Fire is one of the most important disturbances in forest ecosystems ( Bond & Keeley, 2005 ; Pausas & Keeley, 2009 ). It causes changes in forest structure, biomass, or species composition because of different mechanisms involving adaptive or post fire responses. Some studies, however, have reported the positive effect of fire in certain ecosystems, due to the subsequent increase in seedling density ( Gashaw & Michelsen, 2002 ; Walters et al. , 2004 ).

Forest fires are a constant threat in arid lands because they endanger ecological niches with high biodiversity ( Danthu et al. , 2003 ; Savadongo et al. , 2007). The post-fire response of the vegetation depends on many factors, including the pre-fire condition of the species, environmental and ground temperatures, fire inten­sity, ash concentration, nutrient balance or the animal populations associated with the forest ecosystems ( Gashaw & Michelsen, 2002 ; Núñez et al. , 2003 ).

In arid environments, fire tolerance is identified as one of the main factors that controls the distribution of Acacia species ( Burke, 2006 ). Specifically, different adaptive mechanisms to forest fires have been identified. So, among others, heat-shock triggered germination (e.g. hard seed structures in Fabaceae family make the seed impermeable to water until being scarified by heat or other abrasive agent), or germination triggered by combustion chemicals ( Keeley et al. , 2011 ). On the other hand, seed dormancy, a very common feature in many species of the Fabaceae family, derived in many occasions of seed coat structures, is a strategy to ensure survival in adverse conditions ( Danthu et al. , 2003 ). Seed dormancy can be interrupted by several methods ( Teketay, 1996 ; Baskin & Baskin, 1998 ; Tigabu & Odén, 2001 ) such as heat shock, scarification methods, or ingestion by herbivores or birds ( Miller, 1995 ; Baskin & Baskin, 1998 ; Razanamandranto et al. , 2004 ).

Previous studies have analyzed the effect of heat shock on seed germination in forest species in arid zones of Africa ( Teketay, 1996 ; Sy et al. , 2001 ; Gashaw & Michelsen, 2002 ; Danthu et al. , 2003 ; Zida, 2007 ; Maraghni et al. , 2010 ). In Acacia spp., Teketay (1996) tested different heat shocks (60ºC, 80, 100ºC during 15, 30 and 60 minutes) and showed that once the dormancy in leguminous species with hard seed coats was broken, the seeds germinated in a wide range of temperatures. However, temperatures higher than 100ºC and reduced exposure time was not checked by Teketay (1996) , that could mimic fast fires, where different temperatures affect seeds for a short period of time. This information is essential for evaluating the effect of fire on forest regeneration, and for knowing the adaptive significance of seed dormancy in relation to forest fires in these ecosystems.

The African savannah is a good example to study those arid and fragile ecosystems to present periodic droughts and the increase of the over-exploitation of forest resources ( Malagnoux et al. , 2007 ). In particu­lar, the Sudanese savannah, situated in the northern hemisphere of Africa, is an ecosystem with vast open land and few tree species. It covers 5.25 million km ² , from Senegal to the Ethiopian Highlands. Leguminous species from genus Acacia, Boswellia and Comiphora are representative of this ecosystem ( Danthu et al. , 2003 ; Eriksson et al. , 2003 ; Kulkarni et al. , 2007 ). They provide timber and non-timber products, such as firewood, coal, fodder, wood for fences and agricultural implements, natural gums, frankincense, myrrh and medicines ( Loth et al. , 2005 ). Nonetheless, 25-50% of the area is affected by fire, caused by humans or other natural climate factors ( Savadogo et al. , 2007 ; Zida, 2007 ), and therefore it is important to understand the biological and environmental factors that affect seed germination. This knowledge allows us to ensure recruitment ( Sy et al. , 2001 ; Gashaw & Michelsen, 2002 ), or to respond after wildfire disturbances.

Species in the Aculiferum subgenus of Acacia genus, which includes A. senegal , show a soft seed covering ( Danthu et al. , 2003 ). That means that they are sensitive to high fire intensities and prolonged exposure time. Seeds from species in the second sub­genus, Acacia , which includes A. seyal , have a hard and impermeable outer cover that is resistant to high temperature heat shocks ( Danthu et al. , 2003 ). Acacia senegal L. and A. seyal Del., within this ecosystem, are two socioeconomically important species with contrasting seed coat structures. These differences make them suitable to test the hypothesis that species in arid environments, with thick coats, have deeply dormant seeds and that frequency and type of forest fires (e.g. exposure and intensity), influence differently the dormancy of the seeds depending on the seed coat characteristics. To address this hypothesis, the aim of this paper was to assess the effect of different heat shock treatments (differing in temperature and exposure time) on the germination of the Acacia senegal and A. seyal . The results could provide information of great relevance for sustainable forest management and conservation of the species with better silvicultural practices during the regeneration period or in post-fire regeneration treatments.

