The variability of biogenic sulfur flux from a temperate salt marsh on short time and space scales

Three emission chambers were deployed simultaneously to measure rates of emission of dimethyl sulfide, methane thiol and carbonyl sulfide within or across vegetation zones in a New Hampshire salt marsh. Short term (a few hours) variation in fluxes of all S gases from replicate sites were small within a monospecific stand of either Spartina alterniflora or S. patens. The quantity of emergent biomass and the type of vegetation present were the primary factors regulating the rate of emission. Dimethyl sulfide fluxes from the S. alterniflora soils ranged from 800 to 18,000 nmol m⁻² h⁻¹ compared to emissions of 25–120 nmol m⁻² h⁻¹ from S. patens. This difference was probably due to the presence of the dimethyl-sulfide precursor dimethylsulfoniopropionate which is an osmoregulator in S. alterniflora but not in S. patens. Methane thiol emissions from S. alterniflora were 20–280 nmol m⁻² h⁻¹ and they displayed a similar diel trend as dimethyl sulfide, although at much lower rates, suggesting that methane thiol is produced primarily by leaves. Methane thiol emissions from S. patens were 20–70 nmol m⁻² h⁻¹. Net uptake of carbonyl sulfide of 25–40 nmol m⁻² h⁻¹ occurred in stands of S. alterniflora while net efflux of 10–36 nmol m⁻² h⁻¹ of carbonyl sulfide occurred in stands of S. patens. In general, ranges of emissions of sulfur gases were similar to most other published values.     

Formic acid and acetic acid: Emissions,atmospheric formation and dry deposition at two southern California locations

Emission rates, in situ formation rates and removal rates by dry deposition are estimated for formic acid (HCOOH, C1) and acetic acid (CH₃COOH, C2), which are the most abundant acids in southern California air and together account for much of the airborne acidity and are the leading contributors to acid dry deposition. Using data for eight unreactive tracers, direct emission rates during the fall 1987 are estimated to be 5.6 and 12.8 metric tons d⁻¹ for C1 and C2, respectively, at a coastal source-dominated site. These emissions rates increase to 9.6(C1) and 20.4(C2) metric tons d⁻¹ during the summer. In situ formation in the atmosphere via the ozone-olefin reaction is an important source for both acids. This reaction produces an estimated 25.0 and 10.1 metric tons d⁻¹ of C1 and C2, respectively, during the day and 34.5 (C1) and 4.3 (C2) metric tons d⁻¹ at night. More acetic acid than formic acid is emitted by direct sources, with C2/C1 emission rate ratios of 2.1–2.3. The reverse is true of in situ formation, with C1/C2 production rate ratios of 2.5 (day) and 8.0 (night). Dry deposition removal rates depend on season (fall > summer) and location (inland > coastal) and are 22–52 metric tons d⁻¹ for C1, and 32–83 metric tons d⁻¹ for C2. Source (emissions + in situ formation) and sink (dry deposition) terms are of the same magnitude in all six cases studied and balance each other well in three of these cases. Uncertainties in emission, in situ production and removal rates are discussed and reflect uncertaintes in olefin and unreactive tracer emission rates, yields of organic acids from the Criegee biradical (ozone-olefin reaction), and dry deposition velocity, respectively.     

Significance of air humidity and air velocity for fungal spore release into the air

Our previous field studies have shown that the presence of molds in buildings does not necessarily mean elevated airborne spore counts. Therefore, we investigated the release of fungal spores from cultures of Aspergillus fumigatus, Penicillium sp. and Cladosporium sp. at different air velocities and air humidities. Spores of A. fumigatus and Penicillium sp. were released from conidiophores already at air velocity of 0.5 ms⁻¹, whereas Cladosporium spores required at least a velocity of 1.0 ms⁻¹. Airborne spore counts of A. fumigatus and Penicillium sp. were usually higher in dry than moist air, being minimal at relative humidities (r.h.) above 70%, while the effect of r.h. on the release of Cladosporium sp. was ambivalent. The geometric mean diameter of released spores increased when the r.h. exceeded a certain level which depends on fungal genus. Thus, spores of all three fungi were hygroscopic but the hygroscopicity of various spores appeared at different r.h.-ranges. This study indicates that spore release is controlled by external factors and depends on fungal genus which can be one reason for considerable variation of airborne spore counts in buildings with mold problems.     

