Sap Flow Dynamics and Responses to Grazing and Seasonal Changes in Soil Moisture for Acacia Ancistroclada and Comberatum Molle in a Kenyan Savanna

The Kenyan savanna, which is dominated by Acacia ancistroclada and Comberatum molle, has experienced notable changes in rainfall patterns and increased livestock grazing. A signicant decrease in trees spread from 5 % to less than 1 % has been documented for the ecosystem and could be linked to the increased livestock grazing and changes in rainfall patterns, however, scientic evidence is lacking. We utilized sap ow to analyze the hydraulic responses of the prevailing trees to livestock grazing and seasonal changes in soil moisture. Environmental factors including precipitation, air temperature, soil moisture at - 0.3 m, and vapor pressure decit were simultaneously measured. The results showed that the diurnal variation in sap ux density exhibited a single peak curve at around midday and correlated strongly with vapor pressure decit and air temperature. Sap ux density was higher in the grazed (27.47 ± 8.65 g m -2 s -1 ) than the fenced plots (20.17 ± 7.27 g m -2 s -1 ). In all the plots, sap ux density followed seasonality in rainfall patterns, increasing and decreasing in wet and dry seasons respectively. The higher crown projected area was responsible for higher sap ow in the grazed plots. The diurnal variation in sap ux density showed that sap ow was coupled to the atmosphere with relatively low boundary layer resistance and the seasonal variation in sap ow was controlled by stomatal regulation. These ndings point to the possibility that the dominant tree species in Lambwe are isohydric species. However, additional measurements need to be conducted on the eligibility of the species to conrm the conclusion.


Introduction
The structure and function of savannas, especially in Africa are undergoing modi cation which may be  et al., 1998;Plaut et al., 2013). The in uence of soil moisture on sap ow may be related to multiple factors such as tree species (Cheng et al., 2006), rainfall amount and other environmental factors (Ivans et al., 2006). For example, deep rooted Australian Eucalyptus miniata and Eucalyptus tetrodonta exhibited increased sap ow during the dry seasons compared to the wet season and this was linked to increased evaporative demand and exploitation of the belowground water aquifers by the tree roots (O'Grady et al., 1998). On the other hand, shallow rooted Isopogon gardneri exhibited increased sap ow in response to improved moisture levels in Southern Australia (Burgess, 2006), a tendency which was linked to e cient use of available soil moisture. Trees with broad leaves show the greatest increase in sap ow with increasing vapor pressure de cit, while the narrow leaved show the lowest sap ow (Komatsu et al., 2007).
Leaf area and stomatal regulation may account for the varying behavior in response to the changes in environmental and water conditions (Solari et al., 2006;Rodriguez-Gamir et al., 2016).
Based on the stomatal regulation of sap ow, savanna trees may be categorized as either isohydric or anisohydric species (Roman et al., 2015;Zhang et al., 2017). Isohydric species reduce stomatal conductance and sap ow as soil water decreases, hence maintaining constant leaf water potential (Roman et al., 2015). However, a consequence of this strategy is that the trees shut their stomata in response to even mild water stress, a process which though reduces tree water loss, often results in carbon starvation (Cheng, et al., 2014;Sankaran, 2019). In contrast, anisohydric species allow large uctuations in leaf water potential, maintaining high stomatal conductance and sap ow even during periods of water stress (Kinings and Gentine, 2017). This strategy incurs a risk of xylem cavitation, which could ultimately lead to a decline in leaf water supply and tree mortality ( Savannah trees have evolved with grazing, predominantly by wild herbivores, that normally occur at an ecologically sustainable level (Kioko and Seno, 2012;Mureithi et al., 2014;Mojeremane, 2015). However, with expanding livestock numbers, the ecosystems have experienced increased grazing pressure, which likely in uences soil water in ltration and tree water transport (Sankaran 2019). At high grazing intensities and on steep slopes, prolonged trampling by grazers may result in surface runoffs and reduced soil water in ltration (DeLonge and Basche, 2018), thus reducing the water reaching the root zones of trees. Such conditions induce tree water stress and result in reduced sap ow and increased tree mortalities.
Nevertheless, most savanna tree species have well developed deep rooting systems that access below ground water reservoirs (Miller et al., 2010) and can, therefore, overcome the induced water stress. Alternatively, grazing may increase the availability of soil water by reducing competition from the herbaceous layer, as suggested by models of savanna tree-grass coexistence (Sankaran et al., 2004;Baudena et al., 2010). In such scenarios, grazing would, therefore, contribute to the maintenance of water transport in savannah trees.
The Kenyan savanna located in Lambwe Valley has experienced signi cant changes in rainfall patterns and land use over the years (Njoka et al., 2003;Muriuki et al., 2005). The ecosystem has witnessed increased livestock grazing, increased drought periods, and short episodes of intense rainfall (Otieno et al., 2011;Ondier et al., 2019). The changing rainfall patterns might be modifying soil moisture levels with implications on tree survival and water transport. Reports of declining forest cover from 5% to less than 1%, documented for the region (Njoka et al., 2003;Scharsich et al., 2019), could be linked to the changing environmental factors, however, scienti c evidence is lacking. This study used sap ow to investigate the responses of dominant tree species to the increased livestock grazing and seasonal changes in soil moisture. The understanding of water use strategies of the savanna trees under the environmental change scenario will help in decision making aimed at conserving the savannah and other similar ecosystems. were utilized for estimation of sap ux density SFD (g m -2 s -1 ) using uniquely designed, heat dissemination tests developed after Granier's (1987) design. Each sensor consisted of a pair of thermistor of 2-mm-diameter which was vertically aligned ca. 10 cm apart. The two thermocouples were joined at the constantan leads, so that the voltage measured across copper leads provided the temperature difference between the heated probe and the lower reference. The data loggers and the sensors were powered by 120 mA storage cells (Grainer 1987).

