Ring formation in the study species
As evinced from the macroscopic and microscopic examinations, all the studied tree species exhibited clear growth ring boundaries although the degree of distinctiveness varies across species and study sites. Such variations in the distinctiveness of growth rings among the study species are attributed to differences in wood structures, which are known to be species-specific as revealed by Detienne (1989) and Worbes (1995). Moreover, it was believed that the varied site conditions and disturbance factors contributed to this disparity by acting on tree growth.
The growth boundaries in J. procera, which are characterized by the alternating thin- and thick-walled tracheids (cf. Sass-Klaassen et al. 2008; Wils et al. 2011a), are clear and were identified easily; while those of O. europaea, which consisted of scanty (marginal) parenchyma bands with thick-walled fiber, are less distinct compared to the other species, thus were more difficult to distinguish. The results are in line with previous studies elsewhere in Ethiopia which found clearly visible annual growth rings for different tree species. For J. procera trees, the formation of distinct annual growth rings was previously revealed in areas with distinct cycles of wet and dry seasons (Wils et al. 2011b; Couralet et al. 2005; Sass-Klaassen et al. 2008). However, considerable variations were noticed depending on the site and climate conditions where the species grows; it was even reported that J. procera trees form indistinct and non-annual growth rings in some areas in Ethiopia (Wils et al. 2009) and in Kenya (Jacoby 1989) depending on the environment where the species grows (e.g. in well-drained environment).
Similarly, the growth ring boundaries of P. falcatus consists of radially aligned narrow and flattened tracheids, with slightly thickened cell walls in the latewood cells. The cambial activity of the evergreen P. falcatus trees was previously studied in a relatively moist dry Afromontane forests in Ethiopia and revealed the occurrence of intra-annual growth resembling the annual growth rings following the onset of intra-annual climatic conditions (Krepkowski et al. 2011). This phenomenon makes the identification of the annual growth ring boundaries more challenging. The study also indicated that the occurrence of such false rings is irregular and shows considerable variation among individual trees depending on the changes in the microenvironment, thus alerted the need for further analysis to proof the annual nature of the rings. In this case, given the area is subject to extended dry periods with a relatively distinct cycle between dry and wet seasons, it was possible to identify the annual growth ring boundaries. For O. europaea, as to the knowledge of the researcher, no documentation on its dendrochronological study was found in Ethiopia.
In general, while the formation of distinct growth ring boundaries is reported hitherto, their annual nature varies considerably with different agro-ecology, thus urges for intensification and replication of such studies at different ranges of the growing limits of these tree species. The formation of distinct and annual growth rings, but with varying degree of distinctiveness among species, was also reported for various other tree species in Ethiopia (Gebrekirstos et al. 2008; Couralet et al. 2010; Tolera et al. 2013) and elsewhere in the dry tropics (e.g. Rozendaal and Zuidema 2011), even in the moist tropical regions (Brienen and Zuidema 2005; Groenendijk et al. 2014). The formation of annually resolved growth rings in the tropics is most often attributed to the seasonality in rainfall (Worbes 1995, 1999).
Alike to many other dendrochronological studies of tropical tree species, this study confronted problems associated with the occurrence of intra-annual growth zones (false rings) and wedging or missing rings. In many tropical species, the existence of false and wedging rings hampered tree-ring analyses (Worbes 2002). Trees growing in harsh environmental conditions are often prone to develop multiple missing rings (Wils et al. 2011a). In this study, such anomalies were especially apparent in slow growth phases (suppression periods –which are induced by various growth limiting factors), and with narrower rings and eccentric growth as shown with some stem discs. Slow growing trees such as O. europaea trees, were associated with such anomalies and affected the accuracy of growth ring identification. However, such problems were minimized by matching the growth ring boundaries across different radii and by checking their continuity over the entire circumference of a stem disc. Previous works also highlighted the necessity of working with entire stem discs to increase dating accuracy and to address anatomical differences between annual and intra-annual growth zones in tropical species (Worbes 1995; Worbes 2002; Brienen and Zuidema 2005).
In this study, higher values of inter-series correlations and mean sensitivity were found for all studied species across sites. These higher values of inter-series correlations and mean sensitivity (Stahle 1999) coupled with the significant correlations of ring widths with rainfall confirm that the rings were formed annually (Brienen and Zuidema 2005; Brienen et al. 2016). Several other studies on tropical tree species also used successful crossdating (Worbes 1995; Stahle et al. 1999; Worbes et al. 2003; Fichtler et al. 2004; Gaspard et al. 2018) and correlations between tree-ring chronologies and rainfall (Worbes 1999; Fichtler et al. 2004; Brienen and Zuidema 2005; Couralet et al. 2005; Baker et al. 2008; Brienen et al. 2016) as a proof for the formation of annual growth rings for various tree species.
