Photodegradation is the direct breakdown of organic matter by radiation. Photodegradation is known to be an important pathway in aquatic ecosystems . Recently, the possible importance of photodegradation in terrestrial ecosystems has been suggested . Photodegradation can play an important role in arid ecosystems, where microbial decomposition is restricted . estimated that in arid ecosystems 19 % of the annual CO2 flux is induced by photodegradation and, in dry summer conditions, even 92 % of daytime CO2 emissions can be attributed to this process.

Photodegradation is attributed to UV as well as visible radiation . The biochemical mechanisms behind photodegradation-induced carbon fluxes are not clear; it is proposed that solar radiative energy breaks down the bonds of carboxyl, directly producing CO2 and other gas species . It has been hypothesized that rates of photodegradation depend on plant and litter tissue type: lignin, one of the most recalcitrant tissue in plant material (to microbial decomposition), is expected to be most sensitive to photodegradation . However, while studies reporting photodegradation are multiple, recent studies, aiming to further investigate the process, were unable to observe the effects of photodegradation . A reason for this discrepancy has not yet been found . It is important to notice that in literature, the term photodegradation is sometimes also used for the indirect effects of radiation on decomposition. One example is microbial facilitation: radiation breaks down organic compounds into smaller molecules, which are then more easily degradable for microbes. For a review on studies done on photodegradation, please see King et al. (2012).

A less studied abiotic degradation pathway is thermal degradation, the temperature-dependent degradation of carbon in the absence of radiation and possibly oxygen . However, photodegradation is considered the more dominant abiotic CO2 producing process . Besides CO2, CO and CH4 are also reported as products of photo- and thermal degradation .

The role of CO in soils and ecosystems is not well understood. Soils are known for being sources as well as sinks of CO . Most likely, the main cause for soil CO uptake is the oxidation of CO to CO2 or CH4 by soil bacteria or soil enzymes . Soil CO consumption is found to be dependent on atmospheric CO concentrations and the consumption rate is usually expressed in deposition velocity: the uptake rate divided by the CO concentration .

Soil CO emissions have also been reported and are thought to be of non-biological origin . For example, soil CO emissions were found in peatlands and in arid soils . Living plants are also known to emit a small amount of CO . However, senescent plant material has been shown to emit 5 to 10 times more than photosynthesising leaf material . These fluxes, mostly determined in laboratory studies, were attributed to thermal degradation and, to a larger extent, photodegradation .

Studying photodegradation is difficult due to the multiple (indirect) effects radiation has on total biological decomposition. For example, UV-radiation is known to inhibit microbial processes, to change (senescent) tissue chemistry and to alter the dominating microbial and fungal communities, thereby affecting microbial decomposition rates in both directions . Differentiating photodegradation-induced fluxes from biological sources in field experiments can be achieved by the comparison of different flux measurement techniques such as eddy covariance (EC) measurements vs. flux chamber measurements and/or soil gradient measurements, in that one method does not receive solar radiation . This approach requires that the areas or footprints sensed by the different techniques are fully homogeneous, which is not often the case and hard to validate. To study the effects of photodegradation (in field or laboratory), radiation filters can also be used to expose samples to different types or amounts of radiation .

Studying the role of thermal degradation-induced carbon fluxes is challenging, especially for CO2 due to the accompanying effect temperature has on microbial decomposition. To study thermal degradation-induced CO2 production, microbial decomposition should be absent, which can only be achieved in laboratory studies .

Previous field and laboratory studies on the role of direct or indirect abiotic degradation report very contrasting results . More specific studies are thus needed to better understand this process and its role in the carbon cycle. In this study, we present the results of field and laboratory measurements aimed to evaluate the role of direct photodegradation and thermal degradation in an arid ecosystem.

