Net ecosystem CO2 exchange of mixed forest in Belgium over 5 years

Agricultural and Forest Meteorology119(2003)

209–227

Net ecosystem CO2exchange of mixed forest

in Belgium over5years

Arnaud Carrara a,Andrew S.Kowalski a,Johan Neirynck b,

Ivan A.Janssens a,Jorge Curiel Yuste a,Reinhart Ceulemans a,?

a Research Group of Plant and Vegetation Ecology,Department of Biology,

University of Antwerpen(UIA),Universiteitsplein1,B-2610Wilrijk,Belgium

b Institute for Forestry and Game Management(IBW),Ministry of the Flemish Community,

Gaverstraat4,B-9500Geraardsbergen,Belgium

Received20November2002;received in revised form6May2003;accepted14May2003

Abstract

In this paper,we present and discuss the annual net ecosystem exchange(NEE)results from5years(1997–2001)of continuous eddy covariance measurements of CO2?ux above a mixed temperate forest.The forest was a70-year-old coniferous (Scots pine)—deciduous mixture,with slow growth rate and a leaf area index(LAI)of about3,and was part of the European CARBOEUROFLUX research network.Effects of the data pre-treatment and the gap?lling method on annual NEE estimates were analyzed.The u?-correction increased the annual NEE by+61g C m?2per year on average.The maximum difference in annual NEE estimates from different gap?lling methods amounted up to130g C m?2per year in a year with a large gap in the CO2?ux series.The estimated average annual NEE over the5years was+110g C m?2per year(ranging from?9to255g C m?2per year)when using the most defensible gap?lling strategy.We also analyzed the inter-annual variability of carbon balance,which was found to be mainly dependent on the length of the growing season and on the annual temperature.The observation that this forest acted as a CO2source contrasts with previous results from most other temperate forests.

?2003Elsevier B.V.All rights reserved.

Keywords:Carbon balance;Eddy covariance;Data gap?lling;Net ecosystem exchange;Inter-annual variability;CARBOEUROFLUX

1.Introduction

A better understanding of the global carbon bal-ance,and particularly of the role of terrestrial ecosys-tems represents an important current challenge to the scienti?c community,e.g.the“missing sink”issue (Schimel,1995;Ciais et al.,1995;Keeling et al.,1996; Houghton et al.,1998;Sarmiento and Gruber,2002).?Corresponding author.Tel.:+32-3-820-2256;

fax:+32-3-820-2271.

E-mail address:reinhart.ceulemans@ua.ac.be(R.Ceulemans).Given the importance of temperate forests in the global carbon balance(Reich and Bolstad,2001),measuring and modeling the net ecosystem exchange(NEE)of these forests has become a major research activity. Within this context(Baldocchi et al.,1996)net-works of tower-stations have been established per-forming long-term,continuous measurements of CO2exchange between forest ecosystems and the atmosphere(Baldocchi et al.,2001),including EU-ROFLUX(1996–1999)and CARBOEUROFLUX (2000–2003)which supported the present study. Recent investigations have shown that mature and

0168-1923/$–see front matter?2003Elsevier B.V.All rights reserved. doi:10.1016/S0168-1923(03)00120-5

210 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

undisturbed forests were generally sizable carbon sinks,although large variations in their CO2seques-tration rates were observed(Valentini et al.,2000; Falge et al.,2002).However,information about the inter-annual variability of NEE and the effects of management is still rather poor,but essential to char-acterize the role of forests in the global carbon budget. Furthermore,only very few long-term eddy covari-ance sites have been established in mixed conifer-ous/deciduous or patchy forest types,although these are typical of many regions in Europe.Despite the challenging circumstances,measurements over such sites can be helpful to study the speci?c behavior of the existing complex forests and take them into account in SV AT global models.

The purpose of this paper is to present5years (1997–2001)of eddy correlation measurements of CO2?ux above a managed mixed conifer-ous/deciduous forest in Belgium,and the resulting annual NEE.In order to achieve defensible annual sums of NEE,data quality tests and gap?lling proce-dures are required to provide consistent and complete datasets.The different data selection and gap?lling methods used are presented and their effects on the NEE estimates are discussed.We also analyze the inter-annual variability of NEE in relation to climatic variables.Since the forest was rather intensively man-aged during the measurement period,we also discuss the effects of forest management practices on the annual carbon budget.

2.Site description

The forest under investigation is‘De Inslag’,a mixed patchy coniferous/deciduous forest located in Brasschaat(51?18 33 N,4?31 14 E),in the Belgian Campine region(de Pury and Ceulemans,1997), about20km north-east of Antwerp.The site is part of the European CARBOEUROFLUX network (http://www.bgc-jena.mpg.de/public/carboeur/)and is a level-II observation plot of the European network program(ICP-II forests)for intensive monitoring of forest ecosystems(EC-UN/ECE,1996),managed by the Institute for Forestry and Game Management (Flanders,Belgium).The landscape is a coastal plain, almost?at(slope<0.3%)at a mean elevation of 16m.The climate is temperate maritime with a mean annual temperature of9.8?C and750mm of annual precipitation.The site is located in an area with high nitrogen deposition(30–40kg ha?1per year; Neirynck et al.,2002).

This relatively small(150ha)forest consists of many patches of different coniferous and deciduous species,with a variety of understorey species as well (Fig.1).Scots pine(Pinus sylvestris L.,80%of the coniferous species)and pedunculate oak(Quercus robur L.,80%of the deciduous species)dominate the canopy composition.The undergrowth is dominated by black cherry(Prunus serotina Ehrh.),rhododen-dron(Rhododendron ponticum L.)and grass(Molinia caerulea L.Moench),and an extensive moss cover (dominated by Hypnum cupressiforme Hedw.)charac-terizes some patches on the ground.A more complete description of the forest,with vegetation composition of the various patches has been previously published (de Pury and Ceulemans,1997;Janssens et al.,2000). Due to changing management policies over the last decades,the composition and structure of the forest vary among patches,but the forest remains relatively even-aged.The40m scaffold measurement tower is located within an even-aged Scots pine stand planted in1929.In this experimental stand,the current density is375trees ha?1,mean tree height is about21m,mean diameter at breast height(DBH)is30cm and stem basal area is27m2ha?1.About half of the forest con-sists of similar pine stands,and most of the remainder part consists of pedunculate oak stands of similar age (planted in1936)with tree density about320ha?1, mean tree height about17m,mean DBH about24cm and stem basal area about16m2ha?1.Leaf area index (LAI)varies considerably among patches and,in areas characterized by deciduous species,also seasonally. The area-weighted LAI,including both over-and un-derstorey LAI for the entire forest,is about3m2m?2 (between1.7m2m?2in winter and3.6m2m?2in summer,recalculated from Gond et al.,1999).

Most stands have canopy heights similar to the ex-perimental pine stand.The fetch is about500m in the west sector,the prevailing wind direction(Fig.1). Better conditions regarding fetch(>800m)are met in eastern sectors.The roughness length is about1m and the zero plane displacement is about19m.The forest is bordered to the north and west by the residential town of Brasschaat,and to the south and east by rural, partially forested terrain.

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

211

Fig.1.Map with composition of the forest surrounding the tower and wind rose.Wind frequencies are expressed in percent and computed by10?wind sectors for the period1997–2001:()deciduous;()coniferous;()mixed;()grassland;(?)measuring tower.

The soil is loamy sand,moderately wet,with a dis-tinct humus and iron B-horizon(Baeyens et al.,1993) and is classi?ed as Umbric Regosol(FAO classi?ca-tion,Roskams et al.,1997).Below the sandy layer, at a depth of1.5–2m,lies a clay layer.Due to this clay layer,the site has poor drainage,and groundwa-ter depth usually is between1.2and1.5m(Baeyens et al.,1993).A more detailed description of the phys-ical and chemical properties of the soil is available (Janssens et al.,1999;Neirynck et al.,2002).

3.Material/instrumentation

3.1.Eddy correlation measurements

The vertical?ux of CO2between the forest and the atmosphere was measured using the eddy cor-relation technique,originally proposed by Swinbank (1951)and applied to measure the net?ux of car-bon dioxide at ecosystem scale(NEE)since the1980s (Desjardins,1985;Verma et al.,1986;Baldocchi et al., 1988).Fluxes of CO2(F c),water vapor(λE)and sen-sible heat(H)have been measured continuously since mid-1996,from a scaffold tower at a height of41m. The CO2vertical?ux was calculated as the covari-ance between?uctuations in the vertical wind speed (w)and CO2concentration(c).The fast-response sen-sors used were a three-dimensional sonic anemometer (Model SOLENT1012R2,Gill Instruments,Lyming-ton,UK)for wind speed and temperature,and an in-frared gas analyzer(IRGA)(Model LI-6262,LI-COR Inc.,Lincoln,NE,USA)for CO2and water vapor concentrations.The IRGA was used in absolute mode with pure nitrogen?owing through the reference cell as zero air.Air is drawn through a1?m PTFE-Te?on ?lter(Gelman Acro50,PN4258,Ann Arbor,MI, USA)at6.2l min?1by a membrane pump(Model N811KNDC,KNF Neuberger,Freiburg,Germany) installed downstream of the analyzer.It then enters a 3m long PTFE-Te?on tube of4.33mm(inner diam-eter),heated to prevent condensation.A subsequent ?lter(Ballston300-01961,USA)cleans the air prior to sampling by the IRGA.The distance between the air inlet and the central point of the sonic measure-ments is about40cm.Calibration of the IRGA with a reference gas was performed monthly.The data were logged at20.8Hz and the?uxes computed in real

212 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

time using the EDISOL software(Moncrieff et al., 1997).

https://www.wendangku.net/doc/e59789453.html,plementary measurements

Above canopy meteorological measurements were made at the top of the tower(40m)and stored as half-hourly means on a data logger(Campbell CR10, CSI,Logan,UT,USA).They include global radiation (pyranometer,Kipp and Zonen CM6B,Delft,The Netherlands),net radiation(REBS07,Seattle,W A, USA),photosynthetically active radiation(PAR quan-tum sensor,JYP-1000,SDEC,Tours,France),precip-itation(tipping-bucket rain gauge,Didcot DRG-51, Didcot Instrument Co.Ltd.,Abington,UK),rela-tive humidity and temperature(psychrometer,Didcot DTS-5A,UK).

In order to estimate CO2storage in the air layer be-low the eddy correlation measurements height,mea-surements also include a pro?le of CO2concentrations at four levels(10,24,32and40m above the ground). From each inlet,air is drawn through53.5m of tub-ing to an instrument shelter,heated to35?C,and?l-tered through a0.5mm PTFE-Te?on?lter.Each level is sampled for a total duration of6min each half-hour by a gas analyzer(IRGA,LI-800,LI-COR Inc.,Lin-coln,NE,USA).

Other pertinent environmental variables include the soil measurements.Two heat?ux plates(Campbell HFT03,CSI,Logan,UT,USA)measured the soil heat ?ux(G).Soil temperature was measured with probes (Didcot DPS-404,UK)installed at2and9cm depth. Since January2001,CO2soil ef?ux has been mea-sured with a closed dynamic system(IRGA,CIRAS-1, PP SYSTEMS,Hitchin,Herts,UK),in nine patches (different associations of canopy/understorey vegeta-tion)representative for the forest composition. More details about the instruments and methods used for the measurements performed at the site can be found in Overloop and Meiresonne(1999)and in Kowalski et al.(2000)for the complementary mete-orological measurements,and in Curiel Yuste et al. (2003)for the soil ef?ux measurements.

4.Methods

The NEE(biotic CO2exchange,F NEE)can be deduced from the conservation equation of a scalar (CO2),in applying the Reynolds decomposition (Stull,1988).Assuming stationarity and horizontal homogeneity of turbulence,and under the hypothesis of negligible horizontal?ux divergence and molecu-lar diffusion,F NEE can be written as(Aubinet et al., 2000)

F NEE=F c+F S+advection terms(1) where F c=w c is the eddy correlation?ux and F S is the CO2storage in the air layer below the eddy mea-surements height.Since a reasonably accurate estimate of the advection terms was impossible with available measurements,they were neglected in this study and F NEE was simply deduced from the eddy?ux and the storage term.Following the micrometeorological sign convention,these terms are de?ned such that positive values represent release from the ecosystem(upward ?ux),and negative values represent an uptake by the ecosystem(downward?ux).

4.1.Calculation of eddy correlation?uxes

The measurement system and the data treatment both follow the guidelines of the standard EU-ROFLUX methodology(Aubinet et al.,2000).The half-hourly mean?ux values(covariance w c )were computed in real-time(by the EDISOL software, Moncrieff et al.,1997)with a running-mean removal technique based on a digital recursive?lter with a time constant of200s(McMillen,1988).The three-angle co-ordinate rotations of the wind vector were applied in order to remove the effects of instrument tilt or ter-rain irregularity on the air?ow.The time delay of the IRGA signal was set by determining the time lag op-timizing the correlation between vertical wind speed (w)and CO2concentration(c),within the range of 1–4s(the theoretical time lag calculated from?ow rate and tube characteristics:1.6s).

Corrections for high frequency response losses in the eddy?ux system were examined in detail by Aubinet et al.(2000)for the standard EUROFLUX closed chamber system,using description of the transfer functions of eddy covariance systems from the literature(Moore,1986;Leuning and King,1992; Leuning and Judd,1996;Moncrieff et al.,1997).It was shown that the only important effects needing correc-tion were the damping of scalar concentration?uctu-ations in the IRGA sampling tube,and the separation

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227213

between the IRGA sampling tube inlet and the sonic anemometer.In our case,these effects were negligible due to the short tube length and small sensor separation (Kristensen et al.,1997).There was also no need to ap-ply the“Webb”corrections for air density?uctuations (Webb et al.,1980)because of the closed-path con-?guration of the LI-6262gas analyzer.Consequently, no correction was made to the real time computed ?ux.

The applicability of the eddy correlation method is limited by a number of restrictive assumptions (Baldocchi et al.,1988;Dabberdt et al.,1993; Foken and Wichura,1996).These include horizontal homogeneity of the upwind surface,homogeneity of the turbulence and mean?ow,and stationarity.The general reliability of the?ux measurements with this eddy system set-up was?rst veri?ed by examination of spectra and co-spectra(comparison with the model of Kaimal et al.,1972)for several measurements performed under different meteorological conditions. The spectral characteristics of the?uxes were not examined in the routine treatment.However,a qual-ity control(QC)program(Vickers and Mahrt,1997) was run systematically on the raw data in order to reject poor quality data.All data failing the QC were then removed from the set of half-hourly?ux data output by the EDISOL software.Furthermore in the routine treatment,a quality test was applied to in-tegral turbulence characteristic of the vertical wind (σw/u?)as recommended and described by Foken and Wichura(1996).In order to discard the most critically non-stationary CO2?ux measurements,which could signi?cantly in?uence the NEE sums,a simple test was also routinely performed limiting the variances of vertical velocity and CO2concentration(σ2w and σ2c)below a threshold.This simple test has shown (comparative study on a2-month-period)its ability to remove most of the?uxes not ful?lling the insta-tionarity test proposed by Foken and Wichura(1996), and was preferred for its applicability to routine treatment.

