Soil respiration in drylands

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Chapter 1. Introduction

Soil respiration in drylands is poorly understood and largely understudied, despite the fact that deserts cover more than 30% of the Earth's land surface (Thomas, 1997). Given the current interest in global carbon cycling, and in particular the degree to which dryland soil carbon cycling may contribute to it (Lal 2004; Luo & Zhou 2006), there remains a striking paucity of knowledge.

It is only within the past decade that the broad rôles played by Biological Soil Crusts (BSCs) in dryland environments have begun to be seriously appreciated (Belnap & Lange 2003). Interest remains predominantly focussed on their rôle in maintaining soil stability and their influence on hydrological processes. More recently, awareness has increased that BSCs play a 'boundary mediating' rôle, partially governing the exchange between dryland soils and the atmosphere (Belnap et al, 2003a, 2003c; Viles, 2008). Of real importance is our understanding that BSCs can both emit CO2 through respiration, but also sequester carbon (C) through photosynthesis. The latter, in particular, is only possible when crusts are sufficiently hydrated so as to be metabolically active.

This study focuses on the boundary mediating rôle of BSCs, and partially arises from the renewed interest in and recent adaptations of Noy-Meir's (1973) 'pulse-reserve' model (Schwinning & Sala, 2004; Schwinning et al, 2004; Ogle & Reynolds, 2004; Humxan et al, 2004a; Austin et al, 2004). The paradigm shift in investigative approaches to drylands, now understood as existing in a non-equilibrium state (Vetter 2005), underwrites the experimental rationale of the study.

Broadly, the study investigates CO2 surface flux, and the environmental factors that may influence flux activity. The study involves sustained in situ sampling from the southern Kalahari Desert, close to the border between Botswana and South Africa and concentrates on identifying distinct CO2 pulses after simulated precipitation events. The factors that might influence the magnitude of pulse (precipitation amount, Photosynthetically Active Radiation, temperature) are considered. Equally, the study considers those factors that might induce photosynthetic C uptake by BSCs. Subsoil respiration is also sampled, and temporal and spatial variability in flux activity is also investigated.

Literature Review

Despite their often low vascular plant cover, it would be misguided to assume that drylands are devoid of significant biological life (Belnap & Lange, 2003). BSCs are now understood to be present in all arid and semi-arid regions (Thomas & Dougill, 2007; Zaady et al, 2000), constituting more than 70% of living cover (Belnap et al, 2003), and their rôle in vital geomorphic and ecological processes is increasingly recognised (Belnap et al, 2003a; Viles, 2008).

BSCs are "an intimate association between soil particles and cyanobacteria, algae, microfungi, lichens, and bryophytes...which live within, or immediately on top of, the uppermost millimetres of soil" (Belnap et al 2003a, p.3). Cyanobacteria based crusts in particular have the ability to produce exopolysaccharide secretions, which are well recognised to increase soil cohesive strength and reduce erodibility (Thomas & Dougill, 2007). Correspondingly, with the increased focus on land degradation in drylands, significant efforts have been made to research the relationship between BSCs and aeolian and fluvial erosion, with notable contributions by Belnap & Gilette (1998), Argaman et al (2006), Bowker et al (2008) and McKenna Neuman & Maxwell (2002).

Biological Soil Crusts as a mediating boundary

Less attention has been directed towards the ability of BSCs to modulate the boundary between the atmosphere and soil; they are essentially a miniature ecosystem between the two (Bamforth, 2008). Belnap et al (2003c) have produced some seminal work, though studies remain focussed towards the potential rôles of BSCs in dryland hydrological cycles. In particular, Belnap (2006) found the influence of BSCs on infiltration rate, surface run off and sediment production to be largely a function of the stage of succession of the crust. This finding has been reaffirmed by Aguilar et al (2009) through their research in northern Mexico. Dougill & Thomas (2004) propose a three-stage classification system for BSCs as based on their stage of succession (with Stage Three being the most developed). Indirectly, Harper & Belnap (2001) and Berkley et al (2005) discuss the influence of BSCs on vascular plant mineral uptake, finding a significant correlation between BSC presence and the bioessential mineral content of vascular plants. This is thought to affect overland hydrology, as BSCs are likely to support the 'islands of fertility' model (Garcia-Moya & McKell, 1970). Consequently, of particular note is the work of Housman et al (2006) who report that the carbon and nitrogen fixation abilities of BSCs are also differentiated as a function of successional stage. Housman et al's (2006) study is indicative of a broader appreciation of BSCs' ability to capture, transform or deflect atmospheric inputs (Belnap et al, 2003c; Viles 2008). This should not be underestimated; they have been called 'small-scale mantles of fertility' (Garcia-Pichel et al 2003b). Correspondingly, there has been a heightened interest in the Carbon (C) and Nitrogen (N) fixation potential of BSCs.

Biological Soil Crusts: Respiration and photosynthesis

Despite the buoyant interest in dryland soil respiration, Cable & Huxman (2004) stress that knowledge of the rôle BSCs play in ecosystem carbon cycling is still in its infancy. Given that drylands occupy over 30% of the land surface globally, soil crust cyanobacteria are likely to have significant populations; Garcia-Pichel et al (2003a) estimate there to be 54 x 1012gC cyanobacterial biomass in dryland crusts globally. The realisation that CO2 transformation by BSCs contributes significantly to C budgets in drylands has provided further impetus for research (Belnap et al 2003c); Lange in particular has written prolifically (1994, 1997, 1998, 1999, 2003a). Evans and Lange (2003) argue that the more evolved, lichen and moss dominated BSCs, potentially have photosynthetic rates on a par with vascular plants (120-370kgC ha-1 yr-1) although Belnap et al (2003c) stress that cyanobacterial-dominated crusts have lower photosynthetic potential (4-23kgC ha-1 yr-1).

Concerted efforts have been made to characterise the biochemistry that underlies BSCs' respiratory and photosynthetic activity. Focussing in particular on cyanobacterial crusts, they are known to be poikilohydric, i.e. they are only metabolically active given sufficient moisture availability (Zaady et al, 2000; Belnap et al 2003b; Ustin et al, 2008). BSCs are thought by some to be the most effective of any photosynthetic organism (Badger & Price 2003), their pronounced photosynthetic ability being largely attributable to their single cell CO2 concentrating mechanism. Further biochemical evidence of environment adaptation is provided by Ehling-Schulz & Scherer (1999), who argue that cyanobacteria use three different strategies to counteract damage from overexposure to UV. In a seminal paper Garcia-Pichel & Pringault (2001) present findings that cyanobacteria vertically migrate within the soil profile in response to wetting and drying events; this tacit reponse to water is thought to be unique, and likely a coping mechanism towards extreme dryland desiccation-hydration cycles. Bamforth (2004) continues, finding that protozoa are the dominant form of microfauna that contribute to metabolic functioning in hot deserts (though later research (Bamforth 2008) found markedly different protozoa in cold desert BSCs).