We performed a field experiment in a grassland (IT-Ro4, harvested cropland, approximately 250 m by 450 m, lat 42.37∘ N, long 11.92∘ E, 147 ma.s.l.), in the province of Viterbo, Italy. The climate is Mediterranean, with a typical drought period covering approximately 2 months during summer (July–August). Mean annual temperature is 14 ∘C and annual rainfall is 755 mm. Such climatic characteristics make the site suitable for abiotic degradation studies. The underlying material is tuff, soil texture is clay loam and soils are classified as Eutric Cambisol. Yearly, the field site is ploughed to a depth of 20 or 50 cm. Just before the experiment, oat and vetch were cultivated. During the experiment, vegetation was not managed and was a mix of invasive species such as Amaranthus retroflexus, Chenopodium spp., Conyza canadensis, Artemisia vulgaris, Cirsium spp., Mercurialis annua and Polygonum spp. The field study was conducted in July–September 2013. At the beginning of the experiment, most vegetation was dried out, however, patches of active vegetation were observed. Temperature and rainfall during measurements were representative for the period (hot and dry) (Fig  2), however, the preceding spring had been cold and rainy in respect to the average.

IT-Ro4 is an experimental site managed by the University of Tuscia (Viterbo). Continuous EC measurements of scalars and energy fluxes are performed (LI-7500 open path analyzer, Licor, Lincoln, Nebraska, USA; Windmaster Pro sonic anenomemeter, Gill, Hampshire, UK) along with meteorological and environmental measurements (CNR-1, Kipp & Zonen, Delft, the Netherlands; soil water content, CS616, Campbell Scientific, North Logan, USA; soil temperature, CS107, Campbell scientific, North Logan, USA; soil heat flux, HFT3 Soil Heat Flux Plate, Campbell scientific, North Logan, USA).

The analyzer used in this study is based on a Fourier Transform Infrared (FTIR)-spectrometer (Spectronus, Ecotech), for details on the FTIR-analyzer, see . An FTIR is capable of measuring air concentrations of CO2, CH4, N2O, CO and δ13CO2 simultaneously. Before being measured, air samples were dried by a nafion dryer and by a column of magnesium perchlorate. Measurements were corrected for pressure and temperature fluctuations and for cross-sensitivities . Background measurements and a calibration routine using two standard gas cylinders were performed weekly. We designed an external manifold box which allowed us to connect the FTIR to a flux gradient (FG) setup and to two flux chambers (FC), simultaneously. Both methods provide air concentration data as well as flux data. In this paper, only CO2 and CO flux data are presented.

FG measurements were performed once per hour and performed at the same location as the EC tower. More information about the FG system can be found in the Appendix. For FC measurements, six soil collars (50 cm × 50 cm) were inserted to 10 cm depth a week before the start of the experiment. Positions of soil collars were checked for being undisturbed and representative. The two flux chambers (open dynamic chambers, 50 cm × 50 cm × 50 cm, produced by Karlsruhe Institute of Technology, Germany) consisted of a stainless steel frame, UV-transparent acrylic sides (Acryl Glass XT solar, 3mm, UV-transparent) and a vent tube, and were tightened by the use of clamps and rubber air strips. Transparency of the acrylic material was measured and reported to be > 90 % in the UV and visible wavelength band (280–700 nm). Two fans per flux chamber were continuously running, insuring well-mixed headspace air. Automatic chamber closure (once per hour) was made possible by use of a pneumatic system regulated by the valve manifold box. Air flow from the flux chambers to the FTIR was initiated by a membrane pump placed behind the measurement cell, set to 1 Lmin-1. Air flow was measured every 2 min continuously for 20 min in flow mode. Chamber opening and closure was after 4 and 18 min, respectively. Sampling lines from the chambers were of equal size and material and were tested for leaks regularly. Chamber temperatures were recorded by temperature loggers (Voltcraft DL-1181THP). Fluxes were derived from concentration increases after chamber closure, by use of linear regression. Gas fluxes were calculated by: F=VPRSTδCδt, wherein V is the volume of the chamber (m3), P the chamber air pressure (Pa), R the gas constant (8.314 m3PaK-1mol-1), S the chamber surface area (m2), T the chamber air temperature (K) and δC/δt is the gas concentration change over time (mol mol-1s-1). For flux calculations, only the concentration increases between 2 and 10 min after closure were used. Concentration increases were checked for non-linear trends and, if found, not used. Flux standard deviations were derived from the propagated standard deviations of the regression slope.

When homogeneity in footprint can be assured, micrometeorological and FC methods can be compared and used to study the role of photodegradation. Flux chambers can be shielded from incoming radiation, preventing photodegradation-induced carbon production, while micrometeorological methods capture all fluxes. Comparing the two methods therefore gives an indication of the presence and the magnitude of photodegradation-induced carbon fluxes . The use of this method was planned for our field experiment, but could not be applied due to lack of conformity between flux methods footprints, because of sparse photosynthetically active vegetation present in the footprint of the FG technique, causing the methods to be incomparable.

