Relating tree growth to rainfall in Bolivian rain forests: a test for six species using tree ring analysis
- First Online:
- Cite this article as:
- Brienen, R.J.W. & Zuidema, P.A. Oecologia (2005) 146: 1. doi:10.1007/s00442-005-0160-y
- 1k Views
Many tropical regions show one distinct dry season. Often, this seasonality induces cambial dormancy of trees, particularly if these belong to deciduous species. This will often lead to the formation of annual rings. The aim of this study was to determine whether tree species in the Bolivian Amazon region form annual rings and to study the influence of the total amount and seasonal distribution of rainfall on diameter growth. Ring widths were measured on stem discs of a total of 154 trees belonging to six rain forest species. By correlating ring width and monthly rainfall data we proved the annual character of the tree rings for four of our study species. For two other species the annual character was proved by counting rings on trees of known age and by radiocarbon dating. The results of the climate–growth analysis show a positive relationship between tree growth and rainfall in certain periods of the year, indicating that rainfall plays a major role in tree growth. Three species showed a strong relationship with rainfall at the beginning of the rainy season, while one species is most sensitive to the rainfall at the end of the previous growing season. These results clearly demonstrate that tree ring analysis can be successfully applied in the tropics and that it is a promising method for various research disciplines.
KeywordsClimate–growth relationRadiocarbon datingTropical rain forestTropical dendrochronologyWood anatomy
Most tropical rainforests (sensu Whitmore 1998) experience some seasonality in rainfall and have a distinct dry season (Worbes 1995). This leads to annual rhythms in tree physiology (e.g., Borchert 1994b) and often causes leaf-abscission (Worbes 1999). A dry season of at least 2 months with less than 50 mm of rain, results in reduced diameter growth or complete cambial dormancy for many species (Worbes 1999). In this way, a distinct growth boundary is formed in many tree species. Indeed, various studies have proven the occurrence of growth rings in tropical rainforest species throughout the world (Detienne 1989; Vetter and Botosso 1989; Devall et al. 1995; Pumijumnong et al. 1995; Worbes 1999; Stahle et al. 1999; Dunisch et al. 2003; Fichtler et al. 2003, 2004), in spite of the traditional belief that tropical rainforest trees do not produce rings (Lieberman et al. 1985; Whitmore 1998). Yet, the analysis of growth rings of tropical trees is less straightforward than that of temperate regions, as rings are not always formed annually. Some tropical ring studies have reported the occurrence of 2 rings per year (Jacoby 1989; Gourlay 1995) or an irregular formation of rings (Sass et al. 1995).
Deviations of annuality in tree rings can be linked to species-specific differences in physiology and wood anatomy, but can also be due to exogenous factors such as rainfall patterns. Two distinct dry seasons per year for example, can lead to the formation of two rings per year (Jacoby 1989; Gourlay 1995). In the same way irregular patterns of rainfall, such as exceptional droughts during the regular rainy season, can induce the formation of ‘intra-annual’ rings (Borchert 1999; Borchert et al. 2002). The response to rainfall patterns differs among species, often in relation to leaf-fall behaviour (Borchert et al. 2002). Annual tree rings are more often found in deciduous species than in brevi-deciduous or evergreen species (Worbes 1999; Borchert 1999). Furthermore, it is shown that responses vary among sites for the same species (Breitsprecher and Bethel 1990). Given these problems, there is a clear necessity to better understand the relations between rainfall, leaf-fall behaviour and ring formation of tropical rainforest trees.
The potential of tree ring analysis for obtaining reliable age estimates and long-term growth data is very high (Bormann and Berlyn 1981; Baas and Vetter 1989; Eckstein et al. 1995). Information on tree ages and growth rates is crucial to understand the dynamics of tree populations (Enright and Hartshorn 1981) and to develop sustainable management systems for tropical timber species (Stahle et al. 1999; Worbes et al. 2003). Most existing data about tree ages however, are either indirect estimates based on projections (Lieberman et al. 1985; Laurance et al. 2004) or are highly disputed in literature (Martinez-Ramos and Alvarez-Buylla 1998; Chambers and Trumore 1999).
Is tree ring formation annual and can ring analysis be applied for accurate tree growth analysis and age estimates?
Is tree growth correlated with rainfall?
Do the study species differ in response to rainfall?