Real-time,in situ measurements of atmospheric optical absorption in the visible via photoacoustic spectroscopy—IV. Visibility degradation and aerosol optical properties in Los Angeles

Aerosol light absorption (b_abs) has been measured in real-time in Los Angeles with a validated photoacoustic technique, and its impact on visibility degradation has been examined. These measurements were collected during ten days in the summer of 1987 for the Southern California Air Quality Study (SCAQS). Aerosol b_abs (λ = 514.5 nm) varied from an hourly average value of 7 × 10⁻⁶ m⁻¹ in the 3–4 and 4–5 a.m. periods of 13 July to 9 × 10⁻⁵ m⁻¹ in the 7–8 a.m. period of both 28 August and 3 September. This b_abs, which is due solely to elemental carbon (EC) showed a distinct diurnal pattern with low values at night, increasing around sunrise to higher values through mid-afternoon. Comparison of these data with aerosol light scattering data clearly illustrates that the contribution of aerosol light absorption to visibility degradation increases in importance under less polluted conditions. Other urban and rural studies show similar results.     

Kinetics of evaporation of ammonium chloride and ammonium nitrate aerosols

The evaporation rates of ammonium chloride and ammonium nitrate were measured by continuously and rapidly removing gaseous NH₃ and HNO₃ or HCl from aerosols in an annular denuder. The experiments gave the evaporation rates in terms of mass loss of chloride or nitrate which can be expressed conveniently as the rates of reduction of aerosol radius with time. Both dry aerosols (humidity 30–60% r.h.) and aqueous aerosols (humidity ca 97% r.h.) were studied. Dry aerosols evaporate at rates of −1.05 Å s⁻¹ for NH₄Cl and −0.45Å s⁻¹ for NH₄NO₃, while the evaporation rates of aqueous aerosols expressed as for equivalent dry particles are −4.52 Å s⁻¹ for NH₄Cl and −0.49 Å s⁻¹ for NH₄NO₃. The experimentally measured rates are independent of particle radius and remarkably low compared with those predicted from existing theories of aerosol evaporation, thus implying that there is an unknown kinetic constraint to the achievement of equilibrium at atmospheric temperature and pressures.     

Contamination of aqueous samples with formate and acetate from ambient air

A sensitive ion chromatographic technique with detection limits of 1.9 and 6.2 μgℓ⁻¹ has been developed for the determination of formate and acetate ions, respectively, in aqueous solution. Using this technique, uncovered aqueous solutions have been found to absorb the corresponding acids readily from ambient air at rates of approximately 0.02–0.1 nM cm⁻² h⁻¹. Consequently, to prevent vapor diffusion and subsequent contamination of environmental samples with these organic acids, casual exposure to ambient air, particularly in a laboratory, should be minimized.     

Laboratory experiments on membrane filter sampling of airborne mycotoxins produced by Stachybotrys atra corda

A membrane filter method for sampling of airborne stachybotrystoxins was studied in the laboratory. Toxigenic strains of Stachybotrys atra on wallpaper, grain, hay and straw were used as toxin sources in the experiments. Air samples were collected on cellulose nitrate and polycarbonate membrane filters at air flow rates of 10–20 ℓ min⁻¹. After the filter sampling, the air was passed through methanol. The results showed that stachybotrystoxins (trichothecenes) were concentrated in airborne fungal propagules, and thus can be collected on filters. Polycarbonate filters with a pore size of 0.2 μm collected the highest percentage of toxic samples. The laboratory experiments indicated that polycarbonate filter sampling for the collection of airborne mycotoxins is a promising method for extension to field measurements.     

Characterization of atrazine degradation and nitrate reduction by Pseudomonas sp. strain ADP

Concomitant atrazine degradation and nitrate reduction by a pure culture of Pseudomonas sp. strain ADP were studied. Under anoxic conditions, Ps ADP grew well and degraded atrazine efficiently in the presence of nitrate. Similar atrazine degradation rates were observed under both anoxic and aerobic conditions: 30.7±2.83 and 36.2±0.44 mg atrazine g⁻¹ cell h⁻¹, respectively. A high denitrification rate of 90.8±8.22 mg NO₃⁻-N g⁻¹ cell h⁻¹ was also observed using Ps ADP with citrate as the electron donor. The required citrate to nitrate ratio for complete denitrification was 5.11±0.15 g citrate g⁻¹ NO₃⁻-N.     

