Isotope dendroclimatological studies on Juniperus

0 downloads 0 Views 46MB Size Report
were sampled at seven sites in South Wollo and North Gondar. .... 5.4.2 Corrections and coherence ...... Copernicus University and the International Incoming Short Visits ... Walstra (Ghent University, Belgium) and Dr Hans van der Kwast (VITO, ...... 2005), cambial wounding (Détienne, 1989; Sass et al., 1995), radiocarbon ...
Isotope dendroclimatological studies on Juniperus procera from Ethiopia: towards a reconstruction of Blue Nile baseflow

Tommy Hendrikus Gerardus Wils

Submitted to the University of Wales in fulfilment of the requirements for the Degree of Doctor of Philosophy

School of the Environment and Society Swansea University May 2009 2

He dared not look at her, but he felt in his whole being that she was looking at him at that moment and was perhaps looking at him wrathfully, that there must be indignation in her black eyes and that her face was flushed.

The Idiot, Dostoevsky

2

Summary High-resolution climate reconstructions are fundamental to improve the understanding of the climate system. Tree rings are an excellent high-resolution proxy, but currently the spatial coverage of tree-ring chronologies is mostly limited to temperate regions. Here, it is aimed to extend the tree-ring networks into Northwest Ethiopia, the main source area of the Nile River tributaries Blue Nile and Atbara. Juniperus procera trees were sampled at seven sites in South Wollo and North Gondar. The samples were crossdated and a few sub-samples from the post bomb era were radiocarbon dated. The trees in South Wollo did not form annual growth rings. Though multiple false, missing and indistinct growth rings were present, growth rings formed in the main wet season could be identified in North Gondar. Various growth-ring variables were measured, including ring width, minimum blue intensity, and carbon and oxygen isotope ratios, and examined for environmental and physiological signals. A combined chronology of ring widths and carbon-isotope values was developed over three sites in North Gondar. This chronology was calibrated against instrumental records of river discharge (r=0.75, p0.05) or in comparison with the reconstruction of December/January Blue Nile River discharge (r=-0.11, p>0.05). The lack of association is probably related to variation due to Bega rainfall, White Nile River discharge and evaporation downstream. Besides, maximum level is a poor measure for total river discharge during the wet season. Rauda minimum levels are also neither significantly correlated with measured December/January Blue Nile River discharge during AD 1913-1921 (r=0.53, p>0.05, Fig. 92) nor with the reconstructed river discharge values (r=0.09, p>0.05, Fig. 93). However, in the period AD 1875-1921 and around AD 1850, a non-linear association is

7.5

3 7.0 2 6.5 1

0 1913

1914

1915

1916

1917

1918

1919

1920

Figure 92. Detrended Rauda Rauda Water Level (m)

Blue Nile Flow (km3)

4

minimum levels (grey) compared to measured December/January Blue Nile River discharge (black) (r=0.53, p>0.05).

6.0 1921

Year (AD)

221

8.5

5

8.0 7.5

B

4

7.0 3 6.5 2

6.0

1

5.5

0 1750

5.0 1800

1850

6

1900

1950

2000

1950

2000

Rauda Water Level (m)

6

3

Blue Nile Flow (km )

A

Year (AD)

3

Blue Nile Flow (km )

5 4 3 2 1 0 1750

1800

1850

1900 Year (AD)

Figure 93. Reconstructed December/January Blue Nile River discharge (black, alternative in blue) compared to (A) detrended Rauda Nilometer minimum levels (green), and (B) years of drought and famine (grey lines, extreme years in red) (Degefu, 1988; Gebrekirstos, 2006) and years following floods in Khartoum (dashed blue) (Walsh et al., 1994; Holmgren et al., 1997). visible (Fig. 93). In the early part of the record (AD 1755-1770) coherence is present when shifting the reconstruction one year back in time. The weak association is probably related to variation due to rains between January and July, annual variation in

222

White Nile River discharge and poor accuracy and reliability of the Rauda Nilometer record. The shifted coherence in the early part of the record suggests a slight dating error as a result of low replication. Years of extreme drought and famine, plotted in red in Fig. 93, described as 'a holocaust of drought, famine, cattle disease, epidemic, and cholera' (AD 1836-1837) (Degefu, 1988:29) and 'one of the most serious droughts' (AD 1888-1892) (Degefu, 1988:30), correspond to low reconstructed December/January Blue Nile River discharge in AD 1837 and AD 1889. The extreme drought in AD 1836-1837 may have induced the severe rise in carbon isotope ratio in core K4C from AD 1835 to 1836. Damage, either directly by drought or indirectly by man or fauna, seems to have affected the tree irreversibly, though it remains unclear to what extent. Most other years of drought and famine also correspond to low reconstructed December/January Blue Nile River discharge (Fig. 93), except for minor droughts in AD 1895-1896 ('minor drought' (Degefu, 1988:30)) and AD 1964-1966 ('virtually undocumented' (Degefu, 1988:31)). Years following floods in Khartoum coincide with high reconstructed December/January Blue Nile River discharge in AD 1841 and AD 1842 , but hardly in AD 1867 (Fig. 93). As pointed out by Walsh et al. (1994), floods in Khartoum are related to three factors, that is, Blue Nile floods, local rain and urbanisation. The mismatch in AD 1867 suggests that the contribution of the Blue Nile to the flood in AD 1866 was minor.