There has also been a move to establish the dominant environmental variables controlling BSC respiration and photosynthesis. Temperature is understood to be a critical variable, with photosynthetic activity of a rehydrated cyanobacterial crust found to be significantly higher at 25°C and 35°C than 5°C and 15°C (Zhao et al, 2008). Given the significant temperature range in many deserts, BSCs also show remarkable adaptation to extreme temperature ranges. San José & Bravo (1991) report maximum net photosynthesis at 45°C in Venezuelan Savanna algal soil crusts, whilst Lange et al (1998) discuss three different species' varied net photosynthesis and dark respiration in response to a wide temperature range in the Colorado Plateau, Utah.

The relationship between moisture, respiration and photosynthesis

Given cyanobacteria's poikilohydric nature, moisture is also thought to be a fundamentally important variable. Lange et al (1994) report, however, that whilst certain cyanobacteria can undertake net photosynthesis from high humidity water vapour alone, green algae and green algal lichen-formed crusts are reliant on liquid water to become metabolically active. Variable responses to moisture are therefore to be expected within BSC taxa, as reiterated in a later study which found carbon fixation as a function of moisture to be significantly differentiated between four different species of BSC in southern Utah (Lange et al, 1998). It is therefore unsurprising that Belnap et al (2004) conclude that any alteration in precipitation regimes will likely alter the physiological functioning of BSCs.

Dryland soil efflux

Work undertaken on dryland soil respiration and C sequestration is still rather limited (Schlesinger 1999, 2000; Luo & Zhou, 2006). Raich & Schlesinger (1992) consider CO2 flux in soil respiration at the global scale and its relationship to vegetation and climate; Conant et al (2004) more specifically study the numerous environmental controls on soil respiration in semi-arid soils, where soil moisture and temperature are noted as dominant factors. An earlier study also found that soil microbial respiration is largely moisture limited, with increased soil temperature generally leading to greater soil respiration (Conant et al 2000). In a pioneering study, Kieft et al (1987) directly link the wetting of a dry arid soil to a measurable release of organic C from the soil biomass, though it is unlikely to be a simple cause and effect relationship; Liu et al (2002) stress the common finding that the relationship between soil CO2 efflux and soil moisture is highly variable, indicative of complex mechanisms determining dryland soil efflux.

At the global scale, Lal (2004) has been particularly vocal in his research into carbon sequestration in dryland ecosystems, and notes that drylands have been suggested as the 'missing sink' in the global carbon budget (Lal 2003). However, there remains very real uncertainty concerning dryland carbon pools, well characterised by the recent lively exchange between Wohlfahrt et al (2008) and Schlesinger et al (2009). Whilst Wohlfahrt et al (2008) contend that their research demonstrates that an area of the Mojave was a net CO2 sink for 2005 and 2006, Schlesinger et al (2009) strongly question the validity of their findings. Stone (2008) summarises Xie et al's (2008) recent further contribution to the debate, suggesting that high soil salinity or alkalinity positively drive CO2 sequestration in a cold arid desert.

'Pulse-Reserve' Paradigm

The actual nature of dryland precipitation is thought to be critically important in governing both subsoil efflux and BSC activity; episodic water availability irrefutably affects dryland element cycling (Austin et al, 2004). Indeed, BSCs are exposed to significantly different precipitation regimes. Veste et al (2008) consider the respiratory activity of BSCs in Israel's Northern Negev coastal desert, where regular dews contribute cumulatively significant amounts of water to the ecosystem. Lalley et al (2006) note similar fog precipitation in the Namib desert; this is potentially biologically significant, estimated at four times the 19mm mean annual rainfall.

In recent years there has been renewed interest in precipitation patterns and their influence on dryland ecosystem processes (Huxman et al, 2004a, b). The influential 'pulse-reserve' paradigm draws on the fact that dryland precipitation "is not only discontinuous but also stochastic" (Noy-Meir, 1973: 31). It suggests that pulses of BSC activity are tightly coupled to precipitation events. Noy-Meir's (1973) hypothesis has received considerable attention, and its focus on significant temporal and spatial precipitation variability is supported by the more recent non-equilibrium hypothesis (Vetter, 2005). In recent years it has been modified; Schwinning et al (2004) stress both the importance of biologically significant pulses (drawing on Beatley's (1974) work into response thresholds) and hysterisis, the concept that ecosystems have 'memory' of past precipitation events that governs response to future events. To this end, Reynolds et al (2004) propose that sequences of precipitation events, as opposed to individual pulses, are more relevant for understanding dryland carbon cycling and the role BSCs play in mediating it. Schwinning & Sala (2004) further modify Noy-Meir's (1973) original paradigm, arguing that pulse depth and pulse duration are critical in understanding BSC response to pulse events. Indeed, they argue that the length of metabolic activity should constitute the pulse, not the original precipitation event; consequently, pulse events are likely to have different durations for different species.

Empirical evidence for both BSC activity and subsoil respiration supports the pulse-reserve paradigm. Both Huxman et al (2004a) and Ogle & Reynolds (2004) model ecosystem component responses to precipitation pulses, being careful to stress that the influence of pulse events on dryland C dynamics is significantly governed by antecedent moisture conditions. Fierer & Schimel (2003) propose a mechanism for rapid subsoil CO2 production following a simulated pulse event, though Huxman et al (2004b) find high spatial variability in CO2 efflux in Arizona, USA, following simulated pulses. Similarly high, small-scale variability is reported by Maestre & Cortina (2003) in a Mediterranean semi-arid steppe. In a laboratory based experiment, Satoh et al (2002) further reaffirm cyanobacterial adaptation to desiccation, showing rapid photosynthetic activity upon wetting, though Harel et al (2004) conclude that there is very little research into the mechanisms that allow photosynthetic recovery after hydration.

Biological Soil Crusts and subsoil efflux in the Kalahari, Southern Africa

To date, the pulse-reserve model has been little studied in Southern African drylands. Nicholson (1994, 2000) reiterates the current intense spatial and temporal precipitation variability in Southern Africa, and Schwinning et al (2004) note the likelihood of climate change in drylands forcing greater precipitation extremes with more erratic fluctuations. This is particularly true of the Kalahari, where Thomas et al (2005) envisage future climate change further reducing moisture availability in the region. Despite Weltzin et al (2003) reaffirming the criticality of research to understand how ecosystems respond to changes in precipitation, there are very few investigations into subsoil and BSC responses to precipitation pulses in the region. Of the few studies identified, there is a clear division in spatial scale studied. Whilst Wang et al (2007, 2009) compare sites over a 600km Kalahari Transect, Thomas & Hoon (2010) and Thomas et al (2008) focus in particular on a more arid site in southern Botswana. In conclusion, knowledge concerning the spatial distribution of BSCs (Dougill & Thomas, 2004; Berkley et al, 2005; Thomas & Dougill, 2007) and their ability to mediate and respond to precipitation pulse events in this critical area is extremely limited, particularly at finer spatial scales.