Material and Experimental design

Seeds were collected from representative, mature and physically well grown trees in the Rift Valley in Ethiopia in January of 2010. Seeds from A. senegal were collected from Langanno (7º26'N-38º47'E), and A. seyal from the Abijata-Shala National Park (7º32'N - 38º40'E) in stands with a density of 40 trees/ha by personnel of the Ethiopian Environment and Forest Research Institute (EEFRI). The altitude of both areas ranged from 1550 to 1570 m asl ( Argaw et al. , 1999 ). Seeds were kept at a constant temperature (4ºC) in a bag for two months to keep the conditions collected by EEFRI until the time of the experiment.

The seeds were exposed to twelve heat shock treat­ments, consisting of 4 different temperatures (60ºC, 90ºC, 120ºC and 150ºC) and 3 exposure times (1, 5 and 10 minutes). A control, without heat shock treatment, was considered as an additional treatment. Heat shocks were performed in a hot dry oven. So, the experiment consisted of thirteen experimental treatments (4 x 3 heat shock combinations and the control). The range of temperatures chosen is the standard in this type of studies ( Núñez & Calvo, 2000 ). Regular spacing temperatures is common to facilitate the statistical analysis, while the temperature range includes a lot of possibilities ( Trabaud, 1979 ) contrasted with studies of prescribed fires. Similar temperatures were reached in the soil ( Hernando et al. , 2000 ), and in the crown, where the maximum reached temperatures did not exceed the 140ºC ( Vega, 2000 ).

Four replicates (consisting of 25 seeds) for each treatment were established following a randomized complete block design, were the experimental unit was a 9-cm Petri dish with two filter paper discs dampened with sterilized distilled water. In total, the experiment was composed of a total of 2600 seeds (1300 by species and 100 by treatment and species).

The germination test was performed in a germination chamber under constant conditions (30ºC, 80% of re­lative humidity and 12h light/12h dark of photoperiod) for 45 days, following the International Seed Testing Association [ISTA] rules ( Don, 2013 ; ISTA, 2015 ) and adapting for dry areas. In the case of A. seyal seeds, after the application of heat shock, they were immersed in boiling water to interrupt the dormancy for 24 h ( Larsen, 1964 ; Teketay, 1996 ).

Every three days, the germinated seeds were counted and removed and the filter paper was moistened. A seed was considered as germinated when the radicle had emerged at least 2 mm from the testa ( Come, 1970 ). From each Petri dish, accumulative germination was determined as the number of seeds germinated during the entire experiment. The proportion of germinated seeds was defined as the accumulative germination with respect to the total for each combination of block and plot (number of seeds germinated/25 seeds).

At the end of the experiment, the viability of no germinated seeds was assessed by counting live and dead seeds after immersion in a Tetrazolium solution ( Paynter & Dixon, 1990 ). This solution is a biochemical method, which determines seed viability and is used by ISTA.

Statistical analysis

The proportion of germinated seeds was transformed into the arcsine of the square root of the proportion of germinated seeds, which is a very common operation for stabilizing the variance of observed data in experiments of this type ( Núñez & Calvo, 2000 ; Herrero et al. , 2007 ). This transformation was used to correct the statistics' assumptions underlying the normality of the linear model ( Montgomery, 1997 ).

A mixed model was applied in order to detect significant differences in the germination at the end of the experiment as a function of species ( Acacia senegal and Acacia seyal ), heat shock (13 treatments: control and 12 combination of four temperatures and three exposure time) and the interaction between the two factors. There were 52 observations for each species, for a total of 104. These observations corresponded to the four replicate by each heat shock by the two species (4 replicates*13 heat shocks*2 species=104). The mo­del was expressed as follows [Eq. 1]:

where:

y _ijk : Arcsine of the square root of the proportion of germinated seeds by species i with the heat shock j in the replicate k (1 to 4); µ : General mean effect; α _i : The effect of the species factor i (1= A. senegal ; 2= A. seyal ); β _j : The effect of heat shock j , (60ºC_1min; 60ºC_5min; 60ºC_10min; 90ºC_1min; 90ºC_5min; 90ºC_10min; 120ºC_1min; 120ºC_5min; 120ºC_10min; 150ºC_1min; 150ºC_5min, 150ºC_10min; control [30ºC_0]); αβ _ij : The effect of the interaction between species i and heat shock j , and ε _{n (ijk)} : experimental error, where ε _ijk → N (0, σ _i ).