Therefore, as expected, the unimodal rainfall pattern coupled with the extended dry periods induce cambial dormancy during the dry periods followed by active growth during the wet seasons, thus forming annually-resolved growth rings. Results of this study agree with previous studies showing that many tropical species are characterized by distinct seasonal patterns of radial growth. This seasonality and the long dry season particularly in water-deficit forest ecosystems induces periodic cambial dormancy in trees, leading to the formation of annual growth rings (Worbes 1999; Schongart et al. 2006; Gebrekirstos et al. 2008; Sheffer et al. 2011).
Crossdating
Overall, crossdating among trees was successful for each tree species in both study sites. Samples which failed to match with the rest of the ring width series, probably due to the high incidence of false, wedging, and/or missing rings were excluded to minimize dating errors which may compromise the master chronology. Thence, successfully crossdated ring width series which showed higher inter-series correlations were used to establish tree-ring chronologies for the study species. Other studies conducted in a similar environment in Ethiopia also reported successful crossdating for J. procera trees (Couralet et al. 2005; Sass-Klaassen et al. 2008; Wils et al. 2011b) and other tree species from the dry forests and woodlands belonging to the genera Acacia (Eshete and Ståhl 1999; Gebrekirstos et al. 2008) and Boswellia (Tolera et al. 2013; Mokuria et al. 2017).
Although it varies among species and regions, a series intercorrelation value of about 0.4 is, generally, considered crossdatable (Fritts 1976, p 254–268). In this study, high values of mean sensitivity (0.43–0.63), standard deviation (0.59–0.66), series intercorrelation (0.51–0.57), and low values of average autocorrelation (0.04–0.20) were found. All the inter-serial correlations calculated in COFECHA were greater than their respective critical levels (p < 0.001) for all study species, implying that the statistical crossdating for all study species is highly significant.
Several other studies, particularly from the dry tropical regions found similar higher values of mean sensitivity and inter-series correlations and revealed their significance for tree-ring studies in the tropics. These site- and species-specific mean inter-series correlations are generally comparable (e.g. Wils et al. 2011a) reported it to be 0.52–0.59 for J. procera trees sampled from North Gondar, Ethiopia or within the range of other similar studies reported for various tropical tree species elsewhere, which ranges from 0.24 (Trouet et al. 2006) to 0.63 (Therrell et al. 2006). The average mean sensitivity values reported here are also comparable or higher than those reported by several authors elsewhere; for instance, an average mean sensitivity value in the ranges of 0.25–0.50 was reported for J. procera from Ethiopia (Wils et al. 2011b); 0.39–0.49 for Brachystegia spiciformis Benth. From Zambia (Trouet et al. 2006), 0.42–0.50 for three understory tree species from the Democratic Republic of Congo (Couralet et al. 2010); and 0.27–0.29 for Burkea africana Hook and 0.31–0.41 for Pterocarpus angolensis D.C. from Namibia (Fichtler et al. 2004).
These statistical attributes of the tree-ring chronologies, such as high inter-series correlations (Stahle 1999) and high mean sensitivity values show that the chronologies are sensitive to climate variability (He et al. 2013). The higher year-to-year variability, expressed as mean sensitivity (Fritts 1976, p 261) together with low values of autocorrelation indicates strong responses to annually changing environmental conditions (Sass-Klaassen et al. 2008). This is indicative for the synchronous growth patterns among individual trees of a given species, and that the studied tree species respond to a common climate forcing which corresponds with periodic wood formation among the trees (Worbes 1995). Similar to this study region, such strong inter-annual variability of tree-ring chronologies is typical of moisture-sensitive regions (Fritts et al. 1980), indicating that the chronologies developed from the study species have captured strong common climate signals. Besides, the study species showed an EPS value higher than the threshold value (FPS > 0.85) suggested by Wigley et al. (1984) for the reliability of chronology development. These results confirm that trees of the same species synchronize their growth in response to a common environmental factor, providing further proof for the applicability of dendrochronological studies in the study region. Thus, these tree-ring chronologies are believed to provide useful climate proxy data to study regional climate variations in the region and to study the large-scale climate system. In general, the study confirms the potential suitability of tree-ring data from the study region as a proxy for further dendroecological and paleoclimate studies in the region.
Climate-growth relations
All the study species from the remnant dry Afromontane forests of northern Ethiopia showed positive and significant relationships with rainfall of the main growing season (JJAS) and annual rainfall. This supports the claim that plant growth in dry regions is more likely related to the total available soil moisture than water availability during specific periods, and thus the soil-plant system is buffered at seasonal rather than monthly timescales (Krepkowski et al. 2011). This can be ascribed to the phenological characteristics of the evergreen tree species, which are known to radially grow under sufficient soil moisture conditions.