To study photodegradation, two different flux chambers, one with and one without solar radiation exposure were used. During this experiment, the flux chambers were measuring six fixed chamber locations; chambers were manually moved every few days. One flux chamber was made opaque by the use of light excluding aluminum foil (on 5 August). On the days before (28 July–5 August), all positions were compared by measuring the locations with transparent chambers. On 3–5 August, the same locations were measured (with transparent chambers) as on 5–8 August, when one of the two chambers was covered. Both locations showed very similar CO2 and CO flux patterns. Unfortunately, on 8 August, a leak formed in the opaque chamber system, therefore direct comparison between the two treatments is limited to 3 days. Flux measurements made by the opaque chamber after 8 August are not shown. With blank measurements, the flux chambers were tested for internal CO2 and/or CO production. No CO2 production was found. Minor CO production was found during the day, negligible in comparison to field CO production: values presented in this paper are not corrected for this.

Studying thermal degradation-induced CO2 production in the field is not possible due to the simultaneous temperature response of biological CO2 production. For CO, no temperature-dependent biological CO production is expected, wherefore measurement of thermal degradation-induced CO production in the field is possible. To study the role of thermal degradation in field CO exchange, chamber temperature sensors were installed, measuring air temperature every minute.

Two different laboratory experiments were performed to study photo- and thermal degradation. Grass samples (senescent above ground grass material, mix of species as described in Methodology, pieces between 20–80 cm, not ground) for the laboratory experiment were taken from the field site. Mixed soil material samples were taken from the upper 3 cm of the soil, soil samples were not sieved. Both sample types were dried at 35 ∘C for 72 h, to assure microbial activity to be negligible .

Photodegradation of senescent grass material was studied with a system consisting of a metal cylinder, inner diameter = 6.5 cm, height = 25 cm, area = 33 cm2, with an acrylic cap, which could be closed by screws. Transmittance of cap was measured and was 0.2 (250 nm), 6.1 (260 nm), 35.9 (270 nm), 73.9 (280 nm), 89.6 (290 nm) and approximately 94 % for larger wavelengths. The cylinder was placed beneath a UV-A and UV-B source (manufacture instrument: Isitec GmbH, Bremerhaven; UV-A lamp: Philips TL 60W/10R (peak emission at 375 nm), UV-B lamp: Philips TL 40W/12RS (peak emission at 310 nm)). Radiation intensities at the sample location were quantified by use of an OceanOptics USB 2000 spectrometer with an optical fibre patch cord (P200-2-UV/VIS) and by an ILT1700 research radiometer with accompanying optical filters and are reported as comparison to natural radiation measured with the same instruments (determined in summer in Northern Germany, midday, no clouds, pointed at sun). Instrument radiation in the UV-A wavelength band 360–400 nm was measured to be 1.6 times higher than natural radiation, with the peak emission being at 375 nm (2.9 times natural radiation). Instrument radiation in the wavelength band 200–320 nm was measured to be 2.9 times higher than natural radiation, with the peak emission being between 290 and 310 nm (7.7 times natural radiation). During the experiment, different samples (empty cylinder, 2 gram-sample and 4 gram-sample) were exposed to different types/amounts of radiation (no radiation, UV-A and/or UV-B radiation). Grass in the cylinders was positioned so that at least 80 % of the surface bottom was covered with grass material. During the experiments, air was continuously circulated from the cylinder to the FTIR and measured once per minute; emissions were derived from the measured concentration changes. Cylinder temperatures were monitored by an internal temperature probe (GTH 175/PT, Greisinger Electronics) and remained constant over the experiments (21, sd = 0.15 ∘C). Every treatment was performed for 30 min and was duplicated.