Materials and methods
Adult stature, growth potential, leaf-fall behaviour and products of the six study species. Growth potential was based on the five maximum annual diameter increments of different trees (cf. Clark and Clark 1992)
Growth potential (cm/year)
Amburana cearensis (Allemao) A.C.Smith
Bertholletia excelsa H.B.K
Cedrela odorata L.
Cedrelinga catenaeformis (Ducke) Ducke
Peltogyne cf. heterophylla M.F.Silva
Tachigali vasquezii J.J.Pipeloy
The species also differ in leaf-fall behaviour. Amburana, Bertholletia and Cedrela are obligate deciduous and appear leafless for periods of several weeks to months during the year. Amburana and Cedrela lose their leaves in July and produce new leaves in October–November. Bertholletia is leafless in the period from August to October for about 2–4 weeks. Cedrelinga, Peltogyne cf. heterohylla and Tachigali vasquezii can be classified as brevi-deciduous: they never appear completely leafless, but consistently change all their leaves at the end of the dry season or beginning of the rainy season (August–September). As a juvenile, these species behave as evergreen with continuous leaf replacement without annual patterns related to the dry period.
Study areas and site conditions
The vegetation in the Bolivian Amazon can be classified as a tropical lowland moist forest, with a canopy 25–35 m high. In this region 283 species have been recorded, with an average of 75 species per hectare. The average basal area (DBH>20 cm) is 15 m2 ha−1 with 103 trees per hectare (Superintendencia Forestal 1999).
We used two main study areas, approximately 400 km apart. The first area is the private property Purisima (11°24′S, 68°43′W) in the department of Pando, 50 km south of the town of Cobija. This area consists of 850 ha mainly undisturbed tropical moist forest, on undulating terrain. From this site, samples of Cedrela odorata and Amburana cearensis were collected.
Samples of the other four species were collected from several adjacent logging concessions (10°55′S, 65°40′W), approximately 40 km north of the town of Riberalta, consisting of mainly undisturbed tropical moist forest. For Bertholletia excelsa additional samples were collected in and around Riberalta (11°00′S, 66°10′W), including trees growing on the bank of a small creek in a secondary forest and two plantation trees.
Sample collection, treatment and ring measurements
Between October 2002 and September 2003 we took samples from trees that were felled for timber or from trees that had recently died from natural causes. For Tachigali some trees were felled for the purpose of this study. From each tree a cross-sectional wood disc or segments of cross-sections were taken with a chainsaw. The height at which the samples were taken, varied between 0.5 m and 2.5 m above the ground. In cases of buttresses, we preferably took samples above the buttresses, but this was not possible for all trees. Diameters of the sampled trees varied from 30 cm to 200 cm. After air-drying the discs, we polished them mechanically with sandpaper up to grit 600. For each disc we took 2–4 radii for ring counting and measuring. The average values of these radii were taken to describe average diameter increment. This meant that for trees with buttresses we counted and measured the rings both in buttresses and in shorter radii (depressions). Due to this methodology the ring widths between the different radii varied considerably for some trees. Ring boundaries were examined microscopically or by the naked eye. Each ring was marked and we interconnected every tenth ring between the different radii. Due to this procedure wedging rings and discontinuous growth zones were detected and the number of errors in ring marking reduced. Discontinuous growth zones are those bands that occurred only over a part of the disc and are further referred to as intra-annual growth zones. All ring widths were measured to the nearest 0.001 mm using a computer-compatible tree ring measuring system (Velmex Inc. Bloomfield, NY, USA) and a ×40 stereomicroscope. Ring measurements were performed along the predetermined radii in a straight line, and generally perpendicular to ring boundaries.
First we conducted a quality control using correlation analysis (program COFECHA; Holmes 1983; Grissino-Mayer 2001) to detect measurement errors, such as wedging rings or intra-annual growth zones. For this purpose long-term growth trends have to be removed from the raw measurements. A flexible cubic spline is the most appropriate detrending method for trees from closed-canopy stands (Cook and Peters 1980) and the flexibility of the cubic spline determines the amount of variation removed from the growth trend (Grissino-Mayer 2001). We fitted cubic splines of different flexibility, from 5 years to 30 years, to the ring series of each individual tree of all species. The spline stiffness that yielded the highest inter-series correlations was used for the standardization of the tree ring series of a certain species. This procedure was followed to remove the long-term growth trends and filter out low frequency variation that might be caused by, e.g., canopy-dynamics. After fitting a cubic spline curve to a tree ring series the residuals were calculated by dividing each raw value by the value of the fitted spline.