4.2.Storage term computation

The storage of CO2in the layer below the eddy measurements(F S)was estimated by the simple ap-proach using only the change in CO2concentration measured at41m by the LI-6262(Hollinger et al.,1994;Greco and Baldocchi,1996):

F S=

c(z)

t

z(2)

where c(z)is the change in CO2at the height z, t the time period,and z the height of the layer.This method was chosen for the systematic treatment of all?ux data in order to avoid additional data rejection due to failure of the CO2pro?le measurement system during the1997–https://www.wendangku.net/doc/e59789453.html,parison with the storage term F S determined from the full pro?le measurements(method as described by Aubinet et al., 2001)validated this simple approach.The two meth-ods showed good overall agreement(Fig.2)and the associated discrepancies in annual sums of NEE were negligible(<10g C m?2per year).The morning and early evening showed the largest discrepancies,due to the inaccuracy of the simple approach to

estimate

Fig.2.Mean diurnal course of the storage term(F S)during the years2000and2001,estimated with two different methods: the simple approach using the LI-6262CO2measurements(solid line)and the calculation from CO2pro?le measurements(dotted line).For each year,averages are calculated for the periods of quali?ed CO2eddy?ux data and when both systems were properly functioning(LI-6262and pro?le).

214 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

important F S associated with transitory conditions. In order to remove these events for systematic treat-ment,a test excluded data for which:

|[CO2]t+ t?[CO2]t|+|[CO2]t?[CO2]t? t|>10ppm v This test therefore limited the storage term F S to a maximum absolute value of5.1?mol m?2s?1.

4.3.Corrections due to low-mixing

conditions(u?-correction)

4.3.1.Nighttime

It is now accepted by the eddy?ux community that, during stable nighttime conditions,surface exchanges are underestimated by eddy covariance measurements, owing to a lack of turbulent transport(Goulden et al., 1996;Moncrieff et al.,1996;Aubinet et al.,2000; Falge et al.,2001).This underestimation of night-time CO2?uxes constitutes a selective systematic error,and as such can lead to serious problems when long-term budgets are estimated by integration of short-term?ux measurements(Moncrieff et al.,1996). Despite the awareness of the problems in accurately determining nighttime?uxes,no general consensus has been reached for correcting the?ux(Falge et al., 2001).The method most widely used to correct for ?ux underestimation during stable nights was applied in this study.It consists of replacing the?ux mea-sured during stable nighttime periods(de?ned by u?below a threshold)by a value simulated with a tem-perature response function derived during well-mixed nighttime conditions(de?ned by u?above this thresh-old).The function that was used is the Lloyd and Taylor respiration equation(Lloyd and Taylor,1994, Eq.(11)):

F NEE,night=F RE,night

=F RE,283e A[(1/(283.16?T0))?(1/(T K?T0))](3) where T K is air temperature(in K),A is set to309K, whereas T0and F RE,283the respiration rate at refer-ence temperature(283.16K=10?C),are the?tted parameters.In our case,we used air temperature as the principal driver since it gave better results(higher R2values)than soil temperature.

Temperature response functions were evaluated for all the1997–2001nighttime data sorted for u

?Fig.3.Parameter describing the nighttime ecosystem respiration at the reference temperature of283K(F RE,283)estimated using the equation of Lloyd and Taylor(1994)(?tted using the least square method)from all the nighttime data(1997–2001)of storage corrected(?)and not storage-corrected(?)eddy CO2?uxes, sorted by u?classes(0.05m s?1classes).

(0.05m s?1u?classes).In our case,the derived values for the parameter(F RE,283)describing the ecosystem respiration rate at10?C showed a typical saturation above a u?threshold value(Fig.3).Indeed,the CO2 biotic ef?ux corresponding to nighttime ecosystem respiration is considered as independent of turbulence (u?)for biophysical and physiological considera-tions.It appears that for low mixing conditions(u?< 0.2m s?1),storage did not account for the total loss of?ux(Fig.3),and some of the CO2that is released by the ecosystem seemed to leave the forest by as yet undetermined pathways.Therefore,the value of 0.2m s?1was used as threshold for the u?-correction. The potential risk of‘double counting’,if there is a morning?ush out of CO2,is avoided by applying the u?-correction on the storage-corrected?uxes.

4.3.2.Daytime

There is some likelihood that underestimation of the eddy?uxes under low-mixing conditions also occurs during daytime(Goulden et al.,1996;Blanken et al., 1998).In addition to the advection terms which are not accounted for in the eddy measurements,Sakai et al. (2001)argue that a possible explanation for underes-timating daytime?ux is the inadequate sampling of the turbulent long-period?uctuations under low-wind conditions in daytime.By illustrating the importance of low-frequency contributions to eddy?uxes mea-sured over rough surfaces,their study showed that, for a sample summer at a temperate deciduous forest, large eddies with periods from4to30min contributed

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227215 about17%to eddy CO2?ux.In the present study,the

200s running mean used for eddy?ux calculations ef-

fectively?ltered the contribution of the low frequency

eddies.

To estimate the effects of low mixing on daytime

CO2?ux,similarly to nighttime,light response func-

tions describing the daytime CO2?ux were evaluated

for daytime data sorted into u?classes.The function

that was used is a modi?ed form(see Falge et al.,

2001,Eq.(A.8))of the Michaelis–Menten equation

(Michaelis and Menten,1913):

F NEE=

a R g

1?(R g/1000)+(a R g/F GPP,opt)

?F RE,day

(4)

where R g is global radiation(W m?2)and the?tted

parameters are F RE,day,the ecosystem respiration dur-

ing daytime(?mol CO2m?2s?1),a the ecosystem

quantum yield(?mol CO2J?1),and F GPP,opt the op-

timum GPP(?mol CO2m?2s?1)at a R g value of

1000W m?2.

The parameters describing the daytime CO2?ux

were derived for bimonthly seasonal periods(Fig.4).

All parameters exhibited a positive correlation with u?.Since these estimates are derived from temper-ature pooled datasets,potential explanations are the

correlation between low u?and cold early-mornings

following stable nights(leading to lower F RE,day es-

timates for lower u?),and the correlation between

low u?and warm conditions during low-wind sunny

days(inhibition of photosynthesis leading to lower a

and F GPP,opt estimates for lower u?).An additional

explanation is the contribution of the storage term

which may be underestimated for lower u?(storage

only below eddy system height and then invisible

from it)and overestimated for higher u?(?ush out

of CO2),leading to lower estimates of both uptake

(F GPP,opt and a )and respiration(F RE,day)for lower u?.Nevertheless,for u?below0.2m s?1,the values of the regression parameters are too low to be consis-tently explained by these biases,and the u?-correction was applied similarly as for the nighttime.Day-time?uxes measured during low-mixing conditions (u?<0.2m s?1)were replaced by simulated values. Daytime u?-correction did not much affect the annual NEE estimates(average effect was only+5g C m?2 per year).This is because it concerned a

smaller Fig.4.Daytime ecosystem respiration(F RE,day),equivalent quan-tum yield(a )and optimum gross primary production(GPP)esti-mated using the Michaelis–Menten light response equation(?tted using the least square method)from the daytime storage corrected eddy CO2?uxes(1997–2001),sorted by u?(0.1m s?1classes) and by bimonthly seasons.Error bars represent the95%con?-dence intervals.For reasons of clarity,error bars are presented only one-sided in the F RE,day and a graphs.

amount of data than for nighttime and because the

underestimation is a systematic error(always under-

estimation)but not a selective error(occurring for

both positive and negative?uxes).Nonetheless,the u?-correction improved the?t of the light response equations,and signi?cantly increased the daytime

ecosystem respiration estimates(13%average in-

crease in F RE,day estimates from Michaelis–Menten

model).

216 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

4.4.Data gap?lling

4.4.1.Meteorological data gap?lling

Since F NEE?lling methods rely on meteorological

variables,particularly temperature and radiation,it is

essential to have reliable and continuous series of these

variables to achieve defensible NEE estimates.Lin-

ear interpolation was used to?ll small gaps(<2h).

Longer gaps in meteorological variables were?lled us-

ing redundant measurements when available.Gaps in

air temperature measurements(T a)were?lled with the

temperature from the sonic anemometer when avail-

able,or from a nearby meteorological station(about

20km,same elevation).Down-welling global solar ra-

diation(R g)was used as the radiation variable in F NEE

?lling routines,since it was the most robust radiation

measurement and showed a very strong relationship

(R2=0.99)with photosynthetically active radiation (PAR)(Ceulemans et al.,2003).Gaps in R g were?lled

using PAR measurements or net radiation when PAR

was not available.Remaining R g missing data(1%

of daytime)were?lled by the mean diurnal variation

method based on a few adjacent days,selected visu-

ally and manually,exhibiting a similar diurnal tem-

perature pattern.Gaps in precipitation were?lled with

data from a nearby meteorological station.Gaps in soil

temperature T s(1.8%of all values)were inferred by

simple model predicting T s/ t from(T a?T s). 4.4.2.CO2?ux gap?lling

During the period analyzed(1997–2001),the aver-

age CO2?ux(F NEE)data coverage was71%due to

maintenance and system failures.The different qual-

ity tests,including u?-correction,discarded36%of

the measured F NEE data,leading to an average yearly

data coverage of only46%(Table1).This is rather

low,compared to the average coverage of65%for the

28yearly dataset included in the study on gap?lling

by Falge et al.(2001).Therefore,the estimation of

annual or monthly sums of NEE might be highly de-

pendent on the used gap?lling procedure.In order to

test the sensitivity of NEE to the gap?lling procedure,

different commonly used gap?lling strategies were

applied to the5-year dataset,including non-linear

regressions,look-up tables and neural networks.

However,look-up tables were subsequently rejected,

due to poor performance during periods of poor data

coverage.Table1

Percentage of F NEE(net ecosystem exchange)measured and qual-i?ed(not rejected by quality check tests,including u?-selection) data for the5years of measurements

Period Existing data(%)Quali?ed data(%)

Total Daytime Nighttime 199762.035.238.232.1 199871.147.151.942.3 199975.650.658.641.4 200080.552.755.248.9 200167.643.651.335.1

1997–200171.445.851.740.0

4.4.2.1.Semi-empirical methods:non-linear regressions

Semi-empirical gap?lling preserves the response of F NEE to the expected controlling factors(tempera-ture and radiation),as found in the data.Responses are described by non-linear regressions.These methods were applied following procedures and recommenda-tions from Falge et al.(2001).Missing values of F NEE are?lled with regression relationships established be-tween F NEE and controlling factors(temperature and radiation)for bimonthly periods(January–February, March–April,...).Daytime and nighttime data were addressed separately.For nighttime data,the Lloyd and Taylor respiration equation(Eq.(3))was used with air temperature as driver to model F NEE.For daytime F NEE,the Michaelis–Menten light response equation(Eq.(4))was used.The effect of temper-ature on the light response equations was not taken into account,as the regressions were performed on data not temperature sorted.This approach was chosen since splitting the dataset into temperature classes made some of the seasonal regressions un-stable because of the rather high percentage of gaps.

Bimonthly periods were selected to perform the re-gressions because it appeared to optimize the trade-off between having suf?cient data for statistically valid regressions,and describing seasonal variations of for-est carbon exchange.Nevertheless,this choice posed some problems because of some long-term system failures,yielding regressions that were highly biased or were even impossible.In order to be able to?ll the long-term gaps with data measured during the same bimonthly period from other years,regression

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

217

Fig.5.Bimonthly NEE sum differences between NEE N-N(gaps ?lled using the half-hourly neural networks method)and NEE nl-reg (gaps?lled using the non-linear regressions method)versus per-centage of missing data.The standard deviation is calculated from the different versions of half-hourly neural networks. relationships were also established for each bimonthly period pooled over5years.

4.4.2.2.Neural networks

We have used neural networks as a tool for?lling gaps in the F NEE series at two different time scales: half-hourly values and daily series.

Neural networks for gap?lling at half-hourly scale. The F NEE series?rst was screened by quality tests (including u?-selection)and sorted by bimonthly pe-riods.Gaps in the F NEE half-hourly dataset were then ?lled using a single-hidden-layer feed-forward neural network(Saxén and Saxén,1995)by predicting F NEE from four variables:global radiation,air temperature, soil temperature and relative humidity.Attempts to introduce precipitation or other additional variables were unfruitful.Each bimonthly period was treated separately.The neural networks method at half-hourly scale posed some problems for periods with more than 80%missing data.For these periods,the established networks were dependent on training options,and re-sulting NEE therefore was dependent on network ver-sions(Fig.5).For periods with fewer gaps,the neural network technique applied successfully(different ver-sions were very close)and the NEE results were very similar to non-linear regressions(Fig.5),as the most important drivers of neural networks(temperature and radiation)are the driving variables of the non-linear regressions.

Neural networks for gap?lling at daily scale.The

F NEE series was screened by quality tests and sorted

by year.Days with more than a threshold number(12

or18)of missing half-hourly F NEE data were consid-

ered as“missing”,others as“good”.The“missing”

days values were?lled using single-hidden-layer

feed-forward neural networks(Saxén and Saxén,

1995)by predicting daily NEE from daily averages

of global radiation,air temperature,soil temperature,

relative humidity,day length and precipitation.Neu-

ral networks were trained with the daily F NEE sums

of the“good”days,from both measured and modeled

(using the non-linear regression method)half-hourly

F NEE values.The modeling method is assumed to

have a negligible effect on the F NEE daily sums of

the“good”days,as the amount of?lled data remains

low.The average missing data per“good”day repre-

sented respectively11.6and16.7%,for respectively

12and18as the maximum number of half-hourly

gaps allowed per day.The resulting neural-network

?tted the measured daily NEE well(Fig.6),with R2

values ranging from0.91to0.97for the5years,and

with a satisfactory coverage of the entire annual range

of the driving variables.Unlike the methods based on

a bimonthly sorting,this method treats each year as a

whole and therefore enables to?ll the long-term gaps

without special operation.

5.Results and discussion

5.1.Energy balance closure

The energy budget includes energy terms obtained

by eddy covariance(sensible heat?ux H and latent

heat?uxλE)and by other methods(net radiation R n,

soil heat?ux G,heat storage S t).A comparison of

these terms can be used to check the quality of the eddy

covariance measurements(Aubinet et al.,2000).Fig.7

shows the half-hourly values of the eddy covariance

energy?uxes versus the other terms of the energy

budget equation.Despite the strong correlation(R2= 0.91),the sum of the eddy covariance?ux was lower

(slope of the regression=0.76)than the available

energy.The closure de?cit of the energy balance was

similar in the other years(slope between0.7and0.8).