To study thermal degradation, a glass flask (inner diameter = 6.7 cm, height = 6 cm) was placed in a closed loop with the FTIR. For this experiment, only glass and stainless steel materials were used. 4 grass samples of 2 grams and 4 soil samples of 30 grams were taken. The grass sample was distributed equally in the flask. The soil sample was not sieved and filled approximately 1 cm (height) of the glass flask. The samples were heated in temperature steps of 5∘ (20–65 ∘C) by use of a controlled temperature water bath. Temperature time steps were 20 min. During the experiments, air was circulated from the glass flask to the FTIR and measured once per minute. After approximately 3 min, a stabilization in the CO production could be observed. Emissions were derived from the measured concentration changes. Glass flask air temperatures were manually measured to check if water bath temperature was representative for grass and soil material temperatures; after 5 min, the glass flask air temperature had reached the same temperature as the water. All experiments were performed in duplicate and in dark conditions.

In the results sections, the given regression coefficients from polynomial fits are the explained sum of squares divided by the total sum of squares.

FG CO2 fluxes are shown in the Appendix. FG CO uptake (up to 1 nmolm-2s-1) and emission (on average 2 nmolm-2s-1) at night were measured (Fig  1). During the day, large (≥10 nmolm-2s-1) CO emissions were recorded (Fig. 1). Based on the 31 days of FG measurements, on average net 150 µmol CO m-2 per day was estimated to be emitted. FC CO2 and FC CO fluxes of the transparent flux chamber can be seen in Fig. 2, rain events and incoming solar radiation are indicated. FC CO2 fluxes showed a diurnal pattern with small emissions at night (1 µmolm-2s-1) and higher emissions during the day (up to 8 µmolm-2s-1). Large rain events on 20 and 27 August (6.6 and 2 mm) caused a short increase in chamber CO2 fluxes. Locations without organic surface material (indicated as bare soils in Fig  2) showed slightly lower CO2 and CO fluxes.

At night, CO uptake of maximum 0.8 nmolm-2s-1 was observed. During the day, emissions up to 3 nmolm-2s-1 were observed. Over the course of the experiment, nightly CO uptake was continuously decreasing. The rain events caused a clear increase in nightly CO uptake, after which the decreasing continued (Figs. 1 and  2). Based on 36 days of FC measurements, on average net 30 µmol CO m-2 per day was estimated to be emitted.

Photodegradation was studied by comparing opaque and transparent chamber measurements of 3 days (5–8 August) and by analysis of transparent FC data of a period in August (period with fixed location, stable weather conditions and no precipitation). Analysis of different periods (different locations with similar conditions) showed similar patterns.

Possible photo- and/or thermal degradation-induced CH4 fluxes are not shown or evaluated here: FG CH4 fluxes were too small for dependency analysis and CH4 chamber fluxes mostly showed uptake, indicating a different process than photo- or thermal degradation.

EC and FG measurements showed that the arid grassland was not yet in a dormant state; significant CO2 uptake was observed during the day (Appendix Fig. B1). FC CO2 measurements, performed on locations without photosynthetic active vegetation, solely showed positive CO2 fluxes, with peak emissions during the day up to 8 µmolm-2s-1. Figure 3a shows that CO2 fluxes mostly relate to air temperatures, and poorly relate to soil temperatures (not shown). Expected is that most CO2 production takes place close to the surface where the temperature follows air temperatures closer than it follows soil temperatures at 10 cm depth. In the ecosystem, the rain events resulted in an increase in CO2 production for several days, showing the typical water-dependent response of arid ecosystem respiration (Fig. 2 and B1).

In the thermal degradation laboratory experiment, CO2 production from senescent plant and soil material was observed during lower temperatures (20–40 ∘C), indicating remaining biological activity, even after drying. Above 50 ∘C, an increasing CO2 production was observed with increasing temperatures, therefore expected to be (partly) of non-biological origin. Possible abiotic CO2 production of approximately 3 nmolmin-1gr-1 for senescent grass material was observed. Extrapolating the thermal production rates of the senescent grass material to field conditions (assuming 200 gr of senescent plant material per m2 at 55 ∘C), would result in a minor flux of 0.01 µmolm-2s-1, in comparison to observed field fluxes of >1 µmolm-2s-1. Based on the observations in the laboratory, it is expected that the soil material also produces thermal degradation-induced CO2 fluxes. However, considering the relative cold and wet conditions of the subsurface soil material in the field, compared to laboratory conditions and to surface temperatures, it is expected that soil thermal degradation fluxes are minor in comparison to soil biological fluxes.