Results of quality control of the ring width measurements executed with COFECHA. Shown is the number of trees (series) before and after the quality check, the mean inter-series correlation between the detrended series over the time-span 1950–2000 (1975–2001 for Tachigali) and spline length. For Tachigali we analyzed the means of the measured series for each tree and included all individuals in the analysis (for further explanation see text)
Number of trees (series)
Mean inter-series correlation
Spline length (year)
Maximum number of observed rings
Bertholletia excelsa (Riberalta)
Bertholletia excelsa (St. Maria)
Peltogyne cf. heterophylla
For Tachigali we applied a somewhat different procedure, as this is a fast growing, short-lived species, with fewer and extremely wide rings that resulted in a high accuracy of ring dating. All sampled trees with cylindrical growth patterns were included in the analysis. Furthermore, for this species we used averages of the radii instead of the raw series as little variation was generally observed in ring-widths between radii.
From the selected, best correlating tree ring series we established a tree ring chronology for each species with ARSTAN (Cook 1985; Cook and Holmes 1985). Per species the raw tree ring series were standardized applying a cubic spline with the same stiffness as for the quality check. Then, the standardized series were averaged into one chronology per species. For Bertholletia two different chronologies were established; one for the trees that originated from Riberalta and one for the trees that were growing along a small creek 16 km outside Riberalta. These were assumed to respond differently to climate due to the different hydrological site conditions.
To study the relationship between precipitation and ring width we correlated each chronology with monthly rainfall data from the nearest weather station, for each of the six species. For Bertholletia, Cedrelinga and Peltogyne we used the available precipitation data from Riberalta that span the period from 1941 to 2001. For Tachigali the same precipitation data were used, but the chronology only covered the period from 1975 to 2001, because rings near the centre of the discs were not clearly distinguishable. For Cedrela and Amburana we used the rainfall data from Cobija over the period from 1951 to 1981.
Pearson correlations were computed to assess the relationship between the tree ring chronologies and the annual rainfall (from previous July till actual June) and sums of precipitation in distinct periods of the year. These periods were based on the division of seasons according to Walter and Lieth (1964), with the dry season running from June till August and the rainy season defined as the months with a mean of more than 100 mm of rain (October–April, cf. Fig. 1). The rainy season was further divided into early (October–December) and late (January–April) rainy season. We also calculated correlations with the transition from the previous rainy season to the following dry season (April–June) and from the dry season to the following rainy season (September–November), since transition periods have proved to be important for tree growth (Jacoby and D’Arrigo 1990).
Next, we conducted a more detailed analysis to gain insight in the response of trees to rainfall over the year by computing Pearson correlations between the tree ring chronologies and the running three monthly rainfall sums. We did this for 22 months, from November of the previous growing season till August following the current rainy season. Part of the previous year was included, because of the lag effect that rain can have on tree growth, e.g. through filling up water reservoirs in the soil (Fritts 1976). We included some months of the dry season into the analysis, as we did not know how long tree growth continues into the dry season.
For Peltogyne we applied radiocarbon analysis to verify that rings are annual, as we did not find any correlations between rainfall and ring width for this species. We used the elevated concentration of radiocarbon in the atmosphere in the 1960s (resulting from nuclear tests) to determine the exact ages of four rings, which were anatomically predated between 1985 and 1962. Small wood samples (1.6–2.3 mg) were taken and pre-treated using standard techniques (Worbes and Junk 1989). Radiocarbon content was determined using accelerated mass spectrometry (AMS). Values of 14C were expressed in terms of percentage modern, meaning that the 14C activity was corrected to δ13C and expressed in terms of the United States National Bureau of Standards (NBS) oxalic acid standard as of 1950. For calibration of these values to possible calendar dates we used the calibration curve for the southern hemisphere of Wellington, New Zealand (Manning and Melhuish 1994).