Such a closure de?cit of about20%is rather common

with eddy covariance measurements above tall forest

218 A.Carrara et al./Agricultural and Forest Meteorology 119(2003)

209–227

Fig.6.Results of daily net ecosystem exchange (NEE)neural network training (maximum number of half-hourly gaps per day =18)for the years 1998(y =1.0001x ,R 2=0.97)and 1999(y =1.006x ?0.01,R 2=0.94).

canopies (Goulden et al.,1996;Lee,1998;Aubinet et al.,2000).The closure de?cit of the energy balance was found to be independent of the wind direction (data not shown).5.2.CO 2?uxes

The evolution of the biotic CO 2?ux (F NEE )and the main meteorological variables,measured

between

Fig.7.Energy balance closure at half-hourly scale in 2001.Eddy covariance energy ?uxes (H +λE )vs.available energy (R n ?G ?S t ).H sensible heat ?ux;λE latent heat ?ux;R n net radiation;G soil heat ?ux;S t heat storage in canopy air.The linear regression (solid line)equation is (H +λE)=0.76(R n ?G ?S t )?16.8(R 2=0.91).

1997and 2001,are displayed in Fig.8.Climatic con-ditions are characterized by a rather high variability at synoptic scales in all seasons.Precipitation was dis-tributed rather regularly in the course of the year,and drought stress episodes were quite limited (and had no signi?cant effects on tree performance or physiol-ogy,as shown by Meiresonne et al.,2003).Annual total precipitation was above the long-term aver-age (750mm)for every year except 1997(658mm).The average daily air temperature was typically around 5?C during winter and around 15?C during the growing season.The incoming global radiation reached 30MJ m ?2per day in summer and was lower than 5MJ m ?2per day during winter.Maximum measured nighttime ?ux (respiration)was between 5?mol m ?2s ?1in winter and 8?mol m ?2s ?1in summer.Maximum measured negative (uptake)?ux was about ?20?mol m ?2s ?1in https://www.wendangku.net/doc/e59789453.html, uptake ?uxes during daytime were measured even in winter (Fig.8c ),which is mainly due to photosynthesis by the https://www.wendangku.net/doc/e59789453.html, uptake on a daily basis,however,only occurred from the beginning of April until the end of September (Fig.8d ),leading to a maximum length of the growing season of about 6months.From monthly mean diurnal courses (not shown),we observed that the average nighttime ?ux (dark respiration)varied from about 2?mol m ?2s ?1during the coldest months (December–January)to about 5?mol m ?2s ?1during the warmest months (July–August).The average up-take ?ux during mid-day in the months with highest uptake (July and August)was ?10?mol m ?2s ?1.

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

219

Fig.8.Evolution from1January1997to31December2001of:(a)air temperature daily average;(b)daily precipitation;(c)daily incoming radiation;(d)daily NEE,periods with no data were?lled using the non-linear regression method;(e)half-hourly measurements (after quality check)of biotic CO2exchange:F NEE=eddy correlation?ux+storage term.

5.3.Effect of pre-treatment and gap?lling procedures on annual NEE estimates

The effects of pre-treatment(u?-correction and stor-age term)and gap?lling methods on annual NEE es-timates are summarized in Table2.During the5years studied,three large gaps in the CO2?ux series led to unsatisfactory results from the non-linear regres-sion method for three bimonthly periods(May–June 1997;May–June1998and January–February2001).

220 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

Table2

Annual NEE estimates(in g C m?2per year)obtained with different pre-treatment and gap?lling methods for the years1997–2001 Gap?lling method and pre-treatment1997a1998a199920002001a Average(5years)S.D.(5years) Non-linear regression u?-corrected b

(no storage term)

7c116c134229?19c9390

Non-linear regression not u?-corrected?17258176?53c5077

Non-linear regression u?-corrected67c/99d112c/48d130255?9c/?60d111c/94d86c/103d Neural network half-hourly scale u?-corrected8561134238?30c/?82d97c/87d88c/104d Neural network daily(gaps>18)u?-corrected3152133224?687498

Neural network daily(gaps>12)u?-corrected21?18137248?6964114

a Years containing a long-term gap.

b Nighttime u?threshold is0.4m s?1(instead of0.2m s?1for storage-corrected series).

c Missing bimonthly perio

d gap?lled with non-linear regressions from5-year-pooled data.

d Missing bimonthly period gap?lled with neural network at daily scale(gaps>18).

These were?lled with an alternative technique,the non-linear regressions from5-year-pooled datasets or the neural network gap?lling at daily scale.

The u?-correction strongly affected annual NEE es-timates,resulting in a more positive annual sum of NEE,+61g C m?2per year on average,with a maxi-mum of+79g C m?2per year in2000,the year with the highest ecosystem respiration estimates.This ef-fect is similar to results observed in previous studies (Lindroth et al.,1998;Schmid et al.,2000;Falge et al., 2001;Aubinet et al.,2002).In a study on the effects of different gap?lling procedures,Falge et al.(2001) also showed that the largest effect on annual NEE es-timates was due to the u?-correction(+65g C m?2 per year on average,up to+147g C m?2per year, for storage-corrected data).The effect of u?-correction was found to vary seasonally(stronger in summer than in winter),due to higher respiratory?uxes and higher frequency of stable conditions events in summer. The effect of the storage term correction on annual NEE was found to be weak when no long-term gap oc-curred,and providing that the nighttime u?-correction was applied with a suited threshold of0.4m s?1 (Fig.3).

Uncertainty associated with gap?lling can have two sources,the uncertainty introduced by using differ-ent?lling methods and the error introduced by the ?lling process itself.Errors introduced by the?lling process are mostly random errors,and partly com-pensated during time integration.Falge et al.(2001) showed that these random errors did not differ much between the methods tested in their study,and were directly proportional to the percentage of gaps?lled during a period.Their results showed that for the

non-linear regression method,the maximum observed

errors on annual NEE(for four different biomes)were ±0.40g C m?2per percentage?lled in the year for daytime and±0.36g C m?2%?1for nighttime.In our

case,the maximum error introduced thus ranged be-

tween±36and±49g C m?2per year(±41g C m?2

per year on average).

The NEE results from different gap?lling methods

varied mainly when the yearly dataset contained large

gaps.Annual NEE estimates were not very sensitive

to the gap?lling method during1999and2000,but

more sensitive during1997,1998and2001.In this

study,the absolute uncertainty in annual NEE asso-

ciated with the gap?lling strategy,for the same data

pre-treatment(storage term and u?-correction),was

64g C m?2per year on average,ranging between7

and130g C m?2per year.In years where gaps oc-

curred predominantly early in the growing season

(1997,1998and2001;see Fig.8),the neural network

gap?lling methods operating at daily scale produced

lower annual NEE estimates than the other methods.

This is probably due to the absence from the neural

network of a biotic parameter that could describe

the seasonal variation of the forest potential for CO2

exchanges(directly like LAI or indirectly like latent

heat?ux).The neural networks were trained with

annual datasets,but the NEE and the neural network

drivers(e.g.radiation)present asynchronous annual

variations.In those years(1997,1998and2001),

the early growing season gaps may have been?lled

with underestimated daily NEE values because of

the imbalance between early and full/late growing

A.Carrara et al./Agricultural and Forest Meteorology 119(2003)209–227221

season data.Therefore,the most defensible gap ?lling methods appeared to be,in this study,the non-linear regression and the neural networks at half-hourly scale,both operating on bimonthly periods.

For further analysis and discussion on NEE,we will use the annual estimates obtained with the non-linear regression method.5.4.Annual NEE estimates

The NEE estimates from the non-linear regression method exhibited an average of 111(±41)g C m ?2per year,an inter-annual maximum variability of 264g C m ?2per year and a standard deviation of 86g C m ?2per year over the 5years (1997–2001).The ?nding that this forest ecosystem was a CO 2source (positive NEE)contrasts with the previous re-sults from most of the other CARBOEUROFLUX for-est sites (Valentini et al.,2000;Berbigier et al.,2001;Pilegaard et al.,2001;Dolman et al.,2001;Aubinet et al.,2002).Only the Swedish site in Norunda (Lindroth et al.,1998)has reported a posi-tive NEE on long-term.The NEE results from other studies,including most of the CARBOEUROFLUX sites with similar climatic conditions are displayed in Table 3.The amplitude of the inter-NEE annual vari-ability at our site is similar to other studies,although two sites reported higher maximum variability:Hesse (France),which experienced signi?cant changes due to a violent storm in December 1999,and Norunda (Sweden),which reports on a longer period of mea-surement (Table 3).

Table 3

Overview of annual NEE and maximum NEE inter-annual variability observed at different CARBOEUROFLUX sites Site and references

Forest type Latitude NEE average

(g C m ?2per year)NEE inter-annual variability (g C m ?2per year)Period of measurement Brasschaat (present study)Mixed 51?18 N 1112641997–2001Hyytiala (Suni et al.,2003)

Coniferous 61?51 N ?194771997–2001Norunda (Grelle,A.,unpublished)Coniferous 60?05 N 2204501995–2001Soro (Pilegaard,K.,unpublished)Deciduous 55?29 N ?1771651996–2001Loobos (Moors,E.,unpublished)Coniferous 52?10 N ?3191751996–2001Vielsam (Aubinet et al.,2002)Coniferous 50?18 N ?7201901997–2001Deciduous ?4601901997–2001Hesse (Granier,A.,unpublished)Deciduous 48?40 N ?3225141996–2001Le Bray (Berbigier,P.,unpublished)Coniferous 44?05 N ?483881997–2001Harvard a (Goulden et al.,1996)

Deciduous

42?32 N

?220

140

1991–1995

a

From

AMERIFLUX.

Fig.9.Bimonthly sums of NEE for the years 1997–2001.Results obtained with non-linear regression gap ?lling method,presented with standard deviations of the bimonthly sums obtained with different gap ?lling methods.

5.4.1.Inter-annual variability in NEE

The cumulated NEE was computed for each bi-monthly season (Fig.9).The maximum inter-annual variations in bimonthly NEE varied from 23g C m ?2per month (January–February)to 37g C m ?2per month (July–August).This is similar to results from a nearby Belgian coniferous/deciduous forest site (Aubinet et al.,2002),which showed a max-imum inter-annual variation between 15g C m ?2per month in winter and 40g C m ?2per month in summer,over the same period of measurements (1997–2001).

At the annual scale,NEE was not signi?cantly cor-related with radiation (R 2=0.10)and weakly corre-lated with air temperature (R 2=0.26).Both observed effects were in the expected direction:NEE became more positive (towards a source)with increasing

222 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

temperature and more negative with increasing ra-diation.The highest correlation(R2=0.80)was observed between annual NEE and the length of the growing season(LGS).The sensitivity of NEE to LGS was?3.5g C m?2per year per additional day of growing season(Fig.10).It is rather similar to,al-though higher than the dependence observed by Falge et al.(2002)between annual NEE normalized for LAI and LGS results from several sites.Assuming a constant LAI of3m2m?2,we found?1.15g C m?2 leaf area per year per day in our case versus about ?0.85g C m?2leaf area per year per day for the tem-perate evergreen forest sites from Falge et al.(2002). The LGS regression residuals(Fig.10),computed as the difference between the measured NEE and the regression NEE estimates,were strongly correlated with temperature(R2=0.88),but not signi?cantly correlated with radiation(R2=0.04).

A simple linear model(from a multiple regression analysis)of NEE based only on LGS and annual mean temperature was therefore able to simulate the mea-sured annual NEE very well(R2=0.98).Despite the strong correlation observed,this dependence is of lim-ited use for predictive modeling,since LGS was esti-mated with the NEE value(period during which the 10-day averaged NEE was negative).The correlation was signi?cantly lower when the LGS estimate was based on temperature.This may be due to the impor-tance of the variations in temperature at synoptic scales compared to the yearly seasonal variation(difference between winter and summer average temperatures). However estimated,the length of the growing season certainly in?uences annual NEE.For the year2001, the annual NEE was lower due to the exceptionally low NEE during the bimonthly periods of the beginning (March–April)and the end(September–October)of the growing season(Fig.9).This stronger CO2uptake was supported by phenological observations(very late senescence observed for the deciduous species,leaf abscission only started in the beginning of December) and by inter-annual comparison of monthly relation-ships between daytime F NEE and radiation(R g)during autumn,which also suggest an exceptionally late end of the growing season in2001.

Despite the old age of most of the trees,this forest ecosystem was not in equilibrium state because of the rather intensive management over the last decade. During the period of measurements,many of the

for-Fig.10.(a)1997–2001annual NEE sums vs.length of grow-ing season(LGS).Different gap?lling methods used to calculate annual NEE:(?)non-linear regressions;(?)neural-network at half-hourly scale;( )neural network at daily scale.The linear regression is performed on the NEE sums calculated with the non-linear regressions gap?lling method.The linear regression equation is NEE=613?3.46LGS(R2=0.80);(b and c)regres-sion residual vs.mean annual temperature(R2=0.88)and annual incoming radiation(R2=0.04).Residuals are computed as the

difference between the measured NEE and the LGS regression NEE estimates.

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227223

est patches were subject to changes due to thinning or to partial removal of the understorey vegetation. The average LAI of the forest may therefore have de-creased during the5years of measurement,modifying both ecosystem respiration and uptake capacity of the forest.In November1999,the2ha pine stand sur-rounding the tower was subjected to a thinning(har-vest of30%of the trees),leading to a decrease of net primary production(NPP)of about25%(estimated from repeated biomass inventories,Xiao et al.,2003). The large number of different patches in the forest and their relatively small size not only made these changes in LAI quite frequent,but also made their ef-fect on the measured CO2?ux impossible to estimate without a reliable high-resolution footprint analysis. The inter-annual variability in NEE resulted from combined effects of these changes and of the changes in climatic conditions.It is impossible to deconvolute these effects without ancillary information,such as regular forest-scale LAI estimates.The relative contri-butions of changes in climate and management prac-tices to the inter-annual variability in NEE can there-fore not be quanti?ed from the CO2?ux measurements only.

5.4.2.Possible explanations for positive NEE

The average annual NEE over the measurement period(1997–2001)was about+110g C m?2per year.The net annual C increment in wood biomass of the pines was estimated to about180g C m?2per year for the period(1995–2001)(Xiao et al.,2003). Using standing biomass to extrapolate NPP from the pine stands to the oaks,resulted in a mean annual C increment for the entire forest of about150g C m?2 per year for(1995–2001).Different hypotheses,such as the effects of climatic anomalies and forest man-agement practices,can be put forward to explain this large discrepancy between the annual biomass C increment and NEE.