Laboratory tests of KI and alkaline annular denuders

Annular denuders coated with KI and with alkaline solutions have been tested for their ability to remove atmospheric pollutants including ozone, NO₂, SO₂, formaldehyde, methyl nitrate and peroxyacetyl nitrate. Tests were carried out at flow rates of 0.4–2.0 ℓ min⁻¹, using particle-free ambient air as well as purified air to study the effect of atmospheric CO₂ on alkaline denuder performance. Denuders coated with KI were efficient in removing O₃, NO₂, SO₂ (> 95%) and PAN (84±3%) but not methyl nitrate (44%) and formaldehyde (<5%). Selective removal of PAN from NO₂, and vice versa, could be obtained with annular denuders coated with NaOH, which removed 100% PAN and ⩽15% NO₂, and with alkaline guaiacol, which removed ⩾99% NO₂ and ⩽6% PAN.     

Environmental chamber for study of the deposition flux of gaseous pollutants to material surfaces

A specially designed recirculating environmental chamber was constructed to study the environmental factors affecting the deposition of pollutant gases to the surface of stone and other building materials. The chamber and sample holder are designed to place samples in an aerodynamically well-defined air flow. The system is designed to permit use of radioactive ³⁵SO₂ as a tracer if necessary. A wide range of typical environmental conditions can be continuously maintained in the chamber. Wind speeds in the test section can range up to about 5 ms⁻¹, exposing replicate samples to air flow that is uniform to within approximately 3%. Relative humidity in the chamber can be maintained to within 3%, and SO₂, NO₂ and O₃ concentrations in the chamber air can be maintained to within 4%. Test results indicate SO₂ deposition and wind speed in the chamber are closely correlated, allowing for a direct determination of the surface resistance (r_c) component of the SO₂ deposition velocity to various test materials. Initial studies of SO₂ deposition to limestone and marble indicate the r_c values are approximately 1.3 s cm⁻¹ for fresh limestone and 34 s cm⁻¹ for fresh marble at 75% relative humidity, 26°C and 50 ppb SO₂.     

Precision of the radiochemical OH measurement method

Twenty-eight radiochemical ¹⁴C tracer measurement of tropospheric hydroxyl radical (OH) concentrations were obtained at a rural site near Washington State University, Pullman, WA (117°W, 47°N). Diurnal OH concentration variations were observed for the five days between 9 August and 14 August 1990. These data made it possible to estimate the midday precision and detection limit of the radiochemical OH measurement method. Experiments performed at a peak O₃ photolysis rate J (O(¹D)) of (3.0±0.2) × 10⁻⁵ s⁻¹ yielded a mean midday OH concentration of (5.6±0.1 (1σ)) × 10⁶ cm⁻³. Other data put an upper bound of 16% on fluctuations of instrument sensitivity. Low-light or nighttime background OH concentrations were less than (2.6±2) × 10⁵ cm⁻³. A lower detection limit of 10⁵ cm⁻³ was obtained when extra care was taken with the low-level ¹⁴C counting procedure.     

Laboratory studies of some halogenated ethanes and ethers: Measurements of rates of reaction with OH and of infrared absorption cross-sections

We have measured, using a conventional discharge-flow resonance-fluorescence technique, the rates of reaction between the hydroxyl radical and a series of halogenated ethanes and ethers for the temperature range 230–423 K. Our measurements gave the following Arrhenius expressions (units are cm³ molecule⁻¹ s⁻¹): CF₂HCH₃ (HFC-152), 14.2 × 10⁻¹³ exp-(1050/T); CF₂ClCH₃ (HCFC-142b), 2.6 × 10⁻¹³ exp-(1230/T); CFCl₂CH₃ (HCFC-141b), 5.8 × 10⁻¹³ exp-(1100/T); CF₃CFH₂ (HFC-134a), 5.8 × 10⁻¹³ exp-(1350/T); CF₃CF₂H (HFC-125), 2.8 × 10⁻¹³ exp-(1350/T); CF₃CCl₂H (HCFC-123), 11.8 × 10⁻¹³ exp-(900/T); CF₂HOCF₂CFClH, (enflurane), 6.1 × 10⁻¹³ exp-(1080/T); CFH₂OCH(CF₃)₂, (sevoflurane), 15.3 × 10⁻¹³ exp-(900/T). In two cases, we measured rate constants only at room temperature: CF₃CClBrH (halothane), 6 × 10⁻¹⁴ and CF₂HOCClHCF₃ (isoflurane), 2.1 × 10⁻¹⁴.We also report the following values for the integrated absorption cross-sections of the compounds in the spectral region 800–1200 cm⁻¹ in units of cm⁻² atm⁻¹: CF₂HCH₃, 1155; CF₂ClCH₃, 1422; CFCl₂CH₃, 1995; CF₃CFH₂, 2686; CF₃CF₂H, 1970, CF₃CCl₂H, 1411; CF₃CClBrH, 1400; CF₂HOCF₂CFClH, 4800; CF₂HOCClHCF₃, 3900; CFH₂OCH(CF₃)₂, 2550. We use our measurements to calculate ozone depletion potentials and greenhouse warming potentials relative to CFCl₃ for each compound.     