This study aims to further the work of Wang et al (2007, 2009), Thomas et al (2008) and Thomas & Hoon (2010). There is at present an unsatisfactory dearth of research into in situ dryland soil CO2 fluxes, the vast majority of studies to date being artificial laboratory simulations. Those that are field based are on the whole small scale and temporally coarse, and result in unacceptable disturbance of Biological Soil Crust integrity. Additionally, our understanding of the temporal and spatial variability that characterise surface and subsurface fluxes in the Kalahari is still largely in its infancy.

This study therefore investigates the effects of simulated precipitation events on both BSC respiration and photosynthesis and subsoil respiration at two sites in the southern Kalahari, with three broad aims:

  • To discover how Biological Soil Crust-mediated surface efflux/uptake varies internally within sites.
  • To discover how Biological Soil Crust-mediated surface efflux/uptake and subsurface respiration vary between two nearby (350 metres) sites on different substrates.
  • To establish what environmental factors affect Biological Soil Crust-mediated surface efflux/uptake and subsurface respiration.

Chapter 2. Location of Research: The Southern Kalahari Desert

The Kalahari's climate is subject to a distinct seasonality, crudely distinguished between wet (October to April/May) and dry. Lying within the southern hemisphere subtropical high pressure belt, winter precipitation in the northern Kalahari is particularly influenced by fluctuations in the positions of both the Inter Tropical Convergence Zone and the Zaire Air Boundary (Tyson 1986). Temperatures vary both seasonally and spatially (with frequently large diurnal temperature regimes during the dry season), and consistently high temperatures coupled with low humidity result in high potential evapotranspiration; Upington, South Africa has 3805mm annual potential evapotranspiration, for example (Thomas & Shaw 1991).

The intense temporal and spatial precipitation variability that characterises the Kalahari is of acute importance to the study. Though attempts to climatically delimit areas of the Kalahari are fraught with difficulty due to inter-annual and inter-decadal variability, general characteristics can be noted. In particular, there is a south-west north-east gradient with annual precipitation increasing in easterly and northern directions, ranging from arid to semi-arid/subhumid (Thomas & Shaw, 1991; Dougill & Thomas 2004; Wang et al 2007). Precipitation variability is also found to increase to the south. This results in a significant range in precipitation from the arid south to the more humid north-east, as shown in both Fig. 2.1 and Table. 2.1.

Though Pike (1971) concludes that half of Kalahari storms generate less than 10mm of rainfall, the nature of precipitation remains poorly documented. Whilst small scale spatial variability can often be seen as a function of convective activity, longer term temporal variability is also notable. Tyson (1986) notes a statistically significant 18 year cycle of rainfall fluctuations over southern Africa (Thomas & Shaw 1991; Tyson 1986).

The Kalahari's soils have received scant scientific attention, largely due to their low agricultural potential (Thomas & Shaw 1991), but most studies report a close relationship between geology and soils. Kalahari sand soils in particular consist of over 95% fine sand-sized, aeolian deposited sediment (Thomas & Shaw 1991; Wang et al 2007). Predominantly deep and structureless, soil organic carbon and total N content have their highest concentrations (C max = 1.5%, N max = 0.1%) in the top 10-20 cm of the soil profile (Wang et al 2007). Studies of Biological Soil Crust (BSC) prevalence have, until very recently, been entirely neglected in the Kalahari. However, among recent studies Dougill & Thomas (2004) report cyanobacterial BSC cover ranging from 19% to 40%, at four locations within the Kalahari, and Berkley et al (2005) assert the belief that BSC distribution in the Kalahari is likely to be related to Acacia mellifera (which has a particularly high prevalence in the Kalahari) and Grewia flava populations, finding crust density positively correlated to Acacia mellifera canopies.

The research for this study was undertaken on two sites at Berry Bush Farm, approximately 7km north east of Tsabong, in south west Botswana (25°56'51S 22°25'40E), as shown in Fig 2.2. The two study plots are within an enclosed area of farmland that has a history of grazing (most recently a year before the study), and vegetation is predominantly a mix of grass, woody shrubs and trees (Thomas & Hoon 2010). One plot was sited on Kalahari sand (hereafter referred to as KS), with >97% grain size of fine sand, and a pH 5.9 +/- 0.4 (Thomas & Dougill 2004). The other (hereafter referred to as Calcrete) was sited on a surface desiccated calcrete pan, with significant aeolian deposits of Kalahari sand. The vast majority of crusts in the area are cyanobacterial, typically 3-4mm thick and range in morphological development from Type 1 to Type 3, following Thomas & Dougill's (2007) classification system. Coverage varies widely; from 19-40% in pastoral areas to over 95% in undisturbed wildlife management zones. KS had approximately 55% coverage, and Calcrete 75%.

Berry Bush Farm has a precipitation record dating from September 1996. Fig 2.4 shows a histogram of precipitation events from September 1996 to June 2009.

Berry Bush's precipitation data have been part-processed by the author to produce Table 2.2 which also shows a comparison with the much longer record at nearby Tsabong. The mean precipitation event for Berry Bush is 10.66mm, though mean is a notoriously poor measurement in dryland environments. The % difference from the 1934-1988 mean is highly variable. 66.26% of all precipitation events are 10mm or less, with 85.22% of precipitation in discrete events of 20mm or less. This strengthens the rationale for the simulated precipitation events. Convective driven storms are irregular, low frequency high magnitude events. Over the analysed period there have only been 10 events of =50mm (representing only 2.72% of all precipitation events). Whilst the simulated 50mm event for the subsoil pore space CO2 levels represents a statistically rare event, the precipitation delivered in convective storms of =20mm represents, on average, 47.76% of annual precipitation. Their importance cannot be overlooked.

As Fig 2.5 illustrates, there is a marked seasonality of temperature at the study site, coupled with a significant diurnal temperature range.

This is particularly marked in the winter months (as in the study period), where it is not uncommon for temperatures to fall below freezing during the night.

Chapter 3. Methodology

Field environmental variables

All research was conducted from 25/06/09 to 07/07/09. The Kalahari Sand site was studied from 25/06/09 through to 30/06/09, the Calcrete site from 02/07/09 through to 07/07/09. At each site a number of background variables were measured.