5.4.2.1.Climatic anomalies

During the5years of measurements,the annual temperatures were all above the normal average (5-years average was10.8?C and normal is9.8?C). Normal climatic conditions are the30-year means from a nearby(30km)meteorological station.Since respiratory processes are sensitive to temperature, annual NEE could be sensitive to a change in temper-ature.A sensitivity test was made by adding1?C to all half-hourly temperatures,then using the nighttime F NEE temperature response equations to estimate RE, the total annual ecosystem respiration(Lindroth et al., 1998).This rough estimate indicated a79g C m?2per year increase in RE(5-year average)due to a1?C in-crease in temperature.To estimate the effect on NEE, we should also consider the effect of temperature on GPP,but this is not trivial,as it largely depends on the distribution of temperature within the year.In fact, higher temperatures can increase GPP by causing an earlier start of the growing season,but can also de-crease GPP by occasionally limiting photosynthesis during warm days(stomatal closure).Assuming that the effect of increased temperature on GPP is smaller than on RE,NEE should be more negative(more up-take by the ecosystem)in“normal”years.The yearly total incoming global radiation was not higher than the long-term normal(5-year average was3535MJ m?2 per year and normal is3598MJ m?2per year).It also suggests that NEE should be more negative during “normal”years,as annual global radiation and GPP can be supposed to be positively correlated,although they are not expected to be strongly correlated on a yearly scale.

5.4.2.2.Forest management practices

The observed net C losses from the forest may also be partly due to the legacy of past management practices.Until the1980s,the forest was privately owned and not exploited for wood production but for hunting purposes.The original,homogeneous stock-ing density was very high and thinning was neglected (1390tree ha?1in pine stands in1980;Cermak et al., 1995),resulting in tall,slender stems.Furthermore, Rhododendron ponticum and Prunus serotina prolifer-ated and established a3m tall undergrowth.Biomass of this dense undergrowth was on average7t ha?1 where Rhododendron dominated(15%of the forest area),and35t ha?1where Prunus dominated(30%of the forest area).During the1980s,the forest became property of the Flemish region and a new management policy was implemented,including intensive thinning and removal of invasive species such as Rhododen-dron and Prunus.Due to this recent intensive thinning, some forest patches are approaching their degrada-tion stage;more openings in the canopy create better light conditions triggering decomposition of the thick

224 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

forest?oor and consequently enhancing respiration losses.

Based on detailed information on the harvested stems(number and DBH),and applying site-speci?c allometric relationships(Xiao et al.,2003),we calcu-lated that since1987an average of80g C m?2per year of slash was added to the forest?oor as a consequence of the thinning of pines and oaks.Decomposition of this slash remaining after the thinning was estimated to have released about50g C m?2per year during the 1997–2001period(based on exponential decay model with different parameterization for woody tissues and foliage/?ne roots;Janssens et al.,unpublished re-sults).Removal of Rhododendron and Prunus,could therefore easily have doubled the C ef?ux due to de-composition of slash.The resulting enhancement of the C ef?ux from the soil,could therefore have in-creased ecosystem respiration by about100g C m?2 per year,contributing to the observed positive NEE. This situation was possibly exacerbated by the re-moval of Rhododendron from large areas in the forest. Rhododendron inhibits germination of seeds,as well as decomposition of litter(Cross,1975;Rotherham, 1983).Therefore,the prevalence of Rhododendron may have resulted in retarded decomposition rates, and thus higher equilibrium soil organic C(SOC) contents.Since the removal of Rhododendron,the organic molecules retarding decomposition may have been broken down or leached out of the system, which may have restored the original,lower SOC equilibrium content.Thus,the forest?oor may have been losing C during the late1990’s,and thus have contributed to the observed positive C?uxes. However,the effects of this change in manage-ment should be transient.Now that the forest has been intensively thinned and most Rhododendron has been cleared,soil C losses should be reduced and also NEE should go down.At this site,continued, long-term eddy?ux measurements could therefore clarify whether our speculated management effect is indeed transient and over which time period these management effects are prolonged.

5.5.Annual respiration estimates

The annual ecosystem respiration(RE)was esti-mated using the F NEE nighttime temperature response functions(Eq.(3))and,optionally the ecosystem res-piration rate during daytime F RE,day from the light response relationships Eq.(4).We found a5years average of1300(±40)g C m?2per year(the standard error represents only the two different computation methods).This mean RE is close to the highest RE estimates reported within the EUROFLUX network sites(RE between600and1400g C m?2per year, mean value1100g C m?2per year;Janssens et al., 2001).This agreed with our hypothesis that the pos-itive NEE observed at our site could be due to a high ecosystem respiration,potentially caused by manage-ment and climate.

The annual soil respiration(SR)of the entire for-est was estimated to650(±60)g C m?2per year in 2001(Curiel Yuste et al.,unpublished results)using soil ef?ux measurements made in different patches. According to these estimates,soil respiration would account for about63%of the total ecosystem respira-tion(RE=1030g C m?2in2001),which is slightly below the mean SR/RE-ratio observed within the EU-ROFLUX network(69%;Janssens et al.,2001).How-ever,these soil respiration measurements were made on collars that did not contain slash remaining after thinning.Therefore,the true contribution of the soil to RE could be higher than63%.

6.Conclusions

As observed by previous similar studies,the u?-correction strongly affected the NEE annual sums, increasing the annual NEE by+61g C m?2per year on average.For the same data pre-treatment,the uncertainty in annual NEE associated with the gap ?lling strategy was found to be up to130g C m?2per year(in year with large gap).Studies on inter-annual variability in NEE should therefore avoid potential confounding effects by the gap?lling methodology. Over the5years of measurements,the uncertainty associated with the different gap?lling methods was reduced and the studied forest appeared to be a consistent net source of CO2110g C m?2per year (on average for1997–2001).This is different from the results of the other CARBOEUROFLUX for-est ecosystems under in?uence of similar temperate climate,which appeared to be net sinks.Forest man-agement practices and unusually elevated tempera-tures are suspected to have increased the ecosystem

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227225

respiration,and thereby the NEE,over the5years of measurement.

Acknowledgements

This research was?nancially supported by the EC’s Fourth(EUROFLUX contract ENV4-CT95-0078)and Fifth(CARBOEUROFLUX contract EVKL-CT-1999-00032)Framework Programs.The authors acknowl-edge the Division of Forests&Green Areas of the Ministry of the Flemish Community for access to the forest site,the Institute for Forestry and Game Management(IFG)and forest ranger M.Schuermans for logistic support.We are also grateful to F.Kock-elbergh(UIA),N.Calluy(UIA)and Y.Buidin(IFG) for technical assistance as well as to two anonymous reviewers for their useful and constructive comments. IAJ is a post-doctoral researcher of the Fund for Sci-enti?c Research,Flanders(FWO).This study also contributes to the GCTE Core Project of the Interna-tional Geosphere–Biosphere Program(IGBP). References

Aubinet,M.,Grelle,A.,Ibrom,A.,Rannik,ü.,Moncrieff,J., Foken,T.,Kowalski,A.S.,Martin,P.H.,Berbigier,P.,Clement, R.,Elbers,J.,Granier, A.,Grünwald,T.,Morgenstern,K., Pilegaard,K.,Rebmann,C.,Snijders,W.,Valentini,R.,Vesala, T.,2000.Estimates of the annual net carbon and water exchange of forests:the EUROFLUX methodology.Adv.Ecol.Res.30, 113–175.

Aubinet,M.,Chermanne, B.,Vandenhaute,M.,Longdoz, B., Yernaux,M.,Laitat, E.,2001.Long-term carbon dioxide exchange above a mixed forest in the Belgian Ardennes.Agric.

For.Meteorol.108,293–315.

Aubinet,M.,Heinesh, B.,Longdoz, B.,2002.Estimation of the carbon sequestration by a heterogeneous forest:night-?ux corrections,heterogeneity of the site and inter-annual variability.

Global Change Biol.8,1–19.

Baeyens,L.,Van Slycken,J.,Stevens,D.,1993.Description of the soil pro?le in Brasschaat.Internal research paper.Institute for Forestry and Game Management,Geraardsbergen,Belgium. Baldocchi,D.D.,Valentini,R.,Oechel,W.,Dahlman,R.,1996.

Strategies for measuring and modeling carbon dioxide and water vapor?uxes over terrestrial ecosystems.Global Change Biol.

2,159–168.

Baldocchi, D.D.,Hicks, B.B.,Meyers,T.P.,1988.Measuring biosphere–atmosphere exchange of biologically related gases with micrometeorological methods.Ecology69,1331–1340. Baldocchi, D.D.,Falge E.,Gu,L.,Olson,R.,Hollinger, D., Running,S.,Anthoni,P.,Bernhofer,C.,Davis,K.,Fuentes,J.,

Goldstein,A.,Katul,G.,Law,B.,Lee,X.,Mahli,Y.,Meyers, T.,Munger,W.,Oechel,W.,Paw,U.K.T.,Pilegaard,K.,Schmid,

H.P.,Valentini,R.,Verma,S.,Vesala,T.,Wilson,K.,Wofsy,S.,

2001.FLUXNET:a new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide,water vapor and energy?ux densities.Bull.Am.Meteorol.82,2415–2434. Berbigier,P.,Bonnefond,J.M.,Mellmann,P.,2001.CO2and water vapor?uxes for two years above a EUROFLUX forest site.

Agric.For.Meteorol.108,183–197.

Blanken,P.D.,Black,T.A.,Neumann,H.H.,Den Hartog,G.,Yang, P.C.,Nesic,Z.,Staebler,R.,Chen,W.,Novak,M.D.,1998.

Turbulent?ux measurements above and below the overstorey of

a boreal aspen forest.Boundary-Layer Meteorol.89,109–140. Ceulemans,R.,Kowalski,A.S.,Berbigier,P.,Dolman,A.J.,Grelle,

A.,Janssens,I.A.,Lindroth,A.,Moors,E.,Rannik,U.,Vesala,

T.,2003.Coniferous forests(Scots and Maritime Pine):carbon and water?uxes,balances,ecological and ecophysiological determinants.In:Valentini,R.(Ed.),Fluxes of Carbon,Water and Energy of European Forests.Ecological Studies,vol.164, Springer-Verlag,Berlin,Heidelberg,pp.71–97.

Ciais,P.,Tans,P.P.,White,J.C.,Francey,R.J.,1995.A large northern hemisphere terrestrial CO2sink indicated by the 13C/12C ratio of atmospheric CO2.Science269,1098–1102. Cross,J.R.,1975.Biological?ora of the British Isles.List Br.

Vasc.Pl.(1958)No.345.1:Rhododendron ponticum L.J.Ecol.

63,345–364.

Curiel Yuste,J.,Janssens,I.A.,Carrara, A.,Meiresonne,L., Ceulemans,R.,2003.Interactive effects of temperature and precipitation on soil respiration in a temperate maritime pine forest.Tree Physiol.,in press.

Dabberdt,W.F.,Lenschow,D.H.,Horst,T.W.,Zimmerman,P.R., Oncley,S.P.,Delany,A.C.,1993.Atmosphere–surface exchange measurements.Science260,1472–1480.

de Pury,D.G.G.,Ceulemans,R.,1997.Scaling-up carbon?uxes from leaves to stands in a patchy coniferous/deciduous forest.

In:Mohren,G.M.J.,Kramer,K.,Sabaté,S.(Eds.),Impacts of Global Change on Tree Physiology and Forest Ecosystems.

Kluwer Academic Publishers,Dordrecht,pp.263–272. Desjardins,R.L.,1985.Carbon dioxide budget of maize.Agric.

For.Meteorol.36,29–41.

Dolman,A.J.,Moors,E.J.,Elbers,J.A.,2001.The carbon uptake of a mid latitude pine forest growing on sandy soil.Agric.For.

Meteorol.111,157–170.

EC-UN/ECE,1996.European programme for the intensive monitoring of forest ecosystems.General information on the permanent observation plots in Europe(Level II).European Commission,DG VI F.II.2,in cooperation with ICP forests of UN/ECE.

Falge, E.,Baldocchi, D.D.,Olson,R.J.,Anthoni,P.,Aubinet, M.,Bernhofer, C.,Burba,G.,Ceulemans,R.,Dolman,H., Granier,A.,Gross,P.,Grünwald,T.,Hollinger,D.,Jensen,N.O., Katul,G.,Keronen,P.,Kowalski,A.S.,Ta Lai,C.,Law,B.E., Meyers,T.,Moncrieff,J.,Moors,E.,Munger,J.W.,Pilegaard, K.,Rannik,U.,Rebmann,C.,Suyker,A.,Tenhunen,J.,Tu, K.,Verma,S.,Vesala,T.,Wilson,K.,Wofsy,S.,2001.Gap ?lling strategies for defensible annual sums of net ecosystem exchange.Agric.For.Meteorol.107,43–69.

226 A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227

Falge,E.,Tenhunen,J.,Baldocchi,D.D.,Aubinet,M.,Bakwin, P.,Berbigier,P.,Bernhofer,C.,Bonnefond,J.-M.,Burba,G., Clement,R.,Davis,K.J.,Elbers,J.A.,Falk,M.,Goldstein,

A.H.,Grelle,A.,Granier,A.,Grunwald,T.,Gudmundsson,J.,

Hollinger,D.,Janssens,I.A.,Keronen,P.,Kowalski,A.S.,Katul,

G.,Law,B.E.,Malhi,Y.,Meyers,T.,Monson,R.K.,Moors,

E.,Munger,J.W.,Oechel,W.,Paw,U.K.T.,Pilegaard,K.,

Rannik,U.,Rebmann,C.,Suyker,A.,Thorgeirsson,H.,Tirone,

G.,Turnipseed,A.,Wilson,K.,Wofsy,S.,2002.Phase and

amplitude of ecosystem carbon release and uptake potentials as derived from FLUXNET measurements.Agric.For.Meteorol.

113,75–95.

Foken,T.,Wichura,B.,1996.Tools for quality assessment of surface-based?ux measurements.Agric.For.Meteorol.78,83–105.

Gond,V.,de Pury,D.G.,Eerens,H.,Veroustraete,F.,Ceulemans, R.,1999.Seasonal variation of leaf-area index,leaf chlorophyll and water content scaled up to fAPAR to estimate the carbon balance of a temperate multi-layer,multi-species forest.Tree Physiol.19,673–679.

Goulden,M.L.,Munger,J.W.,Fan,S.M.,Daube,B.C.,Wofsy, S.C.,1996.Measurements of carbon storage by long-term eddy correlation:methods and a critical assessment of accuracy.

Global Change Biol.2,169–182.

Greco,S.,Baldocchi,D.D.,1996.Seasonal variations of CO2and water vapor exchange rates over a temperate deciduous forest.

Global Change Biol.2,183–197.

Hollinger,D.Y.,Kelliher,F.M.,Byers,J.N.,Hunt,J.E.,McSeveny, T.M.,Weir,P.L.,1994.Carbon dioxide exchange between an undisturbed old-growth temperate forest and the atmosphere.