A variance for each species was determined by the mixed model. On the other hand, Dunnett's test was used to compare the germination of each treatment with the control and Tukey's test was used for multiple comparisons of the treatments. The data were fitted using SAS 9.4 Proc Mixed software ( SAS Institute Inc., 2017 ).

Logistic regression was used to model the ger­mination probability as a function of the exposure time and the temperature at which the seeds were subjected. The logistic model was expressed as follows [Eq. 2]:

where: p : probability of germination, z : a linear function of the explanatory variables considered in the analysis.

The coefficients of the z function were obtained using maximum likelihood method. The full model, including independent and quadratic terms, and all reduced models were tested. The residual mean deviation was used to scale the standard errors and test statistics in the fitting process. The models were evaluated using the value of the statistical parameters AIC and -2 log likelihood. In order to choose the best model to explain the germination process, score selection variables was used at the beginning of the process to know the behavior of the variables in the fitting process. After this, other different models were tested to introduce combinations of different variables. Some of the candidate models are shown in Tab. 1. The final model was chosen taking account the smallest values of statistical parameters and the best combination of significant variables with a reasonable biological interpretation. The data was fitted using SAS 9.4 Proc Logistic software ( SAS Institute Inc., 2017 ).

The mixed model (Tab. 2) showed that the pro­portion of germinated seeds depended on the species (Pr>F=0.0004), applied heat shock (Pr>F=<0.0001) and the interaction between both factors (Pr>F=0.0007). So, heat shock treatments in each species showed different impact on the proportion of germination. Germination was not different to the control for A. senegal under exposure times of 1 minute for all the temperatures, or for temperatures smaller than 90º for exposure time higher than 1 minute (Figure 1a). However, for temperatures higher or equal to 120ºC and time of exposure higher or equal to 5 minutes, germination was reduced significantly in comparison to the control. For 5 and 10 minutes and temperatures higher than 120ºC, the seeds did not germinate. On the other hand, in the case of A. seyal , the germination was reduced respect to the control, in exposure times of 1 minute and temperature equal to 90ºC (Figure 1a), and when the exposure time was higher than 5 minutes for all the temperatures. Tukey's test showed (Figure 1b) that the germination of A. seyal , was smaller than A. senegal in the most of the heat shock treatments with time exposure higher than 5 minutes. Higher time exposures decreased the germination for all the temperatures in comparison to the same heat shock in A. senegal . Tab. 3a showed that, in A. senegal , hardly non-germinated seeds were viable. In contrast, A. seyal showed a high proportion of viable non germinated seeds, even under high exposure times and temperatures (Tab. 3b). At the highest exposure time (10 minutes) and higher temperatures (120 and 150ºC), there were a different percentages of viable seeds, although the percentage of germination was clearly reduced. The results of control were 59 Germinated and 2 Viable seeds in Acacia senegal (G+V=61) and 70 Germinated and 6 Viable in Acacia seyal (G+V=76).

The logistic model chosen to explain the effects of Acacia species probability of germination, revealed that all variables and their interactions exert significant influences (Pr>χ ² <0.05) (Tab. 4) on the germination of the species studied. The logistic regression model provided additional information on the differences bet­ween species and the influence of the different factors considered in the study on the germination probabili­ty. According to the logistic model, the probability of germination in A. senegal (Fig. 2) decreased as temperature increased, and this occurs in a similar way for all three exposure times. Germination probability remained above 60% up to 120ºC and above 50% in the highest temperatures at 1 minute exposure. In A. seyal (Fig. 2), the decrease in probability of germination was slightly more pronounced as temperature increased. Although the behaviour was consistent in both spe­cies, with exposure time of 1 min, the probability was reduced to 60% at 120°C approximately. While in A. senegal , the differences were observed from 60ºC among the exposure times and were pronounced at 90ºC, in A. seyal , differences were observed from 60ºC.

This work has analyzed the effects of heat shock treatments on the germination of two important species of arid ecosystems. Both species, under ex­treme conditions (exposure time of 5 minutes and tem­peratures of 150ºC) did not germinate, however, the species differed in their behaviour in the other range of temperatures. A. senegal was more tolerant to heat shock treatments because showed higher germination than A. seyal . This information is essential to improve the sustainable forest management prescriptions to the conservation of these fragile ecosystems in relation to forest fires.