Photosynthetically Active Radiation (PAR) was measured every four minutes using a Delta-T QS2 Quantum sensor, connected to a data logger (Delta-T, UK). Air temperature, humidity and dew point were logged every 20 minutes using a Measurement Science USB502 sensor (Adept Science, UK). A sensor had been left on the KS site the previous year, recording from 12/08/09 to 23/06/09 (as shown in Fig. 2.5). Precipitation has been continuously recorded at the study site, using a basic rain gauge, since September 1996 (see Table 2.2) by a researcher trained by the Botswanan Meteorological Organisation.

Soil surface CO2 flux

Six portable in situ closed chambers (ISCC) were used to quantify soil CO2 flux. It is well established that crusts are highly sensitive to changes in moisture, temperature and disturbance (Lange 2003a; Zaady et al., 2000). Indeed, their photosynthetic ability is dependent on maintenance of their structural integrity, in particular the chlorophyll band (Plate 3.4; Thomas 2009 pers. comm). Given the inherent importance to the study of maintaining the integrity of the crust surface, specially designed ISCC were used (also referred to as 'cells').

True in situ studies investigating dryland soil CO2 fluxes are rare; most are reliant on crust removal and laboratory simulation. Until now, those undertaken in situ have tended to alter the ambient environment (significantly interrupting the soil/environment equilibrium) or unacceptably damage the crust surface, undermining the validity of results. The ISCC used in the study are unique, as patented by Hoon et al (2009). Their defining characteristics are shown in Table 3.1.

Each site was sampled over a 6-day period. Of the six ISCC at each site, one was a control cell, and the other five experimental cells. When not being used, the ISCC lids were removed to ensure maintenance of ambient environmental conditions. Each site was subjected to simulated precipitation pulses. The simulated precipitation regimes (Table 3.2; they represent a spectrum of modal amounts derived from analysis of Berry Bush's 13-year precipitation record) were applied to all experimental cells at 06:45, using either a pipette or plant spray, dependent on amount. Sample periods started at 07:00, 10:00, 13:00, 14:00, 16:00, 19:00, 22:00. A generic sample period was as follows: Attach ISCC lid. Sample ISCC internal CO2. Wait 45 minutes. Inject syringe, and manually 'mix' (inject/purge x 2 to ensure ISCC CO2 was adequately mixed, avoiding gas stratification.) Sample CO2. Remove lid.

Subsoil Pore CO2 concentrations

A soil pit, adjacent to each plot, was dug to investigate changes in pore space CO2 concentrations as a function of simulated rainfall events. The methodology used largely followed that of Thomas & Hoon (2010). The square pit (2 x 2m) was approximately 1m deep. One face of the pit was smoothed, and measured to a depth of 75cm. Marks were made linearly every 5cm from 5cm to 75cm depth, resulting in 15 sampling points. At each point, soil temperature and soil moisture content were measured using a Delta T multivariable probe, inserted horizontally to approximately 20cm.

CO2 samples were taken at 6cm, 15cm, 30cm and 75cm. This involved inserting a gas sampling syringe (Fig 3.1) into a predrilled sampling hole in the soil pit face at the four depths, which were then left in situ for 24 hours. A 9mm sample was then taken using a gas syringe, which was injected into a pre-evacuated Exetainer® 4.5mm vial resulting in a slightly over-pressurised sample.

Samples were taken for 6 consecutive days to see the response of pore space CO2 concentrations to a 50mm simulated precipitation pulse at 07:00 on the second day. When not in use, the pit was covered with galvanised aluminium sheeting to prevent undue evaporation from the exposed soil face. Thomas & Hoon (2010) give a more detailed explanation of the methodology's rationale.

Laboratory Analysis

The Exetainer® vials containing the CO2 samples for both surface soil efflux and subsurface soil pores were analysed on return to the United Kingdom. This was manually undertaken using a portable Gas Chromatograph (GC 3000, Agilent, USA) "employing a silicon channel time of flight and thermal conductivity detector (TOF/TCD) and high purity helium (99.999%) carrier gas." (Thomas & Hoon 2010:133) Each sample had a 45 second run time.

Scanning Electron Microscopy (SEM) images were taken in early November 2009 of the Kalahari Sand crust samples harvested on 07/07/09.

Chapter 4: Results and Analysis

Local Climate

The dry season winter June-July period selected for this study was designed to maximise the effects of simulated precipitation events. The wet season usually ends in April, but 2009 was an anomalous year. Precipitation recorded for the two-month period in 2009 is the highest on the 13-year record. Given the intense temporal variability the data period is too limited to be considered representative. Nonetheless, the region's wet season usually ends in early April (Torrance, 1971), and certainly by May. The past two years can therefore be considered anomalous. This anomaly has great significance for the research conducted, as antecedent soil moisture was significantly elevated. Whilst no precipitation was recorded in the two six-day study periods (25/06/09 - 07/07/09), a 34.4mm convective storm occurred on the 10/06/09, the severity of which the landowner called "unprecedented" for the time of year. This is shown in Fig 4.1.

Consequently, the degree to which the effects of very low magnitude simulated precipitation events could be observed was questionable, given the abnormal antecedent moisture conditions.

Diurnal variability in temperature, humidity and dew point are shown in Figs. 4.2a (Kalahari Sand site) and 4.2b (Calcrete pan site).

KS site has a maximum diurnal temperature range of 35°C (-3°C to 32°C), whilst the Calcrete site has a slightly reduced range of 30.5°C (-1°C to 29.5°C). Nonetheless, both sites reflect the severe conditions to which the crusts are exposed. Particularly revealingly, both sites show strong association between temperature and dew point in the early hours of each experimental day. In the early morning on days 3 (see Plate 4.1), 4, 5 and 6 KS records temperatures low enough to permit dew formation. This quantitatively supports field observations (Plate 4.1), and suggests a previously unappreciated source of moisture for BSCs in the Kalahari (though Danin & Orshan (1999) have investigated dew extensively in Israel's Negev desert, and Ullmann & Budel (2003) review the extensive literature on fog and dew inputs to BSC communities in the Namib).

Scanning Electron Microscope (SEM)

Figs 4.3 a - e. are SEM images taken of cyanobacterial soil crust harvested in situ from the Kalahari Sand site on 07/07/09. These images are thought to be unique, for any previous literature detailing SEM analysis on BSCs generally uses artificial, laboratory environmental chamber cultivated soil crusts.

Figs. 4.3 a-c show a coarse scale overview of the sample. They clearly illustrate the cyanobacterial strands that bind together the quartz grains, thus playing an instrumental role in maintaining the structural integrity of BSCs. Fig 4.3d, and to a lesser extent figures 4.3a, b and c seem to indicate that the cyanobacterial strands stabilise the soil by enveloping quartz grains, thus furthering stability by forming a cohesive 'mat' around the quartz grains.