Ecology75,134–150.

Houghton,R.A.,Davidson,E.A.,Woodwell,G.M.,1998.Missing sinks,feedbacks,and understanding the role of terrestrial ecosystems in the global carbon balance.Global Biogeochem.

Cycles12,25–34.

Janssens,I.A.,Sampson, D.A.,Cermak,J.,Meiresonne,L., Riguzzi,F.,Overloop,S.,Ceulemans,R.,1999.Above-and below-ground phytomass and carbon storage in a Belgian Scots pine stand.Ann.For.Sci.56,81–90.

Janssens,I.A.,Meiresonne,L.,Ceulemans,R.,2000.Mean soil CO2ef?ux from a mixed forest:temporal and spatial integration.In:Ceulemans,R.,Veroustraete, F.,Gond,V., Van Rensbergen,J.(Eds.),Forest Ecosystem Modeling, Upscaling and Remote Sensing.SPB Academic Publishing,The Netherlands,pp.19–33.

Janssens,I.A.,Lankreijer,H.,Matteucci,G.,Kowalski, A.S., Buchmann,N.,Epron,D.,Pilegaard,K.,Kutsch,W.,Longdoz,

B.,Grunwald,T.,Dore,S.,Montagnani,L.,Rebmann,

C.,

Moors,E.J.,Grelle,A.,Rannik,U.,Morgenstern,K.,Oltchev, S.,Clement,R.,Gudmundsson,J.,Minerbi,S.,Berbigier,P., Ibrom,A.,Moncrieff,J.,Aubinet,M.,Bernhofer,C.,Jensen, N.O.,Vesala,T.,Granier, A.,Schulze, E.D.,Lindroth, A., Dolman,A.J.,Jarvis,P.G.,Ceulemans,R.,Valentini,R.,2001.

Productivity overshadows temperature in determining soil and ecosystem respiration across European forests.Global Change Biol.7,269–278.Kaimal,J.C.,Wyngaard,J.C.,Izumi,Y.,Cote,O.R.,1972.Spectral characteristics of surface layer turbulence.Q.J.R.Meteorol.

Soc.98,563–589.

Keeling,R.F.,Piper,S.C.,Heimann,M.,1996.Global and hemispheric CO2sinks deduced from changes in atmospheric O2concentration.Nature381,218–221.

Kowalski,A.S.,Overloop,S.,Ceulemans,R.,2000.Eddy?uxes above a Belgian,Campine forest and their relationship with predicting variables.In:Ceulemans,R.,Veroustraete,F.,Gond, V.,Van Rensbergen,J.(Eds.),Forest Ecosystem Modeling, Upscaling and Remote Sensing.SPB Academic Publishing,The Hague,pp.3–17.

Kristensen,L.,Mann,J.,Oncley,S.,Wyngaard,J.C.,1997.How close is close enough when measuring scalar?uxes with displaced sensors?J.Atm.Oceanic Technol.14,814–821. Lee,X.H.,1998.On meteorological observations of surface–air exchange above tall vegetation.Agric.For.Meteorol.91,39–

49.

Leuning,R.,Judd,M.J.,1996.The relative merits of open-and closed-path analysers for measurements of eddy?uxes.Global Change Biol.2,241–254.

Leuning,R.,King,K.M.,https://www.wendangku.net/doc/e59789453.html,parison of eddy-covariance measurements of CO2?uxes by open-and closed-path CO2 analysers.Boundary-Layer Meteorol.59,297–311. Lindroth, A.,Grelle, A.,Moren, A.-S.,1998.Long-term measurements of boreal forest carbon balance reveal large temperature sensitivity.Global Change Biol.4,443–450. Lloyd,J.,Taylor,J.A.,1994.On the temperature dependence of soil respiration.Funct.Ecol.8,315–323.

McMillen,R.T.,1988.An eddy correlation technique with extended applicability to non-simple terrain.Boundary-Layer Meteorol.43,231–245.

Meiresonne,L.,Sampson,D.A.,Kowalski,A.S.,Janssens,I.A., Nadezhdina,N.,Cermak,J.,Van Slycken,J.,Ceulemans,R., 2003.Water?ux estimates from a Belgian Scots pine stand:a comparison of different approaches.J.Hydrol.270,230–252. Michaelis,L.,Menten,M.L.,1913.Die Kinetik der Invertinwirkung.Biochemisches Zeitschrift49,333. Moncrieff,J.B.,Malhi,Y.,Leuning,R.,1996.The propagation of errors in long-term measurements of land–atmosphere?uxes of carbon and water.Global Change Biol.2,231–240. Moncrieff,J.B.,Massheder,J.M.,de Bruin,H.,Elbers,J.,Friborg, T.,Heusinkveld,B.,Kabat,P.,Scott,S.,Soegaard,H.,Verhoef,

A.,1997.A system to measure surface?uxes of momentum,

sensible heat,water vapor and carbon dioxide.J.Hydrol.

188–189,589–611.

Moore, C.,1986.Frequency response corrections for eddy correlation systems.Boundary-Layer Meteorol.37,17–35. Neirynck,J.,Van Ranst,E.,Roskams,P.,Lust,N.,2002.Impact of decreasing throughfall depositions on soil solution chemistry at coniferous monitoring sites in northern Belgium.For.Ecol.

Manage.160,127–142.

Overloop,S.,Meiresonne,L.,1999.Basiskarakteristieken van het proefvlak Brasschaat,domeinbos de Inslag.Mededelingen van het Institut voor Bosbouw en Wildbeheer1,11–21(in Dutch). Pilegaard,K.,Hummelshoj,P.,Jensen,N.O.,Chen,Z.,2001.Two years of continuous CO2eddy?ux measurements over a Danish beech forest.Agric.For.Meteorol.107,29–41.

A.Carrara et al./Agricultural and Forest Meteorology119(2003)209–227227

Reich,P.B.,Bolstad,P.,2001.Productivity of evergreen and deciduous temperate forests.In:Roy,J.,Saugier,B.,Mooney,

H.A.(Eds.),Terrestrial Global Productivity.Physiological

Ecology.Academic Press,San Diego,pp.245–283. Roskams,P.,Sioen,G.,Overloop,S.,1997.Meetnet voor de intensieve monitoring van het bosecosysteem in het Vlaamse Gewest-resultaten1991–1992.Institute for Forestry and Game Management,Ministry of the Flemish Community(in Dutch). Rotherham,I.D.,1983.The ecology of Rhododendron ponticum with special reference to its competitive and invasive capabilities.Ph.D.Thesis,University of Schef?eld,UK. Sakai,R.K.,Fitzjarrald,D.R.,Moore,K.E.,2001.Importance of low-frequency contributions to eddy?uxes observed over rough surfaces.J.Appl.Meteorol.40,2178–2192.

Sarmiento,J.L.,Gruber,N.,2002.Sinks for anthropogenic carbon.

Phys.Today208,30–36.

Saxén, B.,Saxén,H.,1995.NNDT—A Neural Network Development Tool,Version1.2.?bo Akademi University,?bo, Finland.

Schimel,D.S.,1995.Terrestrial ecosystems and the carbon cycle.

Global Change Biol.1,77–91.

Schmid,H.P.,Grimmond,C.S.B.,Cropley,F.,Offerle,B.,Su,

H.-B.,2000.Measurements of CO2and energy?uxes over a

mixed hardwood forest in the mid-western United States.Agric.

For.Meteorol.103,357–374.

Stull,R.B.,1988.An Introduction to Boundary Layer Meteorology.

Kluwer Academic Publishers,The Netherlands.

Suni,T.,Berninger, F.,Markkanen,T.,Keronen,P.,Rannik, U.,Vesala,T.,2003.Inter-annual variability and timing of

growing-season CO2exchange in a boreal forest.J.Geophys.

Res.108,No.D9-10.1029/2002JD002381.

Swinbank,W.C.,1951.The measurement of vertical transfer of heat and water vapor by eddies in the lower atmosphere.J.

Meteorol.8,135–145.

Valentini,R.,Matteucci,G.,Dolman, A.J.,Schulze, E.D., Rebmann, C.,Moors, E.J.,Granier, A.,Gross,P.,Jensen, N.O.,Pilegaard,K.,Lindroth, A.,Grelle, A.,Bernhofer,

C.,Grunwald,T.,Aubinet,M.,Ceulemans,R.,Kowalski,

A.S.,Vesala,T.,Rannik,U.,Berbigier,P.,Loustau, D.,

Gudmundsson,J.,Thorgeisrsson,H.,Montagnani,L.,Minerbi, S.,Jarvis,P.G.,2000.Respiration as the main determinant of carbon balance in European forests.Nature404,861–865.

Verma,S.B.,Baldocchi, D.D.,Anderson, D.E.,Matt, D.R., Clement,R.J.,1986.Eddy?uxes of CO2,water vapor and sensible heat over a deciduous forest.Boundary-Layer Meteorol.36,71–91.

Vickers,D.,Mahrt,L.,1997.Quality control and?ux sampling problems for tower and aircraft data.J.Atm.Oceanic Technol.

14,512–526.

Webb,E.K.,Pearman,G.I.,Leuning,R.,1980.Correction of?ux measurements for density effects due to heat and water vapor transfer.Q.J.R.Meteorol.Soc.106,85–100.

Xiao, C.W.,Curiel Yuste,J.,Janssens,I.A.,Roskams,P., Nachtergale,L.,Carrara,A.,Sanchez,B.Y.,Ceulemans,R., 2003.Above-and belowground biomass and net primary production in a73-year-old Scots pine forest.Tree Physiol.23, 505–516.

避免成为垃圾邮件

如何使邮件营销不被视为垃圾邮件 来源: 邮件营销发布时间: 2011-01-16 18:04 259 次浏览大小: 16px 14px 12px 如何使邮件营销不被视为垃圾邮件 如何使邮件不被拒收，邮件白名单[邮件营销] 关键词：如何群发邮件不被认定为垃圾邮件， 早些时候，我们单位也曾经使用过电子邮件群发进行【网络营销】之【电子邮件营销】。在这里将其中的一些问题和技巧分享给大家 一般的无论您使用的是免费邮箱还是收费的企业邮箱，一个邮箱帐户每天最多的发送量要求是【不能超过1000封】的, 强比科技和网易企业邮箱都是反对群发垃圾邮件的，但对于应用型的邮件和向会员发送邮件，就不在此列。 注：网易企业邮箱是防垃圾邮件协会的主要倡导者，对于发送垃圾邮件并没有特权，使用企业邮箱并不会让您的应用型邮件发的更多更广，但我们会告诉您一些规避被认定为垃圾邮件的规则 1、单邮件单发，收件人里不要一下填上两百个收件人帐号。。。保守的建议：收件人、抄送、密送栏，每栏都不要超过10个以上的收件帐户 2、您发送的收件人是否准确？如果您发送的收件人帐号有错误，请一定要注意将其从列表删除，如果服务器收到太多的退信，会对您的帐户进行监测或被认定为发送的是垃圾邮件 3、群发邮件主题及内容最好每次有所变化和注意相关的垃圾词汇。群发邮件时，一定要注意邮件主题和邮件内容，很多邮件服务器为过滤垃圾邮件设置了垃圾字词过滤，如果邮件主题和邮件内容中包含有如：大量、宣传、赚钱等字词（当然发票，枪支等更是不行的），服务器将会过滤掉该邮件，致使邮件不能发送。因此在书写邮件主题和内容时应尽量避开你认为的有垃圾字词嫌疑的文字和词语，才能顺利群发邮件。另外标题尽量不要太商业化，内容也不宜过多（尽量小于7k），如果一看就是推销邮件，效果就不会太好（有可能直接del了），而内容过多就会使阅读者不耐烦甚至根本不看。 4、会员广告邮件最好能在标题中标注（AD）这个网易等邮箱会直接发送到广告邮件夹里，

怎样避免邮件被当作垃圾邮件

电子邮件送达率是衡量电子邮件营销效果的重要指标之一。随着垃圾邮件越来越泛滥，世界上所有的 ISP 和服务器提供商都采取了越来越严厉的过滤垃圾邮件措施，同时也给正常邮件，以及合法合理、用户欢迎的电子邮件营销带来不便。不过这是大势所趋，不是营销人员能解决的。 垃圾邮件过滤方法垃圾邮件过滤方法 电子邮件营销人员能做的是尽量减少自己的邮件被当作垃圾邮件的机会。要做到这一点，首先需要了解主要的垃圾邮件过滤方法。 第一种是以触发式过滤算法鉴别垃圾邮件，这样的过滤器通常已经装在电子邮件客户端软件或邮件服务器上。其原理是过滤软件检查邮件的发信人，标题，正文内容，邮件中出现的链接和域名，甚至电话号码，当发现带有明显广告性质，或经常出现已知垃圾邮件的典型特征，则给这封邮件打一定的垃圾邮件特征分数。当分数达到一定数值时，邮件将被标志为垃圾邮件，直接过滤到垃圾邮件文件夹。 比如，邮件标题中出现￥、$符号，可能给予2 分垃圾分数。邮件内容中出现“免费”、“发票”、“促销”等典型垃圾邮件中经常出现的词汇时，也各给 1 分。邮件中如果包含已经被确认的经常发垃圾的域名，再加 1 分。甚至邮件内容中出现被确认与垃圾邮件相关联的电话号码，也给个分数。 当这些垃圾分数相加达到某一个数值时，比如达到 10 分，这个邮件将被标志为垃圾。 第二种方法是以黑名单为基础。有一些创建和维护链接邮件黑名单的组织，专门接受用户的垃圾邮件投诉，如果确认确实是垃圾邮件，黑名单运行者将把发送垃圾邮件的服务器和用户IP 地址放入黑名单。 比较有规模的垃圾黑名单通常都与其他ISP 及服务器运营商共享黑名单数据库。一旦某个IP 地址被列入黑名单，世界上很多ISP 和邮件服务器将拒收来自这个 IP 地址的所有邮件。 有的时候用户投诉其实并不是真的因为所收到邮件是垃圾邮件，而是用户忘记了曾经注册这个电子杂志。如果你的IP 地址被错误地投诉而列入黑名单，唯一的方法是联系黑名单维护组织，说明情况，提出证据，要求把你的IP 地址从黑名单中删除。不过这一过程有时非常复杂艰难。 第三种方法是邮件防火墙。很多大公司的服务器是运行在邮件防火墙之后，这些防火墙会综合使用各种过滤器以及黑名单，再加上自行研制的一些算法，来鉴别和剔除垃圾邮件。这些防火墙的算法则更复杂，并且不与其他人分享细节，对正常邮件的送达也可能起到致命的影响。 第四种方法是使用邮件确认。当电子邮件帐号收到一封email 时，这封 email 会首先进入待送达队列中排队，同时自动回复给发信人一封确认邮件。确认邮件中包含有一个确认链接，或标题中包含有一个独特的确认序列号，只有原来的发件人点击确认链接，或回复这封确认邮件，发信人的邮件地址才会被列入白名单，原来所发送的第一封原始邮件才真正被送达到收件箱。 鉴别和阻挡垃圾邮件大致上是这几种方法，有一些邮件服务器可能会综合使用这些方法。 为了避免邮件被这些过滤手段鉴别为垃圾邮件，应该注意下面一些问题。 检查服务器 IP 地址是否在黑名单中？选择邮件服务器时，应该检查服务器提供商的IP 地址是否被列在主要的垃圾黑名单中。国际上主要的垃圾黑名单包括： https://www.wendangku.net/doc/e59789453.html,