Fire has been reported as a tool for stimulating seed germination ( Núñez & Calvo, 2000 ) that also serves to eliminate competing species ( Núñez et al. , 2003 ; Zuloaga-Aguilar et al. , 2011 ). However, under arid conditions, fire characterized by high temperature or prolonged exposure time could cause lower survival or even prevent germination in Acacia species ( Danthu et al. , 2003 ; Walters et al. , 2004 ; Dayamba et al. , 2008 ; Zuloaga-Aguilar et al. , 2011 ), depending on the seed-coat characteristics.

In this study, the effects of several heat shock treatments on germination were evaluated for Acacia senegal and Acacia seyal species from the Sudanese savannah in the semi-arid region of Southwest Ethiopia. Our results confirmed that the proportion of germinated seeds in both species was equal to or above 60% at 1 minute exposure to all temperatures. On the other hand, the probability of germination of the two species also decreased when the temperature increased, being more pronounced in A. seyal . The logistic model provided information regarding how the probability of germination were affected by differences between the species and the heat shocks considered. Additionally, the model allowed us to introduce quadratic terms into the process in order to test how they would affect germination probabilities. In our case, the square of the temperature proved to be an explanatory variable, indicating the asymptotic limit of the germination capacity of these two species (Tab. 4). Logistic model added scientific advances to the germination in arid environments. Our findings could allow stakeholders to fit germination predictive models to wider areas in Ethiopia. This is a highlight result of this work because logistic model is not a method regularly used in similar studies focused on germination of Ethiopian trees and shrubs.

In A. seyal , our results differed from those obtained by Gashaw & Michelsen (2002) . These authors observed that the germination was smaller than 30% over the entire range of studied temperatures (20, 60, 90, 120, 150, 200ºC) at 1 and 5 minutes of exposure time. These differences may be related to the characteristics of the scratch method employed [hot water for 24 hours in our study vs . 1 to 3 hours by Gashaw & Michelsen (2002) ], and the germination conditions [germination chamber with constant conditions of temperature, humidity and photoperiod vs . soil to simulate field conditions]. Therefore, our conditions that did not depend on local conditions (humidity and temperature of the soil). On the other hand, the prolonged scarification methods, allowed us to obtain a greater germination under control treatment, and therefore, a better com­parison of these results to the different heat shock treatments. Gashaw & Michelsen (2002) reported no germination under heat shock with temperatures higher than 120ºC -for time of exposure greater than 1 minute- and with temperatures higher than 90ºC -for time of exposure greater than 5 minutes-. They did not consider time exposure higher than 5 min, however, our results showed that if temperature was smaller than 90ºC, the proportion of germination was similar statistically to the control in Acacia senegal . The increase of germination with higher exposure time obtained in this paper may be caused by a secondary seed dormancy, as the Tetrazolium test confirmed that some of these seeds were viable (Tab. 3a,b). In this paper, a strengthen statistical analysis was carried out respect the previous researchers ( Teketay, 1996 ; Gashaw & Michelsen, 2002 ). A linear mixed model, which consider the experimental error in the fitting process, added robustness to the obtained results. In addition, with the Dunnet's test, additional information was obtained about the germination in the treatments respect to the BAU germination (control treatment). These results allowed us to complete the information obtained by Tukey's tests.

Moreover, the results of the viability of seeds im­proved our knowledge about the consequences of fire in arid species. The germination after fire depended on the intensity of heat shock and also limited to seed survival. Heat shocks could not scarify the integument of the seeds without resultant lethal damage to their embryos. Our findings showed that they could retain viability instead of dying. In these sense, further ana­lysis of Germinated+Viable seeds and Germinated given Germinated+Viable might be useful. In this re­gard, in A. seyal , dormancy would be a strategy or an adaptation to environmental perturbations. This result was agreed with Danthu et al. (2003) , who observed that the percentage of viable seeds at temperatures between 232-241°C was about 80%.

On the other hand, in Acacia senegal , our results agreed with those obtained by Danthu et al. (2003) , who obtained germinations higher than 90% in this species at two fire intensities. The first intensity simulated by Danthu et al. (2003) consisted of fuel load about 100g/m ² that reached temperatures over 154-160°C at 1 m above the ground. The second, 500g/m ² that reached 93-99°C at 3 m above ground. Therefore, our results could be similar than those obtained in studies that evaluate germination at 1 m above ground.

