UEA), FAPEAM, LBA/INPA, the Luxembourg Institute of Science and Technology (LIST), and SDS/CEUC/RDS-Uatumã. Einara Zahn thanks Brazil's CAPES for her master's scholarship. Leonardo Sá, Antônio Manzi, and Nelson L. Dias thank Brazil's National Research & Technology Development Council (CNPq) for their “Productivity in Research” Grants 303728/2010-8, 312431/2013-9 and 303581/2013-1. Edited by: G. Fisch Reviewed by: two anonymous referees