垃圾邮件防护系统分析与应用方法

垃圾邮件防护系统分析与应用方法 【内容提要】: 随着联机上网费用日趋便宜，发送电子邮件广告几近零成本又有利可图，因此造成垃圾邮件如今日混乱猖獗的现况。针对这种问题，许多公司研究出许多垃圾邮件防护和过滤机制产品，本文将对垃圾邮件的有关防护过滤技术和解决方法作一个基本介绍。 【关键词】：垃圾邮件、邮件防护、技术分析、AFS、华硕、过滤、机制 引言---------- 随着互联网的蓬勃发展，E-mail信息的传播达到了前所未有的广度和深度。同时不请自来的电子邮件也以各种形式闯入我们的邮箱- 商品推销、诈骗、政治或宗教抨击、病毒载体以及无法归类的稀奇古怪的形式。有些人每天甚至要收到100 到200 封这样的垃圾电子邮件（甚至更多）。因为更多的人开始使用英特网的关系（自因特网建立以来，人数飞速增长），对于商人、小贩、想入非非者以及蓄意破坏者而言，可以无偿地联系到数目巨大的各类人，诱惑力变得难以抵挡，自此大量的垃圾邮件在世界的各个角落产生，并瞬间传递到世界其他任何地方，这种费时且消耗CPU 的破坏行为迅速对经济产生了极大的负面影响。 现今越来越多的人开始意识到垃圾邮件的传递所带来的严重后果，并不断提出防治的新需求。 一垃圾邮件的定义 一封完整的电子邮件包含以下项目：邮件信封Mail Envelope、邮件标题Mail Header、邮件本文Mail Body 与邮件附檔Mail Attachment。电子邮件传输处理分为两阶段：邮件传输代理Mail Transfer Agent (简称MTA)，例如邮件服务器，以及与邮件使用代理Mail User Agent (简称MUA)，例如Outlook 或Outlook Express。 如果以邮件內容定义垃圾邮件，容易随个人主观认定而异；对银行业、娛乐业，广告业而言，包含其他银行贷款广告、色情广告的邮件，可能是种具有价值的市场资讯，而非垃圾邮件；因此，必需依邮件行为始能，依众人认知、法律规范与国际法规逐一精确定义何为垃圾邮件。 1. 众人认知：不请自來、来路不明、无法拒绝之邮件。 2. 法律规范：造成骚扰、匿名文书或嫁祸他人之邮件。 3. 国际法规： 2003 年底美国立法明定「Can Spam」垃圾邮件法规「Can Spam」字面表示可以「Spam」，惟有「但书」，寄件者必须表明身分，让收件者可以追溯来源不可以匿名、伪造，或者刻意隐匿或篡改资讯等行为发送电子邮件；发送方式方式不可为垃圾邮件滥发者(Spammer) 慣用之垃圾邮件滥发方式或程式，如借用邮件代替(Open Relay)、出现过多邮件转(Received) 或机器自动发送，以及不断尝试各种进入企业信箱方法等，必须提供收件者「选择权」，具有「取消订阅」机制。 综上所述，垃圾邮件之所以恼人并不是因为內容无趣不吸引人，而在于大量滥发，任意长驱直入收信者电子邮件信箱。 二邮件信息安全的影响

Lotus DominoNotes如何有效防止垃圾邮件

[精华] 【经验交流】Lotus Domino/Notes如何有效防止垃圾邮件. 为普及Lotus办公自动化平台(OA)而努力...... 1.前言 随着Internet技术的普及与发展，电子邮件已经成为大多数人必不可少的通讯工具。企业拥有自己的邮件系统，已经随处可见。然而，一个日益严重的问题却困扰着众多的系统管理员，那就是如何有效地防止—垃圾邮件。 垃圾邮件的破坏性是显而易见的，它不仅阻塞有限的网络带宽，而且降低了邮件服务器的信誉度，使得其它邮件服务器可能不接收此服务器发出的邮件。那么，如何有效地预防和限制垃圾邮件呢？在这一方面，Domino提供了各种配置参数，如何有效地使用这些参数，来全面制止垃圾邮件，是本文关心的话题。 下文以lxpServer/test为例，图文并茂进行阐述。 2.Domino提供的SMTP配置参数 2.1.服务器文档： 关于SMTP的参数配置有： ·邮件返回（进入服务器）时是否需要进行身份验证。一般的SMTP在返回邮件时是不需要身份认证的。 ·是否需要启用SSL端口。这主要是为了防止邮件在传输过程中被人截获而泄密。因为SMTP邮件在网络上传输时，采用的是明码的方式，对于敏感数据，可能需要启用SSL加密。 2.2.配置文档 关于SMTP的诸多限制，都是在配置文档中体现的，包括这几个方面的设置： ·Notes用户限制：允许Notes网络域中的哪些用户向外（Internet）发送邮件。 ·转发控制：邮件源地址控制参数，邮件目标地址控制参数，是否需要核实源地址身份。 ·外出控制：允许哪些（Internet）用户向外（Internet）发送邮件。 ·邮件返回控制：允许哪些用户发邮件给本服务器，是否核实对方的身份。 关于参数的详细说明，大家可以参考Domino管理帮助。下面我们通过一些实际的例子，来看看这些参数如何运用。

垃圾邮件的危害及应对措施

０　引言 垃圾邮件的泛滥，已经对互联网和现实社会造成了危害。对垃圾邮件的治理已成为与抗击网络病毒并列的重任。１　垃圾邮件的概况 所谓垃圾邮件，包括下述属性的电子邮件： 收件人事先没有提出要求或者同意接收的广告、电子刊物、各种形式的宣传品等宣传性质的电子邮件；收件人无法拒收的电子邮件；隐藏发件人身份、地址、标题等信息的电子邮件；含有虚假的信息源、发件人、路由等信息的电子邮件。 垃圾邮件的分类从内容上看，主要是商业广告性质的邮件；另外，有少量政治，团体组织的宣传邮件。从邮件的发送形式上看，有直接发送和第三方转发两种。 所谓直接发送，就是邮件的发送者使用自己的服务器，ＩＰ　地址，自己的网络资源传送这些邮件。目前使用更多的是使用第三方服务器转发。对于这种垃圾邮件，只要关闭有关服务器的转发功能就可以了。 垃圾邮件的危害。占用网络带宽，造成邮件服务器拥塞，进而降低整个网络的运行效率；侵犯收件人的隐私权，侵占收件人信箱空间，耗费收件人的时间、精力和金钱。有的垃圾邮件还盗用他人的电子邮件地址做发信地址，严重损害了他人的信誉；被黑客利用成助纣为虐的工具；严重影响ＩＳＰ的服务形象，　ＩＳＰ每年因垃圾邮件要失去７．２％的用户；妖言惑众，骗人钱财，传播色情等内容的垃圾邮件，已经对现实社会造成了危害。 ２　技术上的应对措施 ２．１　找出真正发件人 垃圾邮件一般采用了群发软件发送，发信人的地址是可以任意伪造的，查看信头可以让您找到真正的发件人。查看信头的方法是： （１）如果您是在Ｗｅｂ页面上看邮件的话，直接打开邮件，点击信件显示页面上方菜单中的“原文”，就可以看到信头。 （２）如果是用Ｏｕｔｌｏｏｋ　Ｅｘｐｒｅｓｓ来收信的话，指向邮件，不要打开，点击鼠标右键，看信件的属性，再点击详细资料，就可以看到信头。如果有ｓｅｎｄｅｒ的话，ｓｅｎｄｅｒ后面就是真正的发件人；如果没有ｓｅｎｄｅｒ，最后一个ｒｅ－ｃｅｉｖｅｄ　ｆｒｏｍ就是发件人所用的ＳＭＴＰ服务器。 Ｒｅｃｅｉｖｅ语句的基本表达格式是：ｆｒｏｍ　Ｓｅｒｖｅｒ　Ａ　ｂｙＳｅｒｖｅｒ　Ｂ，Ｓｅｒｖｅｒ　Ａ为发送服务器，Ｓｅｒｖｅｒ　Ｂ为接收服务器。 例如： ＲｅｔｕｒｎＰａｔｈ：＜ｏｗｎｅｒｅｎｅｗｓｄａｉｌｙ＠ｅｎｅｔ．ｃｏｍ．ｃｎ＞ Ｒｅｃｅｉｖｅｄ：ｆｒｏｍ　ｎｓ．ｅｎｅｔ．ｃｏｍ．ｃｎ　（［２０２．１０６．１２４．１６７］） ｂｙｍａｉｌ．７７７．ｎｅｔ．ｃｎ（８．９．３／８．８．７） ｗｉｔｈ　ＳＭＴＰ　ｉｄ　ＴＡＡ１３０４３； Ｔｈｕ，　２８　Ｏｃｔ　１９９９　１９：５１：２８　＋０８００ Ｒｅｃｅｉｖｅｄ：（ｆｒｏｍ　ｌｉｓｔ＠ｌｏｃａｌｈｏｓｔ） ｂｙ　ｎｓ．ｅｎｅｔ．ｃｏｍ．ｃｎ　（８．９．３／８．９．０）　ｉｄ　ＲＡＡ１９７１４ ｆｏｒ　ｅｎｅｗｓｄａｉｌｙｌｉｓｔ；Ｔｈｕ， ２８　Ｏｃｔ　１９９９　１７：５０：３０　＋０８００ Ｒｅｃｅｉｖｅｄ：ｆｒｏｍ　ｃｈｉｎａｎｅｔｗｅｅｋ．ｃｏｍ　（［２１０．７２．２３５．２１８］） ｂｙ　ｎｓ．ｅｎｅｔ．ｃｏｍ．ｃｎ　（８．９．３／８．９．０） ｗｉｔｈ　ＥＳＭＴＰ　ｉｄ　ＲＡＡ１９６９０ ｆｏｒ　＜ｅｎｅｗｓｄａｉｌｙ＠ｅｎｅｔ．ｃｏｍ．ｃｎ＞；　Ｔｈｕ， ２８　Ｏｃｔ　１９９９　１７：５０：２８　＋０８００ Ｒｅｃｅｉｖｅｄ：　ｆｒｏｍ　ｃｈｉｎａｎｅｔｗｅｅｋ．ｃｏｍ　（［１０．１．２．１０５］） ｂｙ　ｃｈｉｎａｎｅｔｗｅｅｋ．ｃｏｍ　（８．９．３／８．９．０） 作者简介：隆益民（１９６５－）　，男，副教授，硕士，研究方向为计算机及网络安全。 垃圾邮件的危害及应对措施 摘要：电子邮件是人们生活中加强沟通的重要交流工具，但垃圾邮件的泛滥已经使整个因特网不堪重负，垃圾邮件严重影响了人们使用互联网。治理垃圾邮件已成为与抗击网络病毒并列的重任。我国就垃圾邮件的处理上尚未制订相关的法律。也可参照国际通行的做法，针对垃圾邮件制订的较为严格的法律和规定，对垃圾邮件问题进行立法研究，从行政管理和法律角度加强对垃圾邮件的管理。 关键词：反垃圾邮件；群发；ＩＤ；电子邮票；黑名单；法律 隆益民 广东女子职业技术学院　计算机系 广东 ５１１４５０

如何使营销邮件不被当作垃圾邮件的几点技巧(精)

如何使营销邮件不被当作垃圾邮件的几点技巧 E-mail营销已经作为行之有效的方法被越来越多的企业营销人员所接受，但是不正当的邮件营销往往会适得其反，同时目前过滤垃圾邮件的软件、措施已越来越普及，如何用好这把行销的双刃剑，避免发出的营销电子邮件被错误地拒收。这里还是有一些技巧可供大家借鉴的： 1．注意标题 避免利用"收件人"、"抄送"和"暗送"将一封邮件发给大量的接收者。确信一封邮件每次仅发给了一个人。 2．选择适当主题 选择一个能够使人信任的主题。避免使用垃圾邮件常用的单词和符号，例如“免费”“派送”、“优惠”以及怪异字符标点等。 3．选用优秀的邮件发送软件 现在的邮件发送，邮件群发软件越来越多了，其中软件质量效果参差不齐，为了您企业的形象还是建议选用优秀的软件进行营销，同时建议使用注册版的软件，否则不少试用版的都在邮件下端留有版权信息，收件人一看就知道是群发，企业形象大打折扣，事与愿违。 4．响应电子邮件的要求。不仅要在电子邮件中含有"退订"的超级链接。订阅者可能会要求您将他的地址从您的数据库中删除。马上去做，并将处理的结果通知他。响应他们的要求有助于避免潜在的垃圾邮件抱怨，避免他们将您列入ISP的黑名单中。 5．加入白名单。诸如像AOL这样的一些国外ISP，允许用户将电子邮件地址加入到他们个人的白名单里。列在这个表里的电子邮件地址就可以通过ISP的垃圾邮件过滤器。浏览者在您的网站上订阅时，请他们将您的地址加入进他们的白名单中。 您也可以给在使用自己电子邮件程序（如微软的Outlook）的垃圾邮件过滤器的用户一个要求，请他们将您的邮件地址或域名添加进他们的"安全发件人"的列表。 6．监视您的退回邮件。当一个特殊的邮件多次被退回，您可以亲自联系接收人，并请他们将您的地址加入他们的白名单。如果您得不到他们的响应，就要从您的数据库中将这些电子邮件地址删除。如果您使用的是电子邮件管理软件，找出无效的邮件地址并删除它们。过多的退回邮件可能被垃圾邮件过滤器过滤。 也要记住，在您邮件"签名"文件中，含有隐私政策和退订。这样可以让您的浏览者不必担心由此引起其他的垃圾邮件，并随时退订不喜欢的邮件。这信息不仅能够增加您的订阅率，而