The samples were analysed on 02/11/09, a little less than four months after being harvested. It is therefore reasonable to expect that the cyanobacteria would be largely dehydrated, with severely reduced exopolysaccharide content. However, 4.3e plainly demonstrates persistent adhesion of the quartz grain to the cyanobacterial sheath, despite desiccation. The binding role of cyanobacterial strands is a rich area for further study.

CO2 flux

Fig. 4.4 illustrates the ambient atmospheric CO2 concentration. Each value represents the mean CO2 concentration for all six cells at the start of each 45 min sampling period; essentially a proxy for background CO2 concentration.

It is apparent that there is significant variation over time. It is evident that the KS site shows more variation over time than the Calcrete site. It also appears that the KS site varies diurnally. This would correspond with observations that the KS site was located in close proximity to an Acacia mellifera shrub, which is likely to have contributed to any CO2 values. Vascular plant populations were not present in such proximity on the calcrete site.

The calculated trend line appears to decline exponentially. The observed decline is reasonable, given the precipitation data for the two months prior to the study. Following the relatively high precipitation recorded in June, the continued elevated background CO2 most likely reflects a CO2 efflux associated with the most recent precipitation event. This supports the modelled decline in background CO2 over the study period, during which no precipitation was recorded. The CO2 level appears to be levelling at 0.045 by around day nine (225 hours) of the research, which represents an expected background level during the dry season.

Surface CO2 flux: Kalahari Sands

Fig 4.5 illustrates the hourly mean for CO2 flux for all six experimental cells at the Kalahari site, over the course of the experimental period. Positive values represent efflux (CO2 respired to the atmosphere), whilst negative values represent sequestration (CO2 drawn down). Photosynthetically Active Radiation (PAR) is also shown.

Four values (around 24, 48, 72 and 120 hours) show significantly raised standard deviations. They have each been skewed by an abnormal value (all from different cells). The anomalous CO2 samples at these points were re-analysed by Gas Chromotography, yet were still abnormal. They most likely represent operator error, as they are too skewed to be conceivably natural. Though included for the purposes of this graph, they have been excluded from statistical calculations.

The mean should, in theory, be a poor reflection of the fluxes recorded at the cells, given that Cell 1 was the control and had no simulated precipitation. However, as Figures. 4.6a - e show, the elevated antecedent soil moisture clearly restricted the ability to observe the effects of simulated precipitation events. As such, no significant difference was found under a Repeated-Measures ANOVA between Cell 1 flux and experimental cell fluxes. Whilst the ANOVA shows no difference between the experimental cells (and thus implies both low variability and highly replicable methodology), that no difference can be established between the control and experimental cells is a very unexpected result.

Some general observations can be made. All experimental cells (bar Cell 6, Fig 4.6e) show a strong period of sequestration at 48 hours (Cell 2 being much higher than the others). This is surprising, as given that this was night-time photosynthesis would not be expected. There is definite evidence of diurnal cycling, with most periods of net efflux occurring during daylight, from around 10:00 till 16:00. Qualitatively, the graphs seem to indicate a degree of variability between cells (the control cell is not shown, as no difference can be established between the control and experimental cells), though a Repeated-Measures ANOVA concludes that no significant difference can be established between any of the cells. Variance thus cannot be statistically established on the Kalahari Sand site for the experimental cells. However, the observed fluxes contrast strongly with those of Thomas et al (2008), as no marked CO2 pulses can be identified after simulated precipitation events. Whilst this may not have been expected at 0.5mm or 2mm, it is revealing that no sizeably bigger CO2 pulses can be identified after the larger simulated precipitation events. This again implies that the antecedent moisture affected the ability to employ simulated precipitation events as an independent parameter.

This is reiterated through comparison with Fig. 4.7, which compares Cell 1 (the control) with a mean value for all experimental cells. This averaging is justified given the lack of statistically significant difference observed between the experimental cells.

There are extended periods of relative parity in response (though Cell 1 is more exaggerated) between control and experimental fluxes; notably, from 30 to 60 hours and 78 to 90. However, whilst at 48 hours both experimental and control cells show carbon sequestration, the control cell net carbon gain is almost double that of the experimental cells. The second notable point of departure, around 68 hours, again illustrates control cell net carbon sequestration, yet net experimental cell carbon efflux. The reverse is true at approximately 120 hours, where experimental cells show net sequestration, and the control cell net efflux. The potential reasons for this are considered in Chapter 5.

On average, the Kalahari Site experienced net CO2 efflux on all except the second day, as shown in Fig 4.8 (the mean daily flux for all six cells). Whilst the simulated precipitation may be activating cyanobacterial photosynthesis, any net sequestration appears to be being countered by CO2 respired from both surface and subsurface sources.

One particular area of interest is the relationship between CO2 flux and temperature. It was hypothesised that a strong association would exist between the two variables, given cyanobacterial activity's dependence on a threshold temperature to be metabolically active (Lange 2003a). Fig. 4.9 illustrates the relationship between the two.

If previous findings by Lange (2003a) were to be replicated, it would be expected that photosynthesis, and thus carbon sequestration would increase as a function of increased temperature. However, there is no statistically significant correlation between flux and temperature. A general observation can be made that efflux seems to increase in a loose association with temperature, though this could quite likely be a reflection of subsurface microbial respiration diffusing to the surface, masking any cyanobacterial photosynthetic effects. Disentangling the influence of the two remains a very real difficulty.

Surface CO2 flux: Calcrete Pan

Fig 4.10 demonstrates tight diurnal coupling, with each value having a relatively low standard deviation in comparison with KS. The two exceptions are around 48 hours, where larger standard deviations and an unexpected increase in efflux can be considered abnormal. However, field notes at this sampling point record difficulty with the instrumentation, meaning that the last sampling point before 48 hours (the 14th) was extended by approximately 18 minutes. This would explain the observed increase in efflux, due to the unintended longer sampling time period likely increasing CO2 concentration in the soil chambers.

As with the KS, a mean has been used for all six cells. This is justified through consideration of Fig 4.11, which shows the individual flux for all six cells.

As Fig. 4.11 indicates, there is a relatively strong association between all cells. Of most interest is that no significant difference between cells can be identified in a Repeated-Measures ANOVA. As such, variation cannot be established between the experimental cells, and no difference can be found between the control and experimental cells. Whilst this suggests confidence in the methods used and their replicability, the lack of difference between experimental and control cells emphasises the challenges of using simulated precipitation as an experimental variable due to the antecedent moisture conditions.