In many species, a similar germination under mo­derate heat shock treatments has also been reported. A threshold of temperatures/times of exposures were found in species adapted to forest fires, such as Genis­ta florida ( Tárrega et al. , 1992 ), Pinus sylvestris (Nú­ñez & Calvo, 2000), C. striatus, G. berberidea and G. triacanthos ( Rivas et al. , 2006 ), and Pinus pinaster ( Herrero et al. , 2007 ).

The differences found in this paper among the two species, reflect an important evolutionary process under arid conditions, related to seed coat characteristics and dormancy of seed. The successful establishment of seedlings depends on long-term seed survival and dispersion strategies ( Gashaw & Michelsen, 2002 ; Danthu et al. , 2003 ; Eriksson et al. , 2003 ; Zida, 2007 ). In this paper, results showed the seed survival of A. seyal after application of heat shock treatments. Although A. senegal showed higher germination after different heat shocks, A. seyal showed poor results in the treatments respect to the control, but viable seeds after heat shocks. It could be related to a secondary seed dormancy, but more studies would be needed to confirm this fact as it would depend on the scarification method applied. Previous researchers such as Teketay (1996) , Danthu et al. (2003) and Villalobos et al. (2002) showed that chemical scarification with Sulphuric acid or mechanical treatment improved the germination under high temperatures.

In this study, one only seed lot origin from each species was used in order to target the germination in the different heat shock treatments with respect to the control. In this sense, one of the limitations of this study was the lack of assessment the intra-specific variation on the tolerance to heat-shock treatments. However, the pre-germination treatments allowed us to obtain a higher germination under control conditions. These treatments could not affect the results because they mostly addressed changes respect the control for the heat-shock treatments.

The combination of knowledge about the biology of a species allowed us to understand the functional characteristics and the evolution of a plant in different regions and ecosystems ( Paula et al. , 2009 ). Models could simulate changes in plant community composition due to different scenarios. In fire-prone ecosystems, changes in fire regime may be more relevant to climate change ( Paula et al. , 2009 ) which may affect the appearance biological, physical and chemical ecosystem, depending on the intensity, duration and the degree of destruction of the fire ( Ursinoa & Rulli, 2010 ). In this sense, our results are very important for the best regeneration management or for the post fire regeneration treatment because, if an intense forest fire happens, additional treatments (e.g. seeding, planting, or selviculture treatments to increase the regeneration) could be beneficial due to the fire damage on the existing seed bank.

Several models have predicted an increase in the frequency and severity of natural disturbances such as forest fires, windthrows, etc., combined with significant drought episodes and an overall increase in temperature as a consequence of climatic change ( IPCC 2007 ). In the Sudanese savannah, this is exacerbated by intense grazing, which creates temporal and spatial heterogeneity in the fuel load ( Savadogo et al. , 2007 ). Our findings showed new data of germination in different fire simulations which could improve the silvicultural strategies designs. Therefore, it would be necessary to avoid prescribed burning in these areas as a method to control the vegetation or to improve the grazing quality in the regeneration areas of these forests, as they could cause a reduction in germination of the two species. Also, control of fuel in these areas would be necessary to reduce the probability of ground forest fires that would also affect the possibility of regeneration of these species. The logistic model could also be included in integrated models of silviculture and forest fires incidences to improve the sustainable management of these species.

In this paper the germination of two species of Acacia ( A. senegal and A. seyal ) from sub-Saharan Africa has been studied after applying different heat shocks. Differences have been found between the two species in their reaction to these treatments. In the two cases, under extreme conditions (exposure time of 5 minutes and temperatures of 150ºC) the two species did not germinate. But the species differed in their behaviour in the other range of temperatures. A. senegal was more tolerant to heat shock treatments because showed higher germination than A. seyal . However, A. seyal showed viable seeds after heat shocks. Non-germinated seeds were viable after heat shocks which is an important finding for the soil bank of this species. Furthermore, the logistic regression model showed the square of temperature as an explanatory variable, indicating the asymptotic limit germination capacity of these species, as well as the interaction between species exposure time and temperature. This information could serve to enhance forest regeneration by developing vegetation management strategies based on prescribed fires.

The authors would like to thank to B. Turrión and R. San Martín for the assistance with statistical analyses and to W. Tadesse for facilitating relevant information related to Ethiopian forest. On the other hand, the authors thank to the University of Valladolid staff for their assistance at the laboratory. Finally, the authors would also like to thank Andrea Blanch for revising the English version and providing generous linguistic advice.