Other studies have observed thermal degradation-induced CO2 fluxes with higher rates (approximately 125 nmolCO2gr-1min-1 for C3-grass at 55 ∘C), but also at lower temperatures . We can not verify this observation for our field material. Based on our observations, we propose that under natural conditions, when soil surface temperatures and especially soil subsurface temperatures rarely exceed 55 ∘C, thermal degradation-induced CO2 fluxes do not play an important role in comparison to biological production, even in arid regions such as our study area.

We observed that chamber design can strongly influence chamber temperatures: during midday, the opaque and transparent chamber temperatures could differ up to 10 ∘C. As observed in the laboratory experiment, unnatural high temperatures might lead to abiotic thermal CO2 production. A research methodology aimed at measuring photodegradation can unintentionally result in high surface temperature levels, which could lead to unrepresentative high abiotic CO2 production estimates.

The simultaneous use of opaque and transparent chambers was employed to study the effect of radiation on carbon fluxes in the field. Blocking radiation had no visible effect on field chamber CO2 fluxes (Fig. 3a and c). CO2 flux measurements performed on bare soil locations (soils without organic surface material) seemed lower than other locations; senescent surface material seemed to contribute to total CO2 fluxes (Fig. 2a). However, only 3 days of bare soil measurements are available and no opaque chamber measurements on bare soil are present, therefore comparison is restricted.

The flux chambers, which were used to assess photodegradation, had a transparency of 90 % or higher in the UV-B, UV-A and visible wavelength band. For our field experiment, we can therefore conclude that no large direct photodegradation fluxes (as suggested by Rutledge et al., 2010 of 1 µmol m-2 s-1) have been induced by natural sunlight intensities. In the laboratory experiment, field site grass samples received above natural-intensity UV-radiation. In this experiment, no direct photodegradation fluxes were observed from field site grass material. While the laboratory experiment presented here does not prove that there are no photodegradation fluxes at all, the results from the laboratory experiment support the conclusion from the field experiment that direct photodegradation fluxes in arid ecosystems are not as important as suggested by a previous study .

The experiment was conducted on a field site situated in a Mediterranean climate. Based on annual precipitation and on measured respiration values, the ecosystem might seem too wet to be suitable to measure arid ecosystem processes. However, the climate is known for the precipitation free summers with high irradiation, causing the soil surface and surface materials to be fully dried out in summer. Since photodegradation is taking place at the soil surface, the ecosystem can be considered suitable for the assessment of this arid ecosystem process. The absolute amounts of possible photodegradation fluxes are not influenced by the respiration fluxes. The expected rates of photodegradation fluxes (of 1 µmol m-2 s-1, ) should have been detectable, even when mixed with respiratory fluxes.

Similar to what has been found by , we did not observe the effects of photodegradation in the field nor in the laboratory: no direct photodegradation-induced CO2 fluxes have been observed. This is in contrast to other photodegradation studies, which have reported photodegradation fluxes in the field or in the laboratory . Potential explanations for this difference are: (a) the used field methodology in the previous study was not suitable for measuring direct abiotic degradation fluxes; (b) the role and significance of photodegradation differs per material and per field site; (c) studies might (partly) have misinterpreted thermal degradation fluxes as photodegradation fluxes or (d) photodegradation fluxes were too small to be observed by the presented method. We therefore do not question the existence of the photodegradation process, but doubt its suggested large role in arid ecosystems. However, as shown, the magnitude and the potential importance of thermal degradation-induced CO2 fluxes in arid ecosystems are still unknown.

During the measurement period, both CO uptake and emission have been observed by the FG method (patches of green active vegetation inside the footprint) as well as by the FC method (no photosynthetic active vegetation contributing to the fluxes) (Figs. 1 and 2). CO exchange measurements from FG and FC differed largely, most likely caused by the difference in footprint.

During the night, uptake of up to 1 nmolm-2s-1 of CO was observed, which is most likely caused by microbial oxidation to CO2 or CH4 . The CO uptake was decreasing over time but a rain event caused an enhanced uptake for some days (Figs. 1 and 2). Soil biota being responsible for the CO uptake seems plausible since the effect of drought (decreasing uptake over time) and the effect of the rain (enhanced uptake) indicate a biological process. Nevertheless, with solely biological CO uptake taking place, one would expect higher uptake during warmer temperatures and no CO emission. It is expected that an abiotic process occurs simultaneously with the biotic uptake of CO, leading to a buffering effect on CO uptake. For this reason, CO deposition velocities could not be calculated.