As Figs. 4.12a - e demonstrate, there are clear diurnal fluctuations in flux, which is in line with pre-existing literature. Peak efflux stays largely constant throughout the period (dropping slightly in all cells on the 20mm simulation day). This is suggestive of higher photosynthetic activity due to greater moisture availability dampening subsoil CO2 efflux. Peak flux recovers on the sixth day, when no additional water was applied.

A revealing comparison can therefore be made between the control and experimental cells, as shown on Fig. 4.13. The mean of all experimental cells shows close parity to the control cell for the first three days of the experimental period. However, the association weakens as the experiment progresses.

Notable deviations include the two periods of marked sequestration for the experimental cells on the fourth and fifth day. This is not replicated by the control cell, and to a large extent this is in line with the theory that greater simulated precipitation should induce photosynthetic activity in the experimental cells, thus drawing down carbon. The association appears then to be reinstated on the sixth day, where no precipitation is simulated (and thus photosynthesis dampened).

Fig 4.14 shows the daily mean flux for all six cells, indicating a net efflux every day (although almost neutral on Day 5). This largely reinforces theory that flux will decline with increased precipitation, due to the increased carbon uptake by photosynthetic cyanobacteria. The marked increase in flux on Day 6 may reflect a time lag of the effects of the infiltration of the 20mm water added on the previous day, thus increasing subsoil respiration.

The relationship between temperature and flux (for all cells) is illustrated in Fig.4.15, which shows a strong association between the two variables. A Pearson test shows a positive (0.652) correlation, which is significant at 99%. One aspect is of particular interest. At the three sampling points around 72 hours (night-time), the temperature is over double (10.7°C) that of previous nights at the same sampling times (a mean of 4.7°C). The increased efflux seen in the last sample on the third day may well reflect higher respiratory activity due to these unexpectedly higher temperatures.

Surface CO2 flux: Comparison between Kalahari Sand and Calcrete sites

Given that no precipitation was simulated on the control cells, flux would not be expected to vary significantly either between days or sites. As Fig. 4.16 shows, however, there is great discordance between the sites.

A T-test indicates that statistically significant difference can be established between the two, at 99.9% confidence. Given that all other conditions remained equal, this implies high flux variability between the two study sites. The Calcrete site shows relatively replicable and consistent diurnal flux cycling, with similar peak fluxes and only slightly differing peak sequestration. KS, however, is far more variable, suggesting a highly complex mechanism driving flux, given that all other environmental factors stay relatively constant. Peak efflux is subdued in comparison with calcrete, with notably greater sequestration over the six day period.

Fig 4.17 compares the KS and Calcrete mean fluxes for the experimental cells. For the first two experimental days, there is a slight association between the two. Whilst the Calcrete site shows higher levels of respiration, there is similarity in the direction of trend.

However, as the level of simulated precipitation increases on both sites, disparity increases between flux activity. Calcrete still maintains a degree of diurnal association, yet KS seems to lose any association. In some respects, the disassociation that begins at around 48 hours is because the Calcrete site shows an unexpected continued rise in respiration at the end of the second day. Were a decrease to have occurred, then association may have been maintained for slightly longer. However, the highly varied responses in the last three days are most likely a reflection of exaggerated activity due to higher simulated precipitation. Overall, Fig. 4.17 does strongly imply that there is high variation in surface soil CO2 flux over a small spatial scale, although no statistically significant difference could be established between the two.

Subsoil CO2 flux: Kalahari Sands

Fig 4.18 shows KS soil pit moisture over the course of the 6-day experimental period (water equivalent to a 50mm precipitation event was added, using a bung, at 07:00 on the second experimental day). All measurements are taken at 13:00.

Prior to wetting, moisture conditions increase slightly with depth, showing a sharp decline below 60cm depth; this depth appears to be a threshold for meaningful infiltration. Infiltration progresses over the course of the 6 days, with moisture peaking in the top half of the soil profile on the third day.

Fig 4.19 shows temperature variation at the same site (all measurements taken at 13.00). Whilst temperature decreases with depth to around 30cm (due to reduced radiative heating from insolation), it largely balances out beyond this depth. Day 2 shows a marked deviation, though given that this was the wetting day, reduced soil temperature is a reasonable observation.

Fig. 4.20 illustrates soil pore CO2 concentration on Day 1, the dry day. CO2 concentration consistently increases with depth (possibly a reflection of increased Soil Organic Carbon).

The particularly large increase between 30cm and 75cm is most likely a reflection of the increased proximity to vascular plant root networks. The noted CO2 concentration increase with depth is temporally stable, with slightly higher concentrations at earlier sample times at all depths (the 75cm concentration at 19:15 is most likely an anomaly).

Fig 4.21 demonstrates CO2 concentration after the simulated precipitation event. Note the immediate rise in CO2 concentration at 27 hours, at 15cm. This is an expression of increased microbial respiration, although is only observed at 6cm and 15cm (possibly due to inadequate infiltration and temperature limitation). At 30 hours, increased CO2 concentration values are still found at the same depths, though it is harder to demarcate, with further time, significant differences between CO2 concentrations. That the highest CO2 values at 75cm are found midway through the experiment is also reasonable, as by this time there would have been adequate infiltration to stimulate respiration. Generally, though small, a CO2 pulse can be noted following the wetting event, though it is smaller than in previous literature (Thomas et al 2008).

A number of statistical analyses were performed on the subsoil data.

The moderate to strong negative correlation between depth and moisture (-0.572) is expected. However, a similarly moderate negative correlation (-0.508) between moisture and CO2 concentration is surprising. The very strong positive correlation (0.919) between depth and CO2 is equally striking. Similarly, no significant correlation between soil temperature and CO2 concentration is an unpredicted outcome. The weak negative correlation between soil moisture and temperature is most likely explained by the confounding influence of surface evaporation.

A One-Way ANOVA and a Post-Hoc Tukey test also indicate that CO2 concentration varies significantly with depth. Whilst no significant difference can be found between values at 6cm and 15cm, significant differences can be found between all other groups. Again, 75cm is significantly different from all other depths at 95% confidence, further emphasising the likely influence of vascular root respiration significantly increasing subsoil pore CO2 concentration.

Subsoil CO2 flux: Calcrete Pan

Soil moisture variation with depth and time at the calcrete soil pit is shown in Fig. 4.22 (the same wetting event was applied on day two as in the KS site).

The immediate rise in moisture post wetting (Day 2) is noticeable, and this quickly subsides with time. In general, moisture levels are higher than KS; whilst KS has a peak moisture of 9.6%, the upper profile on the Calcrete is 17.6% after wetting. There is also greater disparity at depth at the Calcrete site, with the high value at 70cm testament to the greater infiltration capacity. Generally, the calcrete site has more variable soil moisture, both with depth and time.

Fig 4.23 shows soil temperature variance with depth and time. The findings are very similar to the KS site, although Day 2 is an exception. The upper profile values are almost double those of the previous dry day, and the simulated precipitation event does not seem to have induced any cooling.