The vertical distribution of aerosols and acid related compounds in air and cloudwater

Vertical profiles (surface to 5 km) of aerosol particle number concentration, NO_y′ mixing ratio, and cloudwater SO₄²⁻ and NO₃⁻ equivalent concentration were obtained in three field studies: North Bay, Ontario, during the summer of 1982 and the winter of 1983–1984, and Syracuse, New York, during the fall of 1984. The measurements from these locations and different seasons are compared. Generally, airborne concentrations are highest with air-mass back trajectories from the south and lowest with back trajectories from the north. For the southerly trajectories, median particle number concentrations (0.2–2 μm) near ground level (950 mb) vary from 1700 cm⁻³ during the summer project to 800 cm⁻³ during the winter project. At 700 mb, the south trajectory particle number concentration ranged between 60 and 170 cm⁻³. Median NO_y′ mixing ratios for southerly back trajectories were approximately 6 and 9 ppb at 950 mb and 0.4 and 0.8 ppb at 700 mb for the fall and winter projects, respectively. Comparison of particle number concentration profiles outside of cloud with cloud droplet plus interstitial aerosol particle number concentrations inside cloud indicate that cumulus clouds can transport aerosols vertically from below cloud base. In contrast, stratiform clouds have similar concentrations inside the clouds as outside at the same altitude. The vertical variations of cloudwater sulphate and nitrate concentrations and the NO₃⁻/SO₄²⁻ equivalent concentration ratio are discussed for each of the three field studies.     

A comparison of two cloudwater/fogwater collectors: The rotating arm collector and the caltech active strand cloudwater collector

A side-by-side comparison of the Rotating Arm Collector (RAC) and the Caltech Active Strand Cloudwater Collector (CASCC) was conducted at an elevated coastal site near the eastern end of the Santa Barbara Channel in southern California. The CASCC was observed to collect cloudwater at rates of up to 8.5 ml min⁻¹. The ratio of cloudwater collection rates was found to be close to the theoretical prediction of 4.2:1 (CASCC:RAC) over a wide range of liquid water contents (LWC). At low LWC, however, this ratio climbed rapidly, possibly reflecting a predominance of small droplets under these conditions, coupled with a greater collection efficiency of small droplets by the CASCC. Cloudwater samples collected by the RAC had significantly higher concentrations of Na⁺, Ca²⁺, Mg²⁺ and Cl⁻ than those collected by the CASCC. These higher concentrations may be due to differences in the chemical composition of large vs small droplets. No significant differences were observed in concentrations of NO₃⁻, SO₄²⁻ or NH₄⁺ in samples collected by the two instruments.     

Life-cycle environmental and economic impacts of energy-crop fuel-chains: an integrated assessment of potential GHG avoidance in Ireland