223

5.6.3 Lake Naivasha

A high-resolution, semi-quantitative water-depth record of Crescent Island Crater, Lake Naivasha, Kenya (6 in Fig. 90), derived from staff gauge readings (AD 1883-1993) or reconstructed from the sedimentological characteristics of profundal sediments (~AD 880-1883) (Verschuren et al., 2000b), compares well with reconstructed December/January Blue Nile River discharge (Fig. 94). In the early part of the record, high water depths correspond with high river discharge in AD 1757, AD 1767-1768, AD 1780 and AD 1788-1789. In AD 1820-1821, the records do not match, but between AD 1824 and AD 1835 both show an increasing trend. The severe drought of AD 1836-1837, related to the severe rise in carbon isotope ratio in

36

5

30

4

24

3

18

2

12

1

6

0 1750

0 1800

1850

1900

1950

Naivasha Water Depth (m)

6

3

Blue Nile Flow (km )

core K4C and remembered in Ethiopia as a ´holocaust of drought´ (Degefu, 1988:29), is

2000

Year (AD)

Figure 94. Reconstructed December/January Blue Nile River discharge (black, alternative in blue; 7-year running means in red and orange, respectively) compared to water depth of Crescent Island Crater, Lake Naivasha, Kenya (green) (Verschuren et al., 2000a; 2000b; Couralet et al., 2005; Sass-Klaassen et al., 2008).

224

reflected by a severe drop in Lake Naivasha water depth until AD 1846. Low water depths and river discharges also coincide between AD 1864 and AD 1880, except around AD 1872. After AD 1880, the two records diverge, though peaks still coincide in AD 1895-1897, AD 1918 and AD 1930-1931, whereas high river discharge in AD 1936 is reflected by slightly higher water depth in AD 1937. The high lake levels at the end of the 19th century coincide with increased Nile flood frequency as deduced from river discharge at Aswan (Walsh et al., 1994) . The period around AD 1970 and the mid AD 1980s, characterised by extremely low December/January river discharge, show declining trends in water depth. Relative to the entire 1100-year record of water depth from Lake Naivasha, the period AD 1760-1880 was exceptionally dry and for comparison referred to as Naivasha Drought 1 (ND1) (Verschuren, 2004). This drought only matches the reconstruction of December/January Blue Nile River discharge if the section before AD 1836 of core K4C is excluded (Fig. 88). Dating errors may cause minor mismatches in AD 1820-1821 and AD 1936-1937, whereas the mismatch around AD 1872 may be caused by the relative inertia of the response of a lake to precipitation changes. The coincidence around AD 1900 of increased Nile flood frequency and high water depths in Lake Naivasha, but low to average December/January Blue Nile River discharge, suggests that the increased Nile flood frequency was caused by increased White Nile River discharge, as the waters of the White Nile originate from the Equatorial Lake Plateau close to Lake Naivasha. The divergence between AD 1880 and AD 1940 is probably caused by climatic differences between the two areas (see section 6.3). It is remarkable that the two records match to some degree, regarding the lack of correlation between the annual precipitation records from the Equatorial Lake Plateau

225

and Northwest Ethiopia (Hulme, 1994). Even though year-to-year fluctuations in precipitation are different, major and longer term fluctations in water availability seem to correspond.

5.6.4 Other proxy records

A climate record for Africa compiled by Nicholson (2001) from multiple sources reports the key features observed in the reconstruction of December/January Blue Nile River discharge. Wet conditions prevailing until the late 18th century, also dominating Lake Naivasha until AD 1760 and corresponding to colder conditions over Europe in the Little Ice Age (Verschuren et al., 2000b), changed towards dryer conditions at the beginning of the December/January Blue Nile River discharge reconstruction. Therefore, it is difficult to detect the change in the river discharge reconstruction, but a decreasing trend is certainly apparent (Fig. 88). Nicholson (2001) reports dry conditions over Africa during the AD 1820s and AD 1830s, corresponding to the devastating drought of AD 1836-1837. Wetter conditions are reported for the end of the 19th century, corresponding to the Lake Naivasha record, but are only weakly reflected by the AD 1895-1897 peak in the river discharge reconstruction. Dry conditions are again reported in the early 20th century and after AD 1960, corresponding to the AD 1900s, AD 1970s and AD 1980s droughts in the river discharge reconstruction. Wet conditions over Africa in the AD 1950s are reflected by high December/January Blue Nile River discharges. Other proxy records have a low resolution, are short or located in an area climatically different from North Gondar. Nevertheless, dry conditions in the early 1760s and during