Materials And Methods
Sensors were placed in the outer 20 mm of the sap wood (annulus, 1, 0-20 mm radial sapwood depth). In cases where the tree trunk were large with sapwood radius greater than 20 mm, a second sensor was placed 10-15 cm circumferentially, away from the rst sensor pair, on the same side of the stem to avoid azimuth differences. Temperature differences were measured every 5 minutes and 30 minutes mean value logged (DL2e with LAC-1 in single ended mode, Delta-T Devices, England). Sap ux density (SFD, g m -2 s -1 ) for every sensor was determined from T in accordance with Grainer (1987) assuming zero SFD (i.e. T max ) at night and VPD close to zero:

Tree allometrics
The sample trees were selected according to species and tree size distribution within the experimental plots. To estimate the sap wood area (SA) of the sample trees, an increment borer was used to extract cores of sapwood at the sensor installation height (about 1.5 m height) on the same species, but different trees from those installed with sap ow sensors. The depth of the sap wood depth was visually determined at those cores, as wood and heartwood were clearly different. Sap wood area was determined from sapwood depth and tree DBH based on equation ( Where α is a constant and β is allometric scaling exponent, and both are species-speci c coe cients. The ground projected crown area (m 2 ) of sample trees was measured in eight horizontal directions using compass, crown mirror and a measuring tape. The octagonal area was determined as the whole of eight triangles (Schmidt, 2007).

Soil water content
Within the grazed and fenced plots, continuous SWC in the upper 30-cm soil layer was monitored using Theta probes (type ML 2X, Delta-T Devices, Cambridge, UK). Data were collected every 30 min, averaged, and logged hourly using a data logger (DL 2e-Delta-T Devices, Cam-bridge, UK).

Analysis of Plant Root Distribution
Analysis of plant root distribution was done by excavating the selected soil pro les down to a depth of about 1 m or a depth where concrete rock material was found. Soil pro les (~0.5 x 1 m) down to a depth of consolidated rock material were dug within grazed and fenced plots at locations under the tree canopies. Each pro le was divided into individual soil horizons and subsequently analyzed for depth distributions of roots. Coarse (>2 mm diameter) and ne (<2 mm diameter) root densities for each horizon were determined by counting all roots and dividing the number by the area of the pro le wall covered by the associated horizon.