This paper combines life-cycle analyses and economic analyses for Miscanthus and willow heat and electricity fuel-chains in Ireland. Displaced agricultural land-uses and conventional fuels were considered in fuel-chain permutations. Avoided greenhouse gas (GHG) emissions ranged from 7.7 to 35.2 t CO₂ eq. ha⁻¹ a⁻¹. Most fuel-chain permutations exhibited positive discounted financial returns, despite losses for particular entities at a farm-gate processed-biomass price of €100 t⁻¹ dry-matter. Attributing a value of €10 t⁻¹ CO₂ eq. to avoided GHG emissions, but subtracting financial returns associated with displaced fuel supplies, resulted in discounted annual national economic benefits (DANEBs) ranging from −457 to 1887€ ha⁻¹ a⁻¹. Extrapolating a plausible combination of fuel-chains up to a national indicative scenario resulted in GHG emission avoidance of 3.56 Mt CO₂ eq. a⁻¹ (5.2% of national emissions), a DANEB of 167 M€, and required 4.6% of national agricultural land area. As cost-effective national GHG avoidance options, Miscanthus and willow fuel-chains are robust to variation in yields and CO₂ price, and appear to represent an efficient land-use option (e.g. compared with liquid biofuel production). Policies promoting utilisation of these energy-crops could avoid unnecessary, and environmentally questionable, future purchase of carbon credits, as currently required for national Kyoto compliance.     

An analysis of precipitation chemistry measurements in Shimane,Japan

In order to understand the concentration and deposition levels of the major ions in Shimane, on the Japan Sea coast where precipitation chemistry data are scarce, the precipitation was collected at three sites (Matsue, Gotsu and Masuda) from April 1985 to March 1988.The mean precipitation chemistry was very close to each other except for the seasalt concentration. Masuda showed a halved seasalt contribution compared with the other sites. The volume-weighted annual pH mean at each site ranged from 4.6 to 4.9. Nitrate to SO₄⁻² equivalent ratios were in the range of 0.2 and 0.4 throughout the year. Ammonia and calcium species are interpreted to have neutralized approximately 70% of the original sulfuric and nitric acids.The annual depositions of the major ions in g m⁻²y⁻¹ were as follows: H⁺; 0.023–0.037, NH₄⁺; 0.57–0.68, Ca²⁺; 0.51–0.92, SO₄²⁻; 3.29–5.04, NO₃⁻; 1.20–1.70. These levels are of the same intensity as corresponding values of JEA network results.     

In-cloud scavenging and deposition of sulfates and nitrates: Case studies and parameterization

Scavenging of sulfates and nitrates—two most common ions leading the cloudwater acidity—was investigated during field studies atop a site in Mt. Mitchell (35°44′05″N, 82°17′15″W) State Park where the highest peak (2038 m MSL) of the eastern U.S. is located. Experiments were conducted during the growing seasons (15 May–30 September) of 1986 and 1987 using an instrumented meteorological tower (16.5 m tall) and a passive cloudwater collector. A cloud episode that occurred on 12 October 1987, was also comprehensively investigated. Clouds were frequently observed in which the Fraser fir and red spruce stands stayed immersed 28% and 41% of the time during the 1986 and 1987 seasons, respectively. Rate of cloudwater deposition on the forest canopy was determined using an inferential cloud deposition model. It was found by analysing nine short duration (lasting 8 h or less) and 16 long duration cloud events that the ionic concentration (SO₄²⁻ and NO₃⁻) is inversely proportional to the rate (I_c) of cloudwater deposition (in mm h⁻¹) and can be expressed by the following relationship: [SO₄²⁻] = aI_c^−b or [NO₃⁻] = aI_c^−b. Theoretical arguments leading to these relationships are presented. The b values for predicting NO₃²⁻ concentration are found in the range of 0.14–1.24 (mean = 0.48) for short duration and 0.062–0.63 (mean = 0.27) for long duration cloud events, respectively. The corresponding b values for predicting NO₃⁻ concentrations are 0.19–1.16 (mean = 0.49) and 0.072–0.59 (mean = 0.27), respectively. When the b parameter was between 0.2 and 0.6, the correlation coefficients between measured and predicted ionic concentrations were found to exceed 0.7. The parameter a is shown to represent the maximum ionic flux for a given cloud event. The ratio of the a parameter for SO₄²⁻ to NO₃⁻ varied between 1.75 and 6.95, indicating that the SO₄²⁻ contributes to the total ionic concentration substantially more than the NO₃⁻ leading to the conclusion that the cloudwater acidity is primarily due to the presence of sulfuric acid which has been demonstrated to cause foliar injury and growth retardation in red spruce trees. The above parameterization is similar to the one that is frequently used to relate ionic concentration in precipitation to the rainfall rate. In order to understand physico-chemical processes leading to the proposed parameterization schemes, meteorological and chemical variables are comprehensively analysed for one short duration and two long duration cloud events. The concentrations of principal ions (SO₄²⁻, NO₃⁻, H⁺ and NH₄⁺) during the short duration cloud events were found to be much higher than those during the long duration ones, especially at colder temperatures. Such short cloud events have a potential of causing foliar narcosis in red spruce stands because of unusually acidic cloudwater to which these stands stay exposed intermittently during each growing season.     