226

the AD 1836-1837 drought suggested by a record from Lake Hayq, Northern Ethiopia (7 in Fig. 90) (Lamb et al., 2007), correspond to low December/January Blue Nile River discharges. Wetter conditions at Lake Hayq in the early 20th century, corresponding to wet conditions at Lake Naivasha during AD 1880-1940, are only reflected by a slight increase in December/January Blue Nile River discharge between the mid AD 1900s and the late AD 1950s. The historical droughts of AD 1836-1837 and AD 1888-1892, reflected by low December/January Blue Nile River discharge in AD 1837 and AD 1889, may have been detected in a record from Lake Abiyata, Southern Ethiopia (8 in Fig. 90) (Legesse et al., 2002). A short (AD 1963-1996), but well-dated reconstruction of lake level from Lake Hora, Central Ethiopia (9 in Fig. 90) (Lamb et al., 2002), shows similarities with the river discharge reconstruction, particularly during the early AD 1980s drought and the wetter conditions of the late 20th century. Equatorial drought during the late 18th and early 19th century, apparent from Lake Chibwera, Lake Kanyamukali (Western Uganda, 10 in Fig. 90) and Lake Baringo (Central Kenya, 11 in Fig. 90) (Bessems et al., 2008), matches the reconstruction of December/January Blue Nile River discharge if the section before AD 1836 of core K4C is excluded (Fig. 88). A pollen record from the Dega Sala swamp (12 in Fig. 90) in the Arsi Mountains, Southeastern Ethiopia (Bonnefille and Umer, 1994), suggests increasing temperatures since AD 1750, which may have contributed to a decline in December/January Blue Nile River discharge by increasing evaporation. Infilled valley deposit sequences from Tigray, Northern Ethiopia (13 in Fig. 90) (Machado et al., 1998), suggest increased aridity since the early 17th century already, contradicting to wet and cold conditions until the mid or late 18th century reported elsewhere (Bonnefille and Umer, 1994; Verschuren et al., 2000b; Legesse et al., 2002; Lamb et al., 2007). In a review, Umer et

227

al. (2004) suggest that the period AD 1270 to 1850, the latter part of which corresponds with the Little Ice Age in Europe, was generally wet, but interrupted by prolonged periods of severe drought. Unfortunately, the reconstruction of December/January Blue Nile River discharge is too short to add much to this discussion, but a trend of increasingly dry conditions is apparent since the start of the reconstruction in the mid 18th century. Palaeoclimate reconstructions from speleothem records are being developed in Ethiopia (Asrat et al., 2006; Baker et al., 2007; Blyth et al., 2007), but at the time of writing, interpretation is still too difficult to yield records that can be compared to the reconstruction of December/January Blue Nile River discharge.

228

5.7 Forcing

5.7.1 Precipitation, temperature and autocorrelation

Conway (1997) developed a hydrological model of the Blue Nile that predicts monthly runoff from precipitation, potential evapotranspiration, water storage, and soil moisture deficit (Box 2). Potential evapotranspiration depends on temperature, whereas water storage and soil moisture deficit effectively constitute the autocorrelation in the river discharge record. Consequently, the key factors determining December/January Blue Nile River discharge are precipitation and temperature during the preceding months. Correlation analysis, comparing measured December/January Blue Nile River discharge to measured precipitation and temperature data from Gondar, indicates that April-December precipitation (r=0.57, p0.01) 50-ring segment correlations is a consequence of a similar correlation with a lower sample size and can be regarded as a warning. The crossdating process applied to samples from Doba forest (Fig. 30) might have worked if trees had formed growth rings synchronously, but the fact that it produced an erroneous match, emphasises its weaknesses. Visual comparison of ring-width series 243

may mislead the researcher, as there will always be some matching, whereas skeleton plotting may provide a more constraining and robust way of crossdating. The studied

Juniperus procera trees in Doba forest possess a highly ambiguous wood anatomy, leading to wood-anatomical remarks being attached to up to 54% of the growth rings. Although wood-anatomical features did still render some of the potential adjustments improbable, this effect was too minor to yield a trustworthy match. Therefore, the crossdating process was adapted before crossdating the samples from North Gondar, moving the emphasis to direct comparison of the wood and skeleton plotting (Fig. 31A). This revised crossdating process produced matches with statistically significant mean series inter-correlations between 0.56 and 0.59 (individual correlations p