Distinctiveness and structure of rings
Distinctiveness and type of growth zone for the study species. Growth zones typifications follow Coster (1927) and Worbes (1995). ‘Clear’ growth zones are easily and accurately recognizable, while ‘rather clear’ growth zones are more difficult to distinguish and often need careful examination of various parts of the wood disc
Distinctiveness of growth zone
Type of growth zone
Clear to rather clear
Vessel distribution and alternating pattern of parenchyma and fibre bands (4)
Clear to rather clear
Repeated pattern of alternating fibre and parenchyma bands (3)
Marginal parenchyma bands (2)
Density variation (1)
Peltogyne cf. heterophylla
Clear to rather clear
Marginal parenchyma bands and vessel distribution (2,4)
Clear to unclear (centre)
Density variation (1)
Samples with a high number of indistinct rings in the outermost part of the stem were excluded from analysis (10% of the samples). Additional problems related to tree ring identification occurred at very short radii, where a lot of wedging rings occurred. Cross dating between different radii largely solved this problem. Of all species Cedrela showed the most distinct growth rings. Even at low growth rates it was no problem to identify (annual) rings with high accuracy. In Tachigali no rings could be distinguished in the centre of the stems, which made reliable dating only possible for the outer part of the stems at diameters above 10 cm. Rings of Cedrelinga were the most indistinct.
Quality check and inter-series correlation
We rejected a rather high percentage of series in the quality control, but for each species more than half of the initial number of trees remained in the eventual analysis and were used for subsequent calculation of the tree ring chronology (Table 2). The majority of the rejected series came from extreme radii, like depressions or buttresses or different radii from trees that showed irregular, i.e. lobate-growth.
The strength of the inter-series correlations varied between species. Amburana, Bertholletia and Cedrela showed medium to high correlations between the tree ring series of different trees while the correlations for the other species were considerably lower (Table 2).
Correlations between ring widths and rainfall
Pearson correlation coefficients for the relation between the ring width index and whole year sums and sums during distinct periods in the year
Transition rain to dry
Transition dry to rainy
Early rainy season
Late rainy season
Entire rainy season
Peltogyne cf. heterophylla
Further evidence of annual nature of growth rings
For Bertholletia and Peltogyne for which no correlations with rainfall were found, we conducted other methods. For Bertholletia two trees were sampled, which were known to have been planted in 1960. Based on ring counting one of these trees was unequivocally dated to exactly 1960. The stem of the other tree contained five vague rings. Excluding these rings the tree also would be dated to 1960.
For one Peltogyne tree, radiocarbon dates were determined for four rings, which were predated by ring counting. Values of 14C activity expressed in percentage modern carbon are 103.5±0.4, 120.2±0.4, 131.7±0.5 and 123.0±0.5 for the rings predated as 1962, 1965, 1967 and 1985. These values corresponded best to the years 1957, 1960, 1962 and 1982/1983. This reveals that rings of 1957, 1960 and 1962 were 5 years younger than dated initially. Two of these five rings were missed between the rings of years 1962 and 1982/1983 and three rings were missed in the sapwood, between the bark and 1982/1983.
Ring formation in the study species
All species showed the formation of distinct rings, but there were substantial differences in distinctiveness among species and we encountered several problems with intra-annual growth zones. The distinctiveness of rings varied among species and is mainly caused by differences in wood structure, which is species-specific (Detienne 1989). The form of terminal parenchyma bands were the most clear marks to distinguish tree rings, while other types of rings (repeated patterns of parenchyma and fibre bands and variations in wood density) were more difficult to distinguish. This difference was especially apparent at lower growth rates. Generally, the determination of growth rings became less accurate at lower growth rates, i.e. in smaller tree rings.
Intra-annual growth zones (‘false rings’) and wedging or completely missing rings occurred in four of the six species and are a common problem in ring studies of tropical rainforest trees (cf., Ogden 1981; Detienne 1989; Priya and Bhat 1998; Dunisch et al. 2002). Generally, these anomalous rings were successfully detected through differences in the anatomy of the ring and by checking the continuity of rings over the entire stem disc. The high incidence of such false and wedging rings emphasises the necessity to work with entire stem discs and makes the use of increment cores difficult (Detienne 1989; Worbes 1995). Especially when working with new tropical species, it is absolutely necessary to start with entire stem discs for estimating dating accuracy and to learn anatomical differences between annual and non-annual growth zones.