We propose that the observed CO emissions in the flux chambers are caused by thermal degradation. FG measurements showed CO emissions during the day as well as during the night, indicating that CO is not (solely) produced by photodegradation (Fig. 1). By means of opaque chamber measurements, lower CO fluxes, in comparison to transparent chamber measurements, were detected. However, as described before, FC temperatures were strongly affected by the blocking of solar radiation. Analysis of CO fluxes showed a strong correlation with FC temperatures, and no relationship with radiation input, indicating that it was not the absence of radiation, but the indirect effect on temperature that caused the lower CO emissions (Figs. 3 and 4).

FC CO fluxes were ranging between -1 and 2.5 nmolm-2s-1 and only originated from soil or surface litter, since photosynthetic active vegetation was absent. Measured CO emissions are higher than reported for CO emissions from living plants and similar to values found for senescent plant material . However, the measurements are a cumulative signal of uptake and emission and can therefore not be compared directly to other studies.

In the laboratory experiment, where grass from the field site was exposed to above natural intensity UV-radiation, no photodegradation-induced CO fluxes were observed. However, significant thermal degradation-induced fluxes from the senescent grass and soil material were measured, even measurable at low temperatures (20 ∘C). At 50 ∘C, a thermal CO production rate of senescent grass material of 0.13 nmolmin-1gr-1 was found. Extrapolating this observation to field conditions (assuming 200 grams of senescent plant material per m2 at 50 ∘C), would result in a flux of approximately 0.4 nmolm-2s-1, which is approximately 5 times lower than the net measured field CO fluxes. Extrapolating the thermally-induced CO production rate of the soil material to field conditions would result in an estimated production of approximately 1 nmolm-2s-1 from the upper 3 cm of the soil during a summer day. However, while this estimate indicates that abiotic thermal soil CO production indeed might play a major role, for accurate estimates for net soil CO uptake or emission, more information about biological CO uptake and about the soil profile is needed.

The observed field chamber CO fluxes are suggested to be a cumulative signal of biological uptake and abiotic thermal degradation. Both processes were fitted over chamber temperatures. For the fitting of biological CO uptake, a Q10-value of 1.8 was chosen . To match the cumulative measured CO fluxes (purple diamonds in Fig. 5), a higher Q10-value of 2.1 for the abiotic thermal soil CO production was fitted (R2=0.85).

The laboratory measurements were used to experimentally determine the Q10-value of thermal degradation-induced CO fluxes. Q10-values of 2.14 for senescent grass and 2.00 for soil material were measured. These values are similar to the Q10-value which was fitted for the thermal degradation process to match the cumulative field measurements, as described in the previous paragraph (Fig. 5).

The soil CO uptake process, taking place below the surface, is subject to buffered chamber temperatures, and therefore the chosen Q10-value might be an underestimation. Also, the biological soil uptake is not expected to follow the Q10-temperature response at higher temperatures (>35 ∘C). Nevertheless, the difference in temperature response (as a consequence of different Q10-values or as a consequence of buffered temperatures) causes biological CO uptake to be dominant during colder (chamber) temperatures, and thermal degradation to be dominant during warmer (chamber) temperatures. During our field experiment, thermal degradation started to be dominant from approximately 25 ∘C (chamber temperature) and followed an exponential curve with higher temperatures (Fig. 5).

The temperatures inside the chamber were higher than the temperatures outside the chamber. Although this will result in higher fluxes inside the chamber compared to the ecosystem around it, the correlation between temperatures inside the chamber and the CO flux should be representative for the ecosystem. The laboratory study shows a similar relationship between temperature and CO flux. According to our results, the temperatures outside the chamber are high enough to induce significant thermal degradation fluxes. This is supported by the measured CO fluxes by the FG technique. FG CO emissions were higher, likely due to its footprint which contained relatively more dead vegetation (thermal degradation material) since, for practical reasons, the chambers were placed over lower dead vegetation. Also, the FG footprint contained active vegetation, which is another possible CO emitting source .

Overall, the measurements show that the field site is a net source of CO during the summer months, affecting the atmospheric chemistry, at least at plant level, via OH- depletion. More field measurements on annual CO exchange are needed to better understand the role of thermal degradation in CO and CO2 exchange in arid regions.