Stomatal conductance
Monthly measurements of diurnal courses of stomatal conductance was carried out using steady state porometer, model LI-1600, LI COR, USA; with a cuvette for attaching the studied leaves. On each measurement day, three different trees of the same species were selected for measurement. Well exposed leaves were identi ed from each of the tree, and measurements conducted on the same leaves between 08:00 and 17:00 local time to avoid wet canopy conditions caused by morning and late evening dew (leading to high values of g s ). Tree crowns were accessed using a 3 m high ladder. Care was taken not to damage the leaves during measurements by using a soft air tight pad placed over the mouth of the cuvette. Each leaf was mounted in the cuvette so that it received full illumination during measurement.
Every turn of measurement lasted one to two minutes to allow the chamber to equilibrate. Alternate measurements were conducted between species throughout the day.
Statistical analysis SFD and environmental variables were recorded as half hourly average values. These variables, including Gc were converted into daily and monthly averages. Data are presented as mean ± standard deviation (SD). SFD, T air , VPD and Gc were compared between grazed and fenced plots, within months and among tree species using one way anova. Where differences were found among species, a post hoc test was carried out. All statistical analysis were based on 0.05 signi cance level and performed with SAS version 9.1 (California, USA).

Microclimate of the study site
The total rainfall amount in 2014 and 2015 were 1148.4 mm and 1169.5 mm respectively. Rainfall was bimodal, occurring in April to June and September to December. On average, monthly air temperature was 22.95±4 0 C and 22.35±3.2 0 C in 2014 and 2015 respectively. The hottest month was February while the coldest month was July (Fig. 1).

Soil water content within Grazed and Fenced plots
In all the plots, soil water content followed seasonality in rainfall pattern, increasing and decreasing during wet and dry seasons respectively. April-June and September-December were wet periods while January-March and July and August were dry periods (Fig.2).

Root distribution patterns
The depth of soil on which our plots were established ranged between 50 and 70 cm below which there was a strong presence of course rock fragments. Root distribution patterns were not different between the two locations (Fig). In all the plots, the densities of ne roots were highest at around 20 cm depth, most of which belonged to grass. Between 30 and 70 cm depths, coarse roots belonging to trees signi cantly increased.
Characteristics of the studied trees Table 1

Stomatal conductance variations
Monthly variations in leaf stomatal conductance and relationship between conductance and soil water content are shown in gure 7. Stomatal conductance increased and decreased during wet and dry periods irrespective of plot and species (Fig 7 a and b). There was a positive correlation between stomatal conductance and soil water content, with conductance increasing linearly with soil water content (Fig. 7 c).