Fig 4.24 shows CO2 concentration variance with depth/time on the dry day. The absolute CO2 concentrations are very similar to those observed at the KS site. The observed results are largely consistent with theory (Luo & Zhou 2006), showing CO2 concentration increasing throughout the day, reflecting the warmer soil temperatures. This then subsides in the afternoon. There is a noticeable rise in CO2 concentration with depth (as at the KS site), which is likely to be a reflection of vascular plant root respiration.

Fig 4.25 illustrates CO2 concentration variance with depth and time after wetting. Generally, the effects of a pulse event can be observed, with CO2 concentration largely increasing with time at the upper profiles. However, such a clear relationship is not found at 30cm and 75cm. More regular sampling was possible than on the KS site, as the Calcrete soil pit had greater structural integrity. The increase in CO2 concentration at 15cm at 24 hours is similar to that seen in KS at 27 hours at the same depth, indicative of heightened microbial respiration.

Post-hoc tests on a One-Way ANOVA indicate that there is a significant difference (95% confidence) between CO2 concentrations at all depths at all times. This compares with KS, where no significant difference was found between 6cm and 15cm CO2 concentrations. The ANOVA results support the finding that depth and CO2 concentration are positively correlated (0.883, 99% confidence) over all six days.

A number of statistical tests for correlation were undertaken, as shown in Table 4.2:

No correlation was found between soil moisture and temperature, although depth and soil moisture are moderately negatively correlated (-0.419). A surprising negative correlation (-0.452,) was identified between CO2 concentration and moisture, which goes against theory that increased moisture will stimulate higher microbial respiration, and therefore higher CO2 concentration.

Chapter 5. Discussion

The CO2 fluxes observed from both the KS and Calcrete sites do not appear to support the 'pulse-reserve' paradigm. Simulated precipitation had little discernible effect on peak flux, irrespective of the total precipitation amount. This study (excluding anomalous values), reports peak flux at KS of a little over 50 mg C m2 h1, observed on both the 2mm and 20mm simulations. At the Calcrete site the results are lower, the flux over the course of the six days peaking at slightly under 40 mg C m2 h1. This contrasts with the findings of Thomas & Hoon (2010), who report a peak flux from a similar Kalahari Sand site in July 2007 of 339.2 mg C m2 h1. Whilst this was recorded after simulated rainfall of 120mm, higher than in this study, fluxes of 65.6 mg C m2 h1 were observed after just 1.4mm application. In a wider context, the positive effluxes observed at both the KS and Calcrete sites are noticeably lower than those reported in Raich & Tufekcioglu's (2000) global analysis. The lower flux is probably a reflection in part of the limited organic content of the dryland soil, as suggested by Thomas & Hoon (2010). As such, the lower fluxes observed in this study are highly indicative of an ecosystem that is limited by carbon, rather than moisture availability.

This theory is supported by the exponentially declining values recorded for the ambient atmospheric CO2 (see Fig. 4.4) which are in line with the notion that the study site was recovering from a large pulse event, most likely associated with the 34.4mm rainstorm of 10/06/09. At both sites, flux does not appear to vary as a function of simulated precipitation amount. It would justifiably be assumed that a significant difference could be reliably established between the control and experimental cells, yet none exists (as shown in Figs. 4.7 and 4.13). Whilst Thomas et al (2008), Thomas & Hoon (2010) and Wang et al (2009) suggest that CO2 fluxes in Kalahari soils are likely to be moisture limited for the majority of the year, the same conclusion cannot be supported on the basis of this study's findings. Rather, the highly elevated antecedent moisture conditions seem to have undermined the efficacy of employing simulated precipitation events as an independent experimental parameter. Though the ecosystem is usually moisture limited in June, it appeared to be largely carbon limited in June 2009 owing to high rainfall prior to experimentation.

Given the low vascular plant populations, the BSCs are likely to be the only significant source of carbon at the soil surface. Belnap et al. (2003b) report that C fixation by cyanobacteria during photosynthesis can result in soil organic carbon increases of up to 300%, being secreted as extracellular polymeric substances. Whilst Evans & Lange (2003) recognise the potentially significant contribution of BSCs to carbon stores in drylands, the Kalahari remains notable for its relative lack of data. The antecedent moisture conditions following the rainfall prior to this study seem to indicate, however, that the soil carbon stores had been largely exhausted prior to experimentation. This would partially explain why few pulses were observed following the application of water; certainly, it is not possible to differentiate between different magnitudes of precipitation events. This does not suggest that the pulse-reserve paradigm is invalid, but raises the likelihood of the Kalahari being both moisture and carbon limited. In many respects therefore, whilst the study failed to satisfy its intention of demonstrating precipitation event driven CO2 fluxes, it gives an insight into the impact of low frequency, high magnitude unseasonal storm events on carbon cycling.

A key focus of the study was to examine variability both within sites and between the two sites. The findings of the T-Test show that the six-day flux results for the control cells at each site are significantly (99.9% confidence) different. This is worth some consideration. As all other conditions remained equal, the scale of the disparity seen between the two is striking. Numerous environmental variables, both surface and subsurface, are likely to have contributed towards this. Certainly, the proximity of the KS site to a sizeable Acacia Mellifera plant may have influenced samples. As Dougill & Thomas (2004) note, crusts develop preferentially under this plant, so C exchange is unlikely to be spatially even. This has particular implications given the recent bush encroachment recorded in the Kalahari (Dougill 2002). The root network is likely to have impacted subsoil pore CO2 values, and surface vascular CO2 exchange is again likely to have affected sampling. Field observations also suggest that the crusts at the two sites were in a different stage of succession. Despite the fencing at the KS site, ensuring that the crusts were protected for the year prior to study, they were largely destroyed in summer 2008 during research (Thomas 2009, pers comm). However, although the Calcrete site was not protected, pastoral farming had stopped on the site approximately 20 months prior to this investigation. There were large areas of clearly intact, Stage 3 crust (with noticeable surface roughening and microtopography), whilst only Stage 2 crust was identified at KS. The influence of crust succession on its role as an atmospheric boundary has been noted (Housman et al 2006; Zaady et al, 2000), though Zaady et al (2000) suggest that higher level of succession may lead to higher flux rates. This suggestion was not borne out in this study (mean flux was higher at the KS site), where more consistent fluxes were observed on the Calcrete site between all cells. This finding is in line with the theory that greater homogeneity in BSC surface crust may lead to more replicable CO2 exchange (Housman et al, 2006). Equally, the effect of the Calcrete site's more consistent BSC surface coverage on water infiltration should not be overlooked. On the 5mm and 20mm simulation days, infiltration varied greatly between cells on the KS site, yet no such variation was noted at the Calcrete site. Indeed, infiltration took longer at the Calcrete site, which is potentially a reflection of both the more developed BSC population and the significantly higher antecedent soil moisture (pre-wetting, 8.5% as compared to KS 3.5%). It is feasible that the consequently more uniform infiltration may partially explain the more consistent observed fluxes between cells. However, this is only a partial explanation, for the influence of simulated precipitation pulses is likely to have been reduced by the high antecedent soil moisture.