Tachigali does not form distinct tree rings as a juvenile tree (i.e. below approximately 10 cm in diameter), although annual rings are clearly distinguishable at larger diameters. This differentiation in wood structure coincides with the change in leaf-fall behaviour from evergreen as a juvenile to brevi-deciduous as a full-grown canopy tree. Such a change in leaf-fall pattern is probably a physiological adaptation to increasing drought stress, when growing from the understory into the canopy. Canopy trees use more water, due to higher total photosynthesis levels (more light) and higher respiration levels (higher temperatures in the canopy). Similar patterns have been reported in other studies, where ring formation in juvenile trees was non-annual for species, that form annual rings as an adult tree (cf., Detienne 1989; Priya and Bhat 1998; Dunisch et al. 2002). Hence, some caution is necessary when analyzing and interpreting the rings in the juvenile wood. For Cedrela regular formation of annual rings in juvenile wood is proven (Dunisch et al. 2002). The obligate deciduous habit of Amburana juveniles is a strong indication for annual juvenile rings, and we expect the ring formation of the other species to be predominantly annual as well.
For all study species we removed a considerable portion of the sampled trees from the chronology, as they did not show a high correlation with the rest of the tree ring series. There are various explanations for this. First, our method of ring dating did not follow standard dendrochronological techniques (e.g., Stokes and Smiley 1996), as we did not perform a consequent visual quality control of the dated series, but only a statistical one. Therefore, it is possible that we did not detect multiple measurement errors, which may have caused low correlation values for certain series. Second, tree growth was highly variable between different radii of the stem causing low correlations among individual trees.
For four of the six species, we were able to establish tree ring chronologies with relatively high inter-series correlations. These chronologies contained sufficient individuals and could be linked to rainfall, which proves the annual nature of tree ring formation in these species. Correlations between the chronologies of Cedrelinga and Tachigali further supports the annual character of the rings (Stahle 1999). For the two other species we gathered strong evidence of the annual nature of rings through radiocarbon “bomb dating” (Worbes and Junk 1989) and counting rings on discs of known age. This revealed that rings are annual in both species, but we also observed problems with the ring determination in these species. One Bertholletia tree contained some vague intra-annual bands or false rings. For Peltogyne, a comparison of the radiocarbon dates with ring dates revealed that some rings were missed due to the poor visibility of some very thin parenchyma bands. Age determinations by ring analysis will therefore probably give a slight underestimation of the real ages. These species seem unsuitable for palaeoclimatic climate reconstruction.
Relationship between tree growth and rainfall
Variation in growth in four out of six species is at least partially determined by rainfall. Relations between growth and rainfall are generally positive, indicating that growth is limited by the amount of rainfall. Many other studies found that tree growth was influenced by rainfall in climates with comparable precipitation levels (1,000 to 1,700 mm; Jacoby and D’Arrigo 1990; Devall et al. 1995; Worbes 1999; Stahle et al. 1999; Enquist and Leffler 2001) or under wetter conditions (3,000 mm; Dunisch et al. 2003). With a further rise in annual precipitation (up to 3,860 mm), however, it seems that growth is not limited by rainfall anymore, but by fluctuating patterns in photosynthetic radiation (cf., Clark and Clark 1994).
The strength of the correlations we found in this study is comparable to other dendro-chronological studies on tropical rainforest trees (cf., Worbes 1999; Stahle et al. 1999; Enquist and Leffler 2001; Dunisch et al. 2003; Fichtler et al. 2004). Only, for Cedrelinga the response was relatively weak compared to other species.
Each species shows a slightly different response to rainfall. The differences in response did not suggest a relation with the deciduousness of the species, as was found in other studies (Worbes 1999; Enquist and Leffler 2001).
For three out of the four species (Amburana, Cedrelinga and Tachigali) the general pattern of sensitivity to rainfall is quite similar, with the highest sensitivity to the rainfall during the early rainy season and no significant effect of rain later in the season. Several studies have found similar patterns (Pumijumnong et al. 1995; Dunisch et al. 2003; Fichtler et al. 2004).
There are several possible explanations for this pattern. Firstly, the rains at the end of the dry season and beginning of the rainy season could break bud-dormancy, resulting in a longer growing period and hence a higher total growth for that year. Such a rainfall-induced break of bud-dormancy, is observed in deciduous species (Wright and Cornejo 1990; Borchert 1994c; Borchert 1999; Priya and Bhat 1999), but does not satisfactorily explain the pattern in our species: the timing of bud break is not in concurrence with the months of sensitivity to rainfall; bud break occurs in August–September (Cedrelinga and Tachigali) and October–November (Amburana), while growth is sensitive to rain until January.