The formation of distinct annual growth rings in the study species is also an indication of a common climate forcing of tree growth, i.e., the rainfall seasonality which is characterized by a unimodal pattern. This signifies the detrimental role of moisture availability for tree growth.
The strong relations of the tree-ring chronologies with wet-season rainfall is illustrated by a highly synchronous pattern between growth and rainfall in Fig. 8 using part chronologies from J. procera as an example.
The finding of this study agrees with Krepkowski et al. (2011) and Brienen et al. (2016) who found that tropical tree growth is sensitive to the amount of rainfall received. Other similar tree-ring studies from the seasonally dry tropical regions also found tree growth to be positively influenced by rainfall seasonality (Gebrekirstos et al. 2008, Couralet 2010). This positive relationship implies that rainfall is an overriding climatic factor determining plant growth in the study region. Some dendrochronological studies conducted elsewhere in the tropics have also shown that tree growth is influenced by rainfall seasonality in areas with rainfall amount ranging between 1000 and 1700 mm (Worbes 1999; Stahle et al. 1999; Enquist and Leffler 2001). In view of this, the fact that the amount of rainfall in our study sites is generally below 1000 mm signifies that plant growth is highly limited by water-deficit conditions in the study region.
In agreement with some findings of previous tree-ring studies elsewhere in Africa (e.g. Fichtler et al. 2004; Gebrekirstos et al. 2008; Trouet et al. 2010), in this study, the major seasonal precipitation displayed higher regression coefficient values with ring growth compared to annual precipitation (Fig. 7). Strong correlations of tree growth with rainfall during the main rainy season were reported not only for deciduous trees but also for many evergreen tropical tree species (Worbes 1999; Enquist and Leffler 2001; Fichtler et al. 2003; Gebrekirstos et al. 2008). A strong wet-season impact on tree growth is expected as 90% of the annual rainfall in most dry tropical regions falls during that season (Brienen et al. 2010). Trouet et al. (2010) also reported the influence of climate, and of precipitation in particular, on tree growth to be strongest at the core of the rainy season.
In this study, the overall pattern of tree species’ sensitivity to rainfall was found to be more or less similar. Strong correlations with rainfall occurred during the mid-rainy season (July–August), whereas no significant effects of rainfall were observed in the beginning and end of the wet season. This finding is partially comparable to that reported by Brienen and Zuidema (2005), who found the highest sensitivity to rainfall early in the rainy season but insignificant effect of rain late in the season. The authors argued such sensitivity patterns to comply with the early saturation of the soil water reserves which are then retained within the entire season, entailing that there is more significance of rainfall at the beginning of the rainy season than at the end. Unlike this explanation, in this case, it is believed that the small amount of rainfall during the beginning of the rainy season is not sufficient to break the extended dormancy period, but as the rainy season proceeds, the continued enrichment of soil water can significantly enhance radial growth. But then, it shortly becomes limited towards the end of the wet season, making it less significant for tree growth again. Therefore, it seems plausible that the available amount of water during the mid-wet season largely determines tree growth in the study region. The observed variation in the strength of relationships of tree growth and rainfall across species may be explained in relation to the ontogenetic (physiological responses) and site condition (other related environmental factors) differences and the interactions thereof. Such differential growth responses may be linked to functional traits related to water storage and conductance (Mendivelso et al. 2013).
Besides, this study also recognized correlations with minimum and maximum temperatures although they were largely insignificant, alternating between positive and negative relationships. In this way, it can be inferred that temperature has some degree of influence on the growth of the tree species, but was generally complex and weak. Negative relationships between tree growth and temperature were also reported in other seasonally dry tropical regions (Schongart et al. 2006; Gebrekirstos et al. 2008; Pucha-Cofrep et al. 2015; Mokria et al. 2017). The negative relations between temperature and the chronologies implies that high temperatures may negatively affect tree growth, possibly by increasing drought stress due to the enhanced evapotranspiration losses.
In general, studies undertaken in the dry tropics have highlighted the influence of climate on tree growth. These studies revealed that tree growth is significantly correlated predominantly with rainfall seasonality (Bullock 1997; Stahle et al. 1999; Worbes 1999; Fichtler et al. 2004; Therrell et al. 2006; Nath et al. 2006; Schongart et al. 2006; Gebrekirstos et al. 2008; Brienen et al. 2010) in most of these regions, and also with temperature variability in others (Fichtler et al. 2004; Trouet et al. 2006). This implies that climate signal is one of the main controlling factors for tree growth in tropical regions. However, growth response to the climate signal is modified by tree species, provenance, competition and site conditions, among others as revealed by Fritts (1976). Therefore, such studies need to be intensified in different regions considering various tree species, and this is important for planning and implementing restoration and sustainable management endeavors in tropical forests, particularly that of the TDFs.