Introduction

The Amazon basin is one of the regions of highest rainfall in the world and is a major source for atmospheric water vapor (Figueroa and Nobre 1990). Over the last decades it has experienced a more variable climate, with severe droughts in 2005 and 2010, causing increased tree mortality (Phillips et al. 2009) and reducing carbon uptake (Feldpausch et al. 2016). Some models predict climate-induced Amazon die-back for this century (Salazar et al. 2007; Cox et al. 2008).

Tropical tree growth and, therefore, carbon uptake are linked with atmospheric CO2 and micro-climate through photosynthesis, while seasonality is mainly determined by the amount of annual rainfall and its seasonal distribution (Borchert 1994). Rising temperatures (Doughty 2011) and reduced dry season rainfall (Clark et al. 2010) reduce photosynthesis through higher leaf-to-air vapor pressure deficits leading to stomatal closure and, thus, limiting carbon uptake, although rising levels of atmospheric CO2 may compensate for this effect (Lloyd and Farquhar 2008). Studies using remote sensing data, however, analyzed as MODIS Enhanced Vegetation Index (EVI, an index of canopy photosynthetic capacity) report a 25% increase in photosynthesis between 2000 and 2005 during dry seasons due to more sunlight (Huete et al. 2006). Xiao et al. (2006) suggest that leaf phenology in tropical evergreen forests is not determined by the seasonality of precipitation, but by the availability of solar radiation. A pan-tropical analysis of climate effects on tree growth by Wagner et al. (2014), however, showed that, both, precipitation and solar radiation are the main drivers for seasonal variations in tree growth. Wind disturbance can also be an important ecological force as it may accelerate forest fragmentation (Laurance and Curran 2008). Physiognomy of elfin forests at exposed ridges is an adaptive response to frequently high wind speeds (Lawton 1982).

Neotropical montane rainforests receive some of the highest rainfall, but have received much less scientific attention, associated with difficulties of access due to steep and rugged terrain. Montane rainforests are hot spots of biodiversity and endemism (Gentry 1995) due to altitudinal gradients over more than 2000 m, extraordinary edaphic and micro-climatic heterogeneity, vegetation dynamics due to frequent landslides (Lozano et al. 2007), limitation of nutrients (Homeier et al. 2010) and the landscape history since the Pleistocene (Lauer 1986). Concomitant with increasing altitude on wet tropical mountains, forest structure changes (Von Humboldt 1849) and the altitudinal zonation is even compressed on small outlying mountains (Lomolino 2001). Although the physiognomy of tropical montane forests has been studied for over 100 years (Shreve 1914; Grubb 1977; Martin and Bellingham 2016), it remains poorly understood.

The El Sira Mountains in Peru, an eastern parallel extension of the Andes, reaches into the Amazon basin and forms the first barrier for moist air from the tropical Atlantic and the Amazon basin. It rises up to 2400 m asl, and the windward (eastern) side receives more than 7500 mm rainfall per year (Graham et al. 2016) exceeding annual accumulations for rainfall hotspots over the southern tropical Andes reported by Espinoza et al. (2015). Because of its remote location, rugged terrain and poor accessibility, the El Sira uplands are still covered with true pristine forests and therefore represent an ideal landscape-scale laboratory to investigate the effects of droughts on montane forest ecosystems.

In terms of regional climate patterns, Casimiro et al. (2013) reported that mean annual temperatures increased by 0.09 °C for the Peruvian Amazon-Andes region between 1965 and 2007, but no trend for annual rainfall was presented. Other studies report a decrease in rainfall of 3% per decade for the Mantaro River basin (IGP 2005) and 8.3% per decade for the Peruvian-Ecuadorian Amazon watershed between 1970 and 1997 (Espinoza et al. 2006). Tropical montane regions are generally poorly populated, and networks of rain gauges are sparse or absent, but satellite-based sensors can provide remedy on global scale at remote areas. The “Tropical Rainfall Measuring Mission and Precipitation Radar” (TRMM PR) satellite provides a unique opportunity to examine rainfall characteristics in tropical regions that have low density of observations. It is designed to provide information on the three-dimensional structure of rainfall over the tropics and subtropics (Simpson et al. 1988; Kummerow et al. 2000). Several recent studies have validated satellite rainfall estimates against rain gauge records and strongly supported their applicability (Franchito et al. 2009; Semire et al. 2012; Mantas et al. 2015; Manz et al. 2016).