发送邮件如何避免被举报被判垃圾邮件的一些技巧

发送邮件如何避免被举报被判垃圾邮件的一些技巧 发送的邮件的主题及内容躲避欺诈信息、政治敏感信息、及该国家强烈禁止的相关关键词、字段、句子等。 纯广告的邮件最好少发，这是最不受运营商欢迎的，大多数情况下，你的客户不需要牛皮癣广告，倘若真要发的，请放慢速度，细水长流。主题别忘了加上“（AD）”。 避免被举报，避免在一段较短的时间内重复发送相同或者相关类型的邮件给相关的客户。一般如果有意向的话短时间客户就会有回复的，请不要因为没有回复重复发送而造成骚扰。 邮件主题的设计 原则一：电子邮件主题体现邮件内容的精华，要尽量的简短。 这样可以增加客户的信心，通过邮件主题来让他感觉到邮件内容的价值，迅速做出打开邮件详细阅读的决定。 原则二：电子邮件主题体现出发件人信息中无法包含的内容 发件人中除了显示的发件人名称和Email地址之外，很难容纳更为祥尽的信息，对发件人的信任还需要通过邮件主题来进一步强化，所以切中重点很有必要。 原则三：邮件主题含有丰富的关键词 除了增加客户的印象外，也为了让客户在检索收件箱中的邮件时增加发现的机会。 原则四：电子邮件主题不宜过于简单或过于复杂。 尽管没有严格的标准来限制邮件主题的字数，但保持一定合理的范围之内，既能反应比较重要的信息，又不至于在邮件主题栏默认的宽度内看不到有价值的信息。一般说来，电子邮件主题保持在8-20个汉字范围内是比较合适的。 原则五：要避免使用垃圾邮件常用的单词和符号。 例如“免费”、“派送”、“优惠”以及怪异字符标点等。 邮件内容的设计 1.第一次发的邮件最好不要有图片和附件。如果你和客户有联系，客户有回复你的邮件，邮件系统查实你们之间有邮件来往，这样以后的邮件有图片也问题不大。 2.信中最好不要有URL。有链接的话，也是很容易被拦截，因为邮件系统会认为有潜在威胁。如果是公司的网址最好是取消链接效果，把前面的http去掉，这样客户还是能够看到我们的网站。 3.不要用太奇怪的字体，最好用常规的字体。大家都知道用不同的字体，加上不同的颜色，可以吸引客户，但还是和图片一样，再好看，如果连给客户打开的机会都没有，也是白费功夫。奇怪的字体系统会认为是广告邮件。（虽然我们事实上是有广告性质，但是和那些开发

商务智能论文

主流商务智能解决方案的对比和分析 作者：彭潇勇软工一班2010302580155 摘要：针对市场上五种比较流行的商务智能解决方案供应商的产品进行了不同角度的分析与对比，指出了各种解决方案之间的共性和特性，并分析对比了各个产品之间的优劣势。以此作为国内企业寻找切合本企业实际利益的商务智能解决方案的一个依据。 1、引言： 商务智能，指用现代数据仓库技术、在线分析技术、数据挖掘和数据展现技术进行数据分析以实现商业价值。 商务智能（Business Intelligence，简称BI）最早由Gartner Group于1996年提出。Gartner将BI定义为“各种不同的应用程序和技术，可用于收集、存储、分析、共享数据并提供数据访问，从而帮助企业用户做出更好的业务决策。”商务智能的基本理念便是提升业务洞察力，将数据信息转化为商务价值。商业智能的关键是从许多来自不同的企业运作系统的数据中提取出有用的数据并进行清理，以保证数据的正确性，然后经过抽取（Extraction）、转换（Transformation）和装载（Load），即ETL过程，合并到一个企业级的数据仓库里，从而得到企业数据的一个全局视图，在此基础上利用合适的查询和分析工具、数据挖掘工具、OLAP工具等对其进行分析和处理（这时信息变为辅助决策的知识），最后将知识呈现给管理者，为管理者的决策过程提供支持。借助这一手段，企业可以在市场更加多变、竞争更加激烈、组织越来越复杂、规模越来越庞大的商业环境下高效运营、正确决策、快速响应，从而实现从数据到信息、从信息到知识、从知识到利润的转化。因此商务智能应该被看成是一种更好的为企业制定决策提供数据支持和信息知识的解决方案。 既然是解决方案型的技术，便不是一个高下立判的问题。因此我们便有必要对一些主流BI解决方案进行有针对性的分析，提取出其中值得企业关注和投资的信息，以便为企业选择BI解决方案提供一个有针对性的依据。一下便是有关IBM、Microsoft、Oracle、SAS、SAP 这五大BI解决方案提供商的对比分析。 2、五大主流BI解决方案提供商的对比： 2.1、IBM商务智能解决方案特性： IBM 公司在数据仓库/商业智能行业处于领先地位, 根据Survey. com 的2001 报告“数据仓库解决方案”, IBM 的市场占有率将近61% 。全世界跨各行各业的公司都选择了IBM的BI解决方案来帮助他们更具有方向性、更有利润的运营。 IBM具有如此庞大的市场占有率，这与其分布于世界各地的数千名BI咨询专家，以及由在BI方面具备提供解决方案资格的IBM业务合作伙伴组成的巨大网络分不开的。借用集成了各种不同类型的BI工具的DB2，IBM公司具备了相当丰富的经验和专长来开发经济高效、全面的解决方案，可以满足不同行业的具体需求和单个客户的需求。 IBM使用DB2 Warehouse Manager 构建数据仓库。有助于简化和加速构建数据仓库原型，以及后续BI系统的开发与配置。同时，用于构建和管理数据仓库架构的IBM 工具使企业能够构建和提供与Web 数据源相集成的数据仓库, 并且能够构建稳固、可扩展的数据中心和业务应用。 2.2、Microsoft商务智能解决方案： 进入2010年以后，微软的Office 2010、SharePoint2010、SQL Server 2008 R2等重头产品相继发布，从此，微软推出了一个新的组合型商业智能解决方案。 在这个解决方案中，微软强调商务智能的平民化，即企业的每一个员工都可以很容易的上手进行数据分析挖据并得到自己想要的数据。在底层有了SQL Server做保障，用户只要精

如何防止用户邮箱被盗号来发垃圾邮件

如何防止用户邮箱被盗号来发垃圾邮件 1)原因调查 帐户被盗用发垃圾邮件主要情况有以下几种： ①帐户密码设置过于简单，很多邮件用户由于疏忽，设置了一个简单的帐户密码，这种密码很容易被基于字典的破解技术破解。 ②用户使用Outlook ，foxmail 等邮件客户端收发邮件，如果用户的客户端没有安装杀毒 软件，一旦用户的操作系统中了盗取邮件帐户的木马病毒，邮件帐户密码就很容易被盗取。 ③用户在其他网站使用帐户加常用密码注册帐户，如果这些网站的资料泄露，就会导致邮件帐户密码的泄露。 以上几种情况是常见的帐户密码泄露情况，但是一旦帐户密码泄露，影响的不只是一个帐户，而是整个邮件服务器的正常运作。 2)如何预防 无论系统采用如何先进的防盗号技术，最重要的还是要加强用户的安全意识，从根本上杜绝此现象，例如采用复杂度高的密码，避免使用包含用户名或者纪念日之类的数字作为密码，定期修改邮箱密码等等措施。 TurboGate邮件网关系统在防盗号上，做了诸多预防措施和设置，从源头上防止垃圾邮件发送者的侵袭。 ① 防止SMTP盗号 用户邮箱被利用发送垃圾邮件，总是会依据一定的时间频率频繁发送差不多大小的垃圾邮件，根据垃圾邮件发送的特点，TurboGate邮件网关系统会自动扫描用户发邮件状况，在检测 到用户有可能频繁发送垃圾邮件时，能自动锁定账号，及时堵截垃圾邮件的发送，保证其他用户邮件的正常收发。 系统设置 系统管理员postmaster登录后台管理→系统设置→SMTP服务

系统管理员可以根据自己企业的实际情况，通过检测用户定期发送邮件的频率和邮件大小的情况，来判定用户账号是否已被利用来发送垃圾邮件。一旦判定用户已被利用发送大量垃圾邮件，该用户账号将被锁定。该用户需要向管理员申请解锁并修改邮箱密码。 ② 外发邮件使用反垃圾引擎 通过TurboGate邮件网关系统反垃圾引擎，判断系统用户外发的邮件是否为垃圾邮件。若判断为垃圾邮件，可通过邮件网关系统发送“外发垃圾邮件提示”给该用户，及时提醒用户邮箱的使用情况。另外，若外发垃圾邮件达到一定的次数，将自动锁定该用户邮箱账号，避免继续外发垃圾邮件。 系统设置 系统管理员postmaster登录后台管理→反垃圾/反病毒→反垃圾引擎设置 锁定例子 网关系统管理员postmaster登录后台管理→系统监控→SMTP盗号发送垃圾邮件帐号列表 一旦用户邮件账户被锁定，该用户将不能在邮件客户端如outlook、foxmail等使用该邮件账户发送邮件。此时邮件客户端会提示：

如何防止垃圾邮件

如何防止垃圾邮件 以下提供一些如何防止垃圾邮件的设置使用方法，以及一些功能的使用经验供用户参考。 1 “只允许本系统用户发信”以及“启用SMTP发信身份认证功能”是最强烈建议启用的，否则，一些邮局（比如：https://www.wendangku.net/doc/e59789453.html,）会因此拒收您邮局所发出的邮件。 也建议您启用“接收认证功能”，若禁用此功能时，系统内用户互发邮件时将不验证身份，这会使伪造发件人成为可能。 2 以管理员身份登录webmail，然后在“系统设置 | 重要设置”中启用“SMTP域名验证功能”。通常按下图设置即可： 在“未通过检查邮件的处理方式”中建议您可以设置为“拒收”，这样可以避免放入用户垃圾箱后，用户在使用客户端软件时不能及时处理的问题。 如果从其他邮局（如：https://www.wendangku.net/doc/e59789453.html,）发来邮件失败，且退信中有 your domain mismatches client IP [220.181.13.xxx] 这样的内容（注意：只有“拒收”模式才会有此信息），在确认此IP地址是可信任时，您可以将此IP或IP段（如：220.181.13.*）加入到“信任IP(不受限IP地址或IP段)功能” 中。通常情况下建议将 https://www.wendangku.net/doc/e59789453.html, 以及 https://www.wendangku.net/doc/e59789453.html, 的IP加入到此列表中： 220.181.13.* 220.181.31.* 注意：如果接收外部邮件是经由防火墙的25端口再转到您的邮件服务器上时（对邮件服务器来说，所有邮件均来自于防火墙的IP地址），请不要启用此项功能。否则，所有外部的邮件都会因为无法通过域名验证而被放入垃圾箱（或拒收）。 3

启用灰名单防垃圾邮件功能，可以有效减少垃圾邮件数量。但此功能会占用较多的内存。 4 开启RBL (Real-time Black List) 实时黑名单防垃圾邮件功能。方法：在服务器上点击右下角图标后，可以在“系统设置 | 收发规则”中找到此选项。 5 用户服务器对外部发送的邮件，接收人认为是垃圾邮件并向黑名单网站举报后，用户服务器IP地址就会被这些网站加入到黑名单中。采信并定期下载这些黑名单数据的邮局于是就会拒绝接收用户服务器发来的邮件。要想解决被列入黑名单的问题，就要严格控制对系统外部的邮件发送。 1) 病毒爆发引起的大量病毒邮件外发行为最容易造成服务器IP被列入黑名单的问题，所以您必须要 首先确保WinWebMail Server已经做好了正确的杀病毒关联设置，并已经通过了WinWebMail Server 防病毒测试。方法：在服务器上点击右下角图标后，在“系统设置 | 超期设置 | 启用邮件防病毒功能”后点击“设置”按钮，可以找到“保存设置并测试”按钮。 [详细的防病毒软件配置方法] 2) 邮件系统内帐号因为使用弱密码从而被人破解，并利用这些帐号对外发送垃圾邮件，这种情况也极 易引起服务器IP地址被列入黑名单。解决方法：要求用户更改为强密码。 设置方法：管理员登录WebMail后，在“用户管理”中点击用户帐号后的图标，就可以在其中设置要求用户更改强密码。也可以在“用户管理”中先“配置模板”后，再对所选用户全部“应用模板” 来实现。(需安装WinWebMail Server 3.8.2.1版本或更高版本) 3) 用管理员登录WebMail，在“系统设置 | 系统设置”内勾选“启用外发邮件排行统计功能”和“启 用邮件外发超限额自动监控功能”，这两项功能可以帮助管理员查出系统内的垃圾邮件发送人。在“系统设置 | [44] 邮件外发排行”中管理员需要每天查看有哪些用户外发邮件内容或外发邮件数量有不正常之处，然后将发送垃圾邮件的用户帐号禁用或是限制这些帐号允许外发的邮件数量，或是让这些用户更改强密码。 4) 开启“用户外发邮件自动限制功能”。方法：在服务器上点击右下角图标后，可以在“系统设置 | 防护”中找到此选项。管理员可以考虑设置为每小时只允许外发10封或20封邮件，而对于可信任的用户，管理员可以通过将其帐号加入到系统信任帐号中的方式来解除限制，方法：在服务器上点击右下角图标后，可以在“系统设置 | 端口”中找到“设置信任帐号”的选项。 5) 开启“防止利用系统内帐号进行中继发送”功能，并将处理方式设为“拒收”。方法：在服务器上 点击右下角图标后，可以在“系统设置 | 防护”中找到此选项。 6) SMTP发信认证功能必须被启用。方法：在服务器上点击右下角图标后，可以在“系统设置 | 收发 规则”中找到此选项。 7) 关闭“允许对发往本系统不存在帐号的外部邮址退信”功能。方法：在服务器上点击右下角图标后， 可以在“系统设置 | 超期设置”中找到此选项。 8) 关闭“发现病毒邮件时通知发件人”功能。方法：在服务器上点击右下角图标后，在“系统设置 | 超期设置 | 启用邮件防病毒功能”后点击“设置”按钮，可以找到此选项。

防范垃圾邮件的几点技巧(精)