A second explanation for the sensitivity to early rainy season precipitation might be a gradual increase in water reserves stored in the soil as the rainy season proceeds. After a time-lapse of 2–3 months following the start of the rainy season the water reserves in the soil may surpass critical levels and trees can maintain a positive water balance throughout the rest of the rainy season, even in short dryer periods. Hence, only the rainfall at the beginning of the rainy season has a significant impact on diameter growth.
Finally, ring growth is by far strongest soon after the beginning of the rainy season triggered growth (Priya and Bhat 1999). Later in the rainy season ring growth is much less (Worbes 1999). This may be due to decreasing photosynthetic capacities of older leaves (Mooney et al. 1981; Hirose et al. 1989; Ackerly and Bazzaz 1995; Miyaji et al. 1997) or to a shift from growth to storage (Iwasa and Cohen 1989). These factors, or a combination of these factors, most probably explain why early rainfall is so important for tree growth that year.
For Cedrela we found an interesting pattern in its sensitivity to rainfall, which is clearly different from the other three species: its growth is most dependent on the amount of rain well before the actual growing season, in the transitional months from the previous rainy season to the beginning of the dry season and growth is negatively correlated to the rainfall at the end of the current growth period. The same pattern is found in Panama for Annona spraguei (Devall et al. 1995) and in Brazil for Cedrela odorata (Dunisch et al. 2003). Dunisch et al. (2003) explain the positive influence of rainfall of the previous rainy season through storage of reserves during this period, which are used for extra growth at the beginning of the next growth season. The negative correlation of growth with rainfall at the end of the current growth period however, seems contradictory to this mechanism. At present, we have no explanation for this contradiction. Other possible mechanisms causing the observed lag in relations are bud preformation (e.g., Critchfield 1960) and long-term water-table depth or stem water-storage (Borchert 1994a). Further investigation into the physiological reactions of this species to rainfall is needed to clarify this phenomenon.
Outlook and future directions of research
We proved in this study the occurrence of annual tree rings in six tropical tree species in the north of Bolivia. For this region the estimations of the percentage of species with rings vary from 25% (Roig 2000) to 50% (Brienen and Zuidema 2003) and we expect, based on our results, that rings of most of these species are annual. Average estimation of the number of species with clear growth zones is about 50% for the Amazon basis as a whole, but with regional differences depending on the degree of seasonality (Alves and Angyalossy-Alfonso 2000).
The accuracy of ring counting for the determination of tree ages varies among species, depending on the distinctiveness of rings and the regularity of ring formation in the centre. Even for species containing less visible, or non-annual rings in the centre, ring analysis is one of the best methods for determining (long-term) growth rates and for large-scale estimation of tree ages. Ring analysis is relatively cheap and precision of the age estimations are much higher than age-projections of short term growth rates, which are mostly used in tropical trees (Lieberman et al. 1985; Laurance et al. 2004).
Of the six investigated species Cedrela and Amburana show good potential for the development of climate-sensitive chronologies; their rings can be dated precisely, their growth is sensitive to climate and the species grow relatively old (>200 years). This dendrochronological potential is new for Amburana, but was described earlier for Cedrela (Tomazello Fo et al. 2000; Dunisch et al. 2003).
We are very grateful to Adhemar Cassanova Arias, Merlijn Janssens, Henri Noordman, Jeanette Pacajes, Anneke Rijpkema, Jan Rodenburg, Vincent Vos and Oliver Yancke for their indispensable assistance with the ring measurements and the staff of PROMAB and the ‘field team of Purisima’ for their help with the fieldwork. We thank Instituto de Geología y Medio Ambiente (IGEMA) from the Universidad Mayor de San Andres (UMSA) in La Paz and Dr. Jaime Argollo for the use of their measurement equipment. Rene Boot, Ute Sass, Marinus Werger and Martin Worbes and two anonymous reviewers are acknowledged for helpful comments on an earlier draft version of this paper. This research was part of the research program of the Programa de Manejo de Bosques de la Amazonía Boliviana (PROMAB). Radiocarbon dating was skilfully done by dr. K. van den Borg, Physics department, Utrecht University.