Whilst the pulses of positive CO2 flux may be less than expected, some interesting observations can be made on negative flux (carbon uptake). At the Calcrete site, some differentiation can be established between the control and experimental cells. Whilst negative flux does occur at the start of Day 3 for the control cell (-9.7 mg CO2 m2 h1), it is less than half the mean experimental negative flux (-20.7 mg CO2 m2 h1). However, field observations for this day note significant dewfall that night, as supported by the temperature/dew point data. Whilst not quantified, it is likely that sufficient dew was present to establish low metabolic activity for the control cell, thus explaining the C uptake. Considering the experimental cells, some negative flux may have been expected on the second day (0.5mm event), for Lange (2003a) notes that optimal photosynthesis, and thus CO2 fixation, can occur with 0.5mm rehydration (though this is highly species dependent). The fact that the negative flux is recorded immediately after wetting again confirms cyanobacteria's rapid metabolic response to hydration, as suggested by Garcia-Pichel & Pringault (2001). Certainly, the stronger negative flux for the experimental rather than control cells at 2mm rehydration implies photosynthetic reactivation. Considering experimental Day 5 (20mm event), another significant negative flux can be identified. As Fig. 4.12e reiterates, this flux is remarkably consistent (both over time and in magnitude) between all experimental cells. This finding is testament to the high temporal resolution of the sampling methodology developed by Hoon et al (2009). Such subtlety of flux would not be apparent in the studies of Wang et al (2007, 2009) and others, which use one flux measurement as representative for the whole day.

The high sequestration counters previous results found by Thomas et al (2008). Generally, at higher wetting events, any increased photosynthesis which may lead to a CO2 drawdown is masked by increased subsoil respiration. This does not appear to be the case in this study, strengthening arguments that the ecosystem may be carbon exhausted. Equally, the strength of the CO2 sequestration is surprising, for Lange (2003b) reports photosynthetic inhibition at suprasaturation. However, Lange's work is largely focused on lichenous BSCs; our awareness of species differentiated responses is at best currently limited, and further research is needed to determine optimal hydration for photosynthesis for the cyanobacteria at this study site.

The KS site shows more variable periods of CO2 uptake, and uptake is less consistent between cells. Most surprisingly, uptake appears to occur at all experimental cells just before 48 hours. The sampling point was at 22:00, so photosynthetic activity is not an obvious explanation. Had just one cell experienced sequestration then this observation could perhaps be dismissed, yet given that it is replicated over five cells it is worth some consideration. Mairs (2009, pers comm) reports in situ results from the Kalahari, where BSC cyanobacteria were artificially darkened, and wetted with a 5mm event every other day. Using IR Spectrometry, it was found that after six days surface chlorophyll content actually increased in darkened conditions, explained by the ability of chlorophyll to glide vertically within the soil column on EPS sheaths. It is possible that the experimental ISCC used at the KS site may have limited photosynthetically active radiation to the BSC, due both to shading from the Acacia Mellifera and shading from the ISCC themselves. Consequently, a time-lag may have occurred in sequestration. However, the fact that the control cell shows significant sequestration at 22:00 on the second and 19:00 on the third experimental day (-61.6 mg CO2 m2 h1 and -76.4 mg CO2 m2 h1 respectively) further complicates the relationship. As all other conditions remain equal, there is no obvious theory to explain this. It is possible that light independent (dark) photosynthesis may be occurring; Belnap (2003) discusses nitrogen fixation in cyanobacterial BSC, where under laboratory conditions dark fixation has been observed. Equally, Stone (2008) cites episodes of night-time CO2 uptake in China's Gurbantunggut Desert. This is largely attributed to the high alkalinity of the soil, yet the pH at the KS site was 5.9, so this theory is unlikely to be applicable in this instance. This is an area rich for future research, as our understanding is very much in its infancy.

Over the course of the six days, both of the study's sites show net efflux. The six cell daily mean (Fig 4.8) shows net CO2 uptake on the second day at the KS site, although this is skewed by the unexpectedly high uptake of the control cell. However, the variance at the KS site as shown by the error bars, is much higher than the Calcrete site, so any conclusions should be tentative. The Calcrete site (Fig 4.14) shows a decline in the mean net flux from the second to fifth experimental day, which is consistent with theory that increased moisture will lead to increased photosynthetic uptake. As such, the rise in the sixth day may well be explained by the 20mm event being sufficiently large to infiltrate to a depth whereby it could cause subsoil respiration, and thus a larger efflux the following day (by which time surface moisture would have declined to the extent that it might limit photosynthesis). Considering the overall sequestration at both the KS and Calcrete sites, there is a contrast to the work of Wohlfahrt et al (2008), who report net annual CO2 uptake in the Mojave. However, there is likely to be seasonal variance in flux activity, given the significant range in both temperature and moisture availability, so any conclusions from this study are by no means representative.

The environmental variables governing BSC metabolism and soil respiration have been widely studied (Belnap & Lange, 2003). Whilst this study recorded Photosynthetically Active Radiation (PAR), equipment failure resulted in data loss for the entire KS site. As such, a proxy was used, taking the PAR measurements from mid-July 2008. Consequently, it would be unreasonable to undertake considerable analysis of relationships between PAR and respiratory/photosynthetic activity. However, the relationship between temperature and flux activity is of interest. Though Lange (2003a) discusses examples (largely laboratory based) where there is strong coupling of CO2 flux and crust temperature, Thomas & Hoon (2010) argue that in a carbon limited ecosystem, there is unlikely to be a positive correlation between crust CO2 flux and temperature. Thomas & Hoon therefore imply that the relative importance of both moisture and temperature are likely to be dependent on antecedent conditions (Thomas et al 2008; Thomas & Hoon 2010). In this light, the results found in this study are rather surprising. Given the sealed nature of the experimental cells, concern has been raised as to their ability to maintain ambient temperature, though Thomas (2009, pers. comm.) finds a very strong positive correlation between cell and ambient temperature. Whilst the KS site appears to support Thomas & Hoon's (2010) theory, given that there is no significant correlation between the two, the Calcrete site shows a 0.652 correlation, significant at 99%. Such a difference over what is a small spatial scale emphasises the variability between the two sites, and also stresses the importance of crust succession in governing response to environment