防范垃圾邮件的几点技巧 如果你还在为收到很多不需要的垃圾邮件困扰的话，请瞄一瞄一下面这些垃圾邮件防范技巧，也许它们并不能完全地解决垃圾邮件问题，但我保证，你也不会再那么烦扰了。 1、安装合适的过滤软件 在我看来，好的邮件过滤器有两个特征。第一，容易安装和使用;第二，它有自动适配过滤性能，能够知道你认为哪些邮件是垃圾邮件、哪些是有用的。那些以某些关键字(例如“按揭”)为基础的过滤器，是很容易上当的，很可能将好的电子邮件归类为垃圾邮件，反之亦然，也可能将垃圾邮件归类为好的邮件。 2、故意混淆你的邮件地址 如果你运行一个网站的话，我强烈建议你在显示你的邮件地址的时候采取一些特别措施。垃圾邮件发送者找出“受害者”最常用的方法之一是使用特殊的软件来扫描万维网上的电子邮件地址，然后将这些地址存储到数据库，用垃圾邮件来进行轰炸。 如果你使用一些方法，将你的电子邮件地址显示出来，可以让人们很容易地联系到你，但垃圾邮件僵尸机却无法获得你的邮件地址的话，你就可以防止你自己被卷入到大片的电子邮件的数据库中。 有一些网站是专门处理这一问题的，举一个例子，蒂姆威廉姆斯(Tim Williams)的网站，它包含了许多简单的和复杂的方法，可以用来隐藏你的邮件地址。我最喜欢的是HTML编码方法，就目前的情况来说，它是足够的了——但如果你想要更高的安全性的话，你就需要使用Java技术了。 扯的远一些，你还可以用你自己的电子邮件地址做一个Google搜索，这样你就可以知道是否有网站将你的的电子邮件地址以非编码的形式显示出来，并且允许垃圾邮件发送者像采摘成熟的樱桃一样使用它。如果发生这样的事的话，你就需要联系网站的所有者，要求他删除你的邮件地址或将其进行编码。 3、关掉预览功能 对垃圾邮件发送者来说，每个电子邮件地址都是有价值的，如果发送到这些地址中的垃圾邮件被阅读的话，那价值就更高了。有些人使用HTML发送大量的电子邮件，并且在邮件中包含一些图片的链接，而这些链接通常是链到某个服务器上独有的地址，这个地址是垃圾邮件发送者所操控的，他们可以检测到哪些URL 请求查看图片。 简单来说就是，不管你是什么时候打开垃圾邮件的，你都是告诉了垃圾邮件发送者，他发送的废物已经达到了目的。为了防止那样的事发生，你要么就禁用HTML电子邮件，要么就关掉预览功能，并且只打开那些肯定不是垃圾邮件的邮件。

防止邮件变成垃圾邮件

防止邮件变成垃圾邮件注意的事项 1、标点符号不要热情过度！邮件中太多的惊叹号一定会被扣分，还有大量使用引号也会有相同问题。 2、邮件主题过长的邮件主题有时也会被扣分。语言和标点符号的问题很大程度上与此有关，所以邮件主题避免敏感词和惊叹号。 3、HTML质量HTML质量差的邮件会很容易被过滤器认定是垃圾邮件。一句话，越简单越好。尽量不要设置过多的文字格式、过多的颜色，整个内容花里胡哨，最后被邮件服务器一枪给毙掉了。所以尽量不要用过多加大、加粗的文字，如果你的邮件是英文内容，尽量不要全部用大写字母，这样才能增大进入收件箱的机会。 4、文字和图片数量发送的电子邮件包含太多的图片是造成邮件进垃圾箱最常见的原因。邮件的内容到底多长才合适呢？内容里最多能放多少图片呢？这个没有标准答案，只有参考答案：即越简单越好，图片越少越好。过长的内容，过多的图片，尤其是整个邮件内容里只有一样大图片，没有文字内容，被毙掉的可能性极大。所以一定要尽量控制邮件的内容大小，平衡文字与图片的布局与搭配。 5、链接数量邮件中太多链接也属于垃圾邮件特征。很多人喜欢在邮件内容里加入大量的链接，恨不得将整个网站的首页内容全部照搬上去。岂不知，邮件服务器不喜欢过多的链接，在邮件内容里放置链接的前，要充分考虑是否十分必要，只有必要的链接才放置进去，总之，链接能少则少，越少越不容易被过滤。 6、敏感词有些特定的词或短语会被垃圾邮件过滤器扣分。比如“买即送现金或％免费”，或大量使用“机会”或“点击此处”等字样，就算是许可的会员邮件也会遭遇进垃圾箱。 7、纯文本邮件内容如果是纯文本的，发送到收件箱的机会更大。 * 8、文本格式电子邮件文本格式不能太花俏。太多加大加粗字体，颜色鲜艳的文字都有可能被扣分。也要避免邮件字体太大和全部用大写字母拼写（英文情况下）。记住，从设计的角度来看，少即是多。

拦截垃圾邮件的6大技巧

拦截垃圾邮件的6大技巧 作为民航企业，电子邮件已经成为上海航空公司日常办公的支柱，垃圾邮件的困扰也随之而来。公司平均每天收到邮件2.3万封，其中垃圾邮件达到1.6万封，比例占邮件总数的近69%。用户经常抱怨每天处 理的垃圾邮件比正常邮件还多，浪费了大量时间和精力。通过在电子邮件系统中建立垃圾邮件综合防范体系，设置多达六道过滤拦截措施，我们成功地把垃圾邮件过滤掉96%以上。并且垃圾邮件的错误识别率也低于千分之一。 ● 反向DNS解析（PTR） 使用反向 DNS 解析（PTR）机制，可以基本确认邮件的来源是否合法。例如新浪通过https://www.wendangku.net/doc/e59789453.html,（对应IP 202.108.3.172）发送邮件，我们的邮件服务器在接受新浪发来的邮件过程中，去进行反向DNS解析，查询 202.108.3.172该IP是否有对应的固定域名，如果查询成功说明该IP确实是互联网上一个具有固定IP地址的发信服务器。现在国际上的反垃圾邮件组织普遍要求发信邮件服务器需要设置PTR记录，以此来确认这台邮件服务器是否是在Internet上稳定存在的服务器。大家应该为自己的发信邮件服务器设置好PTR 记录，以免发生被国外拒收邮件的情况。如果你从ISP运营商处拿到的不是一个完整的C类网段，你就需要向ISP提交添加PTR记录的申请。一般ISP可以在二周内完成。如果你的邮件系统是租用的，则你需要向出租方提出申请。 由于国内绝大多数中小型公司的邮件系统都没有设置PTR，所以，在邮件系统中应该把反向DNS解析结合其他措施进行拦截，而不应该直接拦截。一般可以把PTR用于判断该邮件是否是正常的邮件，用于减少垃圾邮件的误判率。 ● 黑名单、白名单、实时黑名单（RBL） 黑名单技术是最早出现的一种反垃圾邮件技术，一般的邮件服务器都有该功能。黑名单技术的原理是确定已知垃圾邮件制造者及其ISP的域名或IP地址，然后将其整理成黑名单。邮件服务器在接收邮件时，拒绝任何来自黑名单上的垃圾邮件制造者的邮件。黑名单服务是基于用户投诉和采样积累而建立的、由域名或IP组成的数据库。实时黑名单是指网上提供的即时更新的黑名单数据库，其具有很强的时效性。目前国际上提供免费第三方实时黑名单的组织有不少，根据本人多年的使用体验，推荐下列组织Spamhaus、Spamcop、Dsbl、Sorb，其中Spamhaus和Spamcop的黑名单数据库更庞大，更新更及时。国内的中国反垃圾邮件联盟也提供类似的服务。这些数据库保存了频繁发送垃圾邮件的主机名字或IP地址，供邮件服务器中的邮件传输代理（MTA）进行实时查询，以决定是否拒收相应的邮件。根据四年来的使用经验，靠RBL实时黑名单可以过滤拦截60%以上的垃圾邮件。 黑名单技术可能会拒绝掉来自某站点的正常邮件，从而造成邮件不能正常的投递。例如，国内的大型邮件系统运营商新浪、163、搜狐等发信服务器的IP 地址经常会进入国外的黑名单，甚至Hotmail、Yahoo的发信服务器IP地址也会进入黑名单，这时就需要白名单，我们把国内外大多数大型邮件系统的IP地址

如何避免邮件被过滤

为了避免邮件被这些过滤手段鉴别为垃圾邮件，应该注意下面一些问题。 1.检查服务器IP地址是否在黑名单中。 选择邮件服务器时，应该检查服务器提供商的IP地址是否被列在主要的垃圾黑名单中。 用户可以在网上实时查询自己的服务器IP地址是否被列入黑名单。当然在使用过程中也不能排除某些用户发送垃圾邮件影响到其他用户。如果发现邮件送达率、阅读率有异常降低，应该随时监控IP地址在主要黑名单的情况。 2.邮件撰写的注意点 (1)在邮件标题及正文中都尽量少使用敏感的、典型垃圾邮件常使用的词汇，如英文的伟哥、贷款、色情图片、获奖、赢取，以及中文的免费、促销、发票、礼物、避税、研修班、折扣、财务等。不是说这些词本身有什么问题，也不是完全不能用，而是尽量少用，以免触发垃圾过滤算法。 (2)少使用惊叹号，减少使用夸张的颜色，尤其是加粗的红色字体。这都是典型的垃圾邮件常用的吸引眼球的方法。如果是英文邮件，不要把很多词完全用大写。 (3)邮件内容、标题、发件人姓名都不要使用明显虚构的字符串。比如有的垃圾邮件发送者当然不会告诉别人真名实姓，就在发信人名称中随便写上几个字母。维护垃圾过滤算法的人也不傻，这种莫名其妙的随机字符串通常都是欲盖弥彰的垃圾邮件特征。 (4)HTML邮件代码应该简洁，减少使用图片。虽然HTML邮件允许使用图片美化邮件，但是图片与文字相比应该保持在最低比例。图片越多，被打的垃圾分数可能越高。 3.注册流程的注意点 (1)用户提交注册表格后显示的感谢页面及确认邮件中应该提醒用户把你的域名，以及邮件地址加入到用户自己的白名单和通讯录中。邮件客户端软件通常都在垃圾过滤器设置中有白名单选项，绝大部分免费邮件提供商，如雅虎、hotmail、gmail也都有相应的设置。把电子邮件地址存入通讯录中也起到相同的效果。 (2)如果某封邮件已经被过滤到垃圾邮件夹中，提醒用户单击“不是垃圾”按钮，告诉过滤器判断错误了，这些反馈信息会被邮件服务器的过滤算法所统计和运用在今后的算法中。 (3)给用户最简单方便的退订方法。在发给用户的所有邮件中都应该包含退订链接，用户单击这个链接，程序就会自动将其E-mail地址从数据库中删除。这个退订方法越简单越好，如果搞得很复杂，用户可能宁可去按更简单的“报告垃圾”按钮，造成的损失更大。 (4)及时处理投诉。如果收到用户或ISP的投诉，必须尽快处理。如果是用户忘记自己曾经订阅你的电子杂志，错误投诉，应该把完整证据，包括用户的姓名、电子邮件地址、订阅时的IP地址、精确订阅时间，提供给ISP和垃圾黑名单运营组织。在绝大多数情况下，只要提供确实证据，ISP和垃圾黑名单组织都会理解。

Microsoft Dynamics 365智能商务应用解决方案

LOGO 单位名称 Microsoft Dynamics 365智能商务应用解决 方案 在此输入你的单位名称

Microsoft Dynamics 365智能商务应用解决 方案 首先是关于我们微软人工智能 (AI) 平台的声明。微软人工智能 (AI) 平台是一个开放和灵活的平台, 开发者可以倚重它，将人工智能 (AI) 注入到他们的应用和服务中。云 + 人工智能 (AI) + 数据的组合, 使用正确的工具，您就可以很容易地通过我们的人工智能 (AI) 应用服务和基础设施，把人工智能 (AI) 融入到您的业务应用中去。这包括：认知服务(Cognitive Service)、聊天机器人框架 (Bot Framework) 和Azure 机器学习 (Machine Learning)。 今天已经有超过68.2万开发者在使用这个平台。这其中包括：优步(Uber) 和顶尖在线演示服务商（Brainshark）正在使用认知服务, 而罗克韦尔自动化 (Rockwell Automation) 采用聊天机器人框架 (Bot Framework) 和深度分析 (Deep Analytics)为他们的运营提效。我们还宣布了对Azure 机器学习 (Machine Learning) 相关服务和工具的更新, 为开发者和数据科学家提供了一个单一的解决方案。您可以通过它，在云中、本地或Windows 10浏览器 (Microsoft Edge) 上，创建、部署、管理和监控您的数据及学习模型。 其次是分享了我们如何继续将人工智能 (AI) 注入到现有的解决方案中去。这项工作是基于我们在人工智能 (AI) 核心研究方面的持续投资之上的，主要创新包括：计算机视觉 (Computer Vision), 自然语言处理(Natural Language Processing), 语音识别 (Speech Recognition), 等等。已注入人工智能 (AI) 的产品和服务功能包括: 具有智能图像分析或智能字幕的Office、具有生物特征认证的Windows、必应Bing、微软小娜(Cortana)，和商务应用平台Dynamics 365。 您可以看看我最近在瑞典斯德哥尔摩 (Stockholm) 关于PowerPoint 实时翻译插件的演示。为了使工作环境更有成效, 我们推出了Bing的商务专用版 (Bing for Business)，专为企业提供的智能搜索引擎。它提供了

发开发信,避免成为垃圾邮件的一些方法

发开发信，避免成为垃圾邮件的一些方法 最近一直在修改自己的开发信，经过多次测试，我发现只要信中有图片，就算不是在附件中，都有几乎90%的机会成为垃圾邮件，而且有图片之后，邮件会变大，更难避开拦截。有图片效果，邮件看起来会好很多，但是接近客户的成功率也会大大下降。同时也测试其他情况，和大家分享一下： 1.第一次发的邮件最好不要有图片和附件。如果你和客户有联系，客户有回复 你的邮件，邮件系统查实你们之间有邮件来往，这样以后的邮件有图片也问题不大。 2.信中最好不要有URL。有链接的话，也是很容易被拦截，因为邮件系统会认为有潜 在威胁。如果是公司的网址最好是取消链接效果，把前面的http去掉，这样客户还是能够看到我们的网站。 3.不要用太奇怪的字体，最好用常规的字体。大家都知道用不同的字体，加上 不同的颜色，可以吸引客户，但还是和图片一样，再好看，如果连给客户打开的机会都没有，也是白费功夫。奇怪的字体系统会认为是广告邮件。（虽然我们事实上是有广告性质，但是和那些开发票的相比，我们还算比较好） 4.内容不要有free和low price等字样。这样是很吸引人，但这也是系统拦截 的最普遍的字眼。一般的介绍内容就行了，有时太敏感的文字，系统一样会拦截。 5.标题是最关键的。比较另类的标题可以吸引人，不过也是很危险的，加大了拦截的可 能性。建议用一些朴实的标题，正式一点的，避免标题和内容不一致，客户看到，会很反感的。 6.一定要针对性地发邮件，向那些和自己产品相关的客户发邮件，并 且在发之前，要大概了解客户的情况。一天发几十到一百封左右的针对性客户邮件。在了解客户的同时，你也从客户相关的咨询学到好多东西，这也是一种进步，使自己更专业。 7.邮件要有合理的排版，不要太凌乱。 8.如果条件允许，用付费的企业邮箱最好了，收发邮件稳定、高效；实在没有的话，其次考虑国外的免费邮箱gmail，yahoo也可以。 这些都是减少成为垃圾邮件的办法。邮件的吸引性可能会因此下降，但是我们最重要的目的是，要避免成为垃圾邮件，并且客户愿意打开，其次才是从内容吸引客户。客户一天不打开你的邮件，你写的再好也没有用。我认为这个方向比较正确，希望对大家有用。