Estimates of ion sources in deciduous and coniferous throughfall

Estimates of external and internal sources of ions in net througfall deposition were derived for a deciduous and coniferous canopy by use of multiple regression. The external source component appears to be dominated by dry deposition of Ca²⁺, SO₂ and NO₃⁻ during dormant and growing seasons for the two canopy types. Increases in the leaching rates of K⁺ and Mg²⁺ during the growing season reflect the presence of leaves in the deciduous canopy and increased physiological activity in both canopies. Internal leaching rates for SO₄²⁻ doubled during the growing season presumably caused by increased physiological activity and uptake of SO₂ through stomates. Net deposition of SO₄²⁻ in throughfall during the growing season appears highly dependent on stomatal uptake of SO₂. Estimates of SO₂ deposition velocities were 0.06 cm s⁻¹ and 0.13 cm s⁻¹ for the deciduous and coniferous canopies, respectively, during the dormant seasons, and 0.30 cm s⁻¹ and 0.43 cm s⁻¹ for the deciduous and coniferous canopies, respectively, during the growing season. For the ions of major interest with respect to ecosystem effects, namely H⁺, NO₃⁻ and SO₄²⁻, precipitation inputs generally outweighed estimates of dry deposition input. However, net throughfall deposition of NO₃⁻ and SO₄²⁻ accounted for 20–47 and 34–50 per cent, respectively, of total deposition of those ions. Error estimates of ion sources were at least 50–100 per cent and the method is subject to several assumptions and limitations.     

Analysis of atmospheric ozone measurements over a pine forest

Vertical and horizontal profiles of ozone concentration have been measured within the atmospheric boundary layer over the pine forest located in the southwest of France (Landes Forest). Evidence for an ozone depletion in lower layers is obtained from the analysis of vertical profiles recorded at the end of the night. In terms of deposition at the upper canopy level, this corresponds to a disappearance rate ranging between 0.2 and 0.5 cm s⁻¹. The horizontal profiles obtained at midday reveal that ozone vanishes at a rate of the order of 5 × 10⁻⁵ ppb m⁻¹ when air mass moving in the advection direction passes over the forested area. These results are consistent with those obtained by numeric simulation in the case of low emission rates of nitrogen oxides. On the basis of these measurements, the expression of the ozone budget within the atmospheric boundary layer is discussed and compared with the data obtained from the simulation study.     

The evaluation of some photochemical smog reaction mechanisms—III. Dilution and emissions effects

The effects of dilution and emissions on the performance of four widely-used photochemical smog reaction mechanisms are examined, using newly available experimental data from the CSIRO outdoor smog chambers. It is found that the Caltech, CB-XR and CB-IV mechanisms overpredicted the smog produced (SP) maximum associated with the dilution by 14–29% when the dilution rate was 22% h⁻¹ and by 53–73% when the rate was 55% h⁻¹. The CB-III mechanism underpredicted the maximum by 17% at the lower dilution rate (when the cumulative light flux was moderate), but agreed with the observations at the higher dilution rate (when the cumulative light flux was high). Five emissions experiments were conducted at moderate to very high HC/NO_x ratios (8.3–51.0). When HC were injected during the light-limited regime the rate of formation of SP increased markedly; when NO was injected in this regime there was less effect (a slight decrease in slope). When HC were injected during the NO_x-limited regime there was a gradual decline in SP concentration, but when NO was injected during the NO_x-limited regime there was a dramatic increase in SP, leading to a second light-limited regime and a second NO_x-limited regime. When both HC and NO_x were injected during the light-limited regime, marked increases in slope occured after each injection. The Caltech model gave good to excellent agreement with the data. In one experiment it overestimated the final SP by 8%. The CB-III model performed next best, sometimes in excellent agreement with the data, but at other times underpredicting the final SP by 14–18%. The CB-XR and CB-IV models underpredicted the final SP for most of the experiments by up to 38%.