Sap ow characteristics
This study demonstrated diurnal patterns of sap ux density (SFD) for the tree species within the plots (Fig. 6).
demonstrated a linear relationship between SFD and VPD and revealed that the daily changes of SFD for tree species displayed single peak curves that reached maximum values between 1200hrs and 1400hrs. Our results likewise uncovered positive relationships between SFD and both air temperature and VPD. Further, SFD exhibited distinct curves that reached maximum values at about 1200hrs and 1400 hrs. The maximum SFD values coincided with peak values of both VPD and air temperatures (Fig. 6). The results demonstrate that SFD within our plots is driven by VPD and air temperature. VPD and air temperature are known drivers of sap ow and transpiration attributable to the water potential gradient, and the ability of dry air to pull water from a source of higher water potential (Nadezhdina et al., 2019). Moreover, the correlation with VPD infers that SFD was coupled with the atmosphere with relatively low boundary layer resistance (Gomenez et al., 2019). Similar ndings were reported by Deng et al., (2015) who recorded a gradual increase in tree sap ux density with increasing vapor pressure de cit up to around midday. The linear increase in SFD with VPD was linked to increased evaporative demand that was necessitated by the increased air temperature. According to the report, the peak SFD was only possible when VPD was < 1.4 kPa, beyond which stomatal conductance and transpiration drastically reduced to avoid cavitation. This is true for our experiment where sap ow density peaked when VPD was around 1.0 kPa (Fig. 6 c and d).
Although the dynamics of SFD were comparative in patterns, they were not synchronized. Acacia ancistroclada recorded signi cantly lower sap ow than the other studied species (Fig 4 and 6). This nding could show the uniqueness in habitats to which every species is adopted (Wu et al., 2017). It might also be related to the differences in the anatomical structure of the leaves of individual species. Acacia ancistroclada, which is narrow-leaved compared to the broad-leaved Comberatum molle. Under the same environmental conditions, Acacia is expected to transpire less compared to Comberatum molle due to the limited leaf area exposed to transpiration water loss. This could be the reason for the lower sap ow in Acacia observed within our plots (Fig 4a and b). Moreover, in response to the increasing soil water de cit and atmospheric demand following dry periods within our plots, there was a signi cant reduction of the transpiring leaf surface through shedding (personal observation). This could account for the decline in whole-tree sap ow in Acacia. Gomenez et al., (2019) linked lower transpiration in narrow leaved trees to lower stomatal conductance which was triggered by the accumulation of ABA in the leaves. In our study, however we did not determine the level of aba in the leaves of trees. Further, we could not determine the stomatal conductance of Acacia species due to the limitation of our equipment.
The linear relationship of SFD, with DBH, wood area and the crown projected area reported in our study (table 1) was also documented by other researchers (Kallarackal et al., 2012;Berry et al., 2017). The relationship between crown projected area and sap ow is explained by the pipe model theory (Shinozaki et al., 1964). According to the theory, the crown area and amount of leaves on a tree are supported by a proportionate of the crossectional area of the xylem vessels with equal hydraulic conductance. Consequently, the larger the crown area, the higher the transpiration and sap ow. Additionally, large projected crown areas are linked to higher leaf numbers and higher transpiration rates (Li et al., 2017). The relationship between DBH and sap ow is attributed to the positive relationship between DBH and sapwood area which directly transports sap ow. The report by Kallarackal et al., (2012) that large stems are linked with increased stem water storage and transport capacity agrees with the ndings of our study.
Sap ow responses to livestock grazing and seasonal changes in soil moisture Mean sap ux density was higher in the grazed plots compared to the fenced plots (Table 1). We could link this to a higher mean crown projected area in the grazed (39.69 ±7.24m 2 ) compared to grazed areas (27.28 ±8.64 m 2 ). Higher crown area is linked to a grater transpiring surface and could be the reason for higher SFD (Liu et al., 2017). SFD within our plots was decoupled from soil water content at -0.3 m (Fig  5a and b). This could be expected as tree roots within this study site are concentrated at a depth > 0.4 m (Fig.3). Tree sap ow within our plots followed seasonality in rainfall pattern, increasing and decreasing during wet and dry seasons respectively ( Fig. 4a and b). Such results were also recorded for a seasonally dry and wet forest in Thailand  and were linked to soil moisture stress. For our plots, however, the uctuation in sap ow in response to seasonality in the rainfall pattern was linked to stomatal adjustment (Fig. 7 a and b) to limits water loss during dry seasons and maintain transpiration during wet seasons. Such a phenomenon is common in isohydric tree species. However, it is inherently di cult to establish a rm relationship between sap ow and soil water content mainly because of the large spatial variation in soil properties and soil moisture, but also because of sap ow's strong dependence on other weather parameters.

Conclusions
The higher Sap ux densities in the grazed visa vie fenced plots were explained by the differences in tree cover and crown area. The savanna trees showed a uniform response to the seasonality in the rainfall pattern. They exhibited an increase and decrease in SFD following increased and reduced rainfall which coincided with wet and dry seasons respectively. This is a phenomenon that characterizes isohydric species. Acacia ancistroclada which recorded signi cantly lower sap ow compared to the Comberatum molle could be distinguished as best suited species for survival in Lambwe Valley under prevailing environmental change processes. The species can then be used for reforestation of the ecosystem and other similar areas. However, additional measurements need to be conducted on the eligibility of the species to con rm the conclusion.

Declarations
Ethical approval and consent to participate        Relationship between sap ow and soil water content for studied trees within grazed (A) and fenced (B) plots Figure 6 Diurnal course of mean sap ux density as a function of air temperature in the grazed (A) and fenced plot (B) and vapor pressure de cit in grazed (C) and fenced plot (D) respectively. Data are means±SD