This study is the first climatological description of an altitudinal gradient on the El Sira uplift and uses TRMM PR rainfall estimates and rain gauge records to identify rainfall anomalies over the last years. In parallel, the effects of seasonal precipitation anomalies on tree radial growth variations are analyzed.

Materials and methods

Study sites

The El Sira Mountains form the backbone of the El Sira Communal Reserve (RC El Sira), located roughly 9°–11° south and 74°–75° west (Fig. 1). It is almost entirely surrounded by lowland Amazonian habitats and isolated from all other Andean elements by large river systems. In the northern and most isolated section of the uplift are found the highest elevations of the entire range, reaching above 2400 m asl.

Fig. 1
figure 1

El Sira cordillera and regional precipitation regimes. The map indicates areas of annual rainfall over 4000 mm (satellite estimates, Manz et al. 2016). 1: Pucallpa, 2: Puerto Inca, 3: Panguana (P251), 4: Iparia, 5 (red oval): Altitudinal transect (P845, P1391, P1568, P2230), 6: Altitudinal transect from Graham et al. 2016. a, b show topography and seasonal precipitation along horizontal profiles (dashed lines in map) in east–west and north–south direction, respectively

Full size image

The study was conducted from May 2011 until September 2015 at an altitudinal gradient along the western slope at the northern part of the mountain range (Fig. 1). Trees selected for observation are located within five permanent 1-ha plots along the gradient established by the RAINFOR project (Malhi et al. 2002; Montegudo et al. 2014; Valenzuela et al. 2015) within the RC El Sira (sites P845-P2230) and the private preserve area “ACP Panguana” (site P251) in the central Amazonia of Peru, region Huanúco (Table 1). Montane rain forest sites within the Communal Reserve are quite remote and only accessible by a 2-day long foot walk, Valenzuela et al. (2015) report at least 729 angiosperm plant species within 120 families and 398 genera for this area. The most abundant families are in this order Orchidaceae, Rubiaceae, Fabaceae, Lauraceae, Melastomataceae and Arecaceae.

Table 1 Overview of forest sites
Full size table

Along the transect, different altitudinal forest types can be distinguished (Fig. 2). Tropical lowland rain forests (terra firme) are found between 170 and 650 m asl, tropical submontane or transitional rain forests between 650 and 1100 m asl, tropical montane cloud forests (TMCF) between 1100 and 1850 m asl and tropical elfin forests (or tropical subalpine forests, Frahm and Gradstein 1991) above 1850 m asl. The occurrence of peat mosses (Sphagnum sp.) clearly indicates the perhumid character within elfin forests.

Fig. 2
figure 2

Vegetational profiles of sites. Depth of projections is about 5 m. For further details see Table 1

Full size image

Table 1 shows the location and forest main characteristics of each 1-ha plot. A more comprehensive description of the altitudinal forest types along the transect is given by Valenzuela et al. (2015).

Ground-based climate data

Air temperature (°C), relative air humidity (%), both used to calculate vapor pressure deficit of air (vpd), global solar radiation (W m−2), precipitation (mm), wind direction and speed (m s−1) were recorded every 30 min using Watchdog 2700ET weather stations (Spectrum Technologies, Inc., USA) at 860, 1540 and 2230 m asl, located near to the study sites P845, P1586 and P2230, respectively. The weather station at 860 m ran from August 2011 until August 2015. At 1540 m the station ran only from May 2011 until August 2012 due to an animal attack. The third weather station ran from May 2011 onwards until a lightning stroke in September 2014 destroyed it. Precipitation was measured by tipping buckets but turned out to be inconsistent or even absent for most of the time as a consequence of litter blocking the water discharge of the instrument. Therefore, large cans (volume = 19 L) with funnels collecting rain where additionally installed at the end of March 2015, next to the three weather stations. Total amount of captured rain was measured at the end of August 2015. Long-term rain gauge records of the towns of Pucallpa (154 m asl, about 140 km NE of study area) and Puerto Inca (200 m asl, about 26 km N of study area) were supplied for this study by the National Meteorology and Hydrology Service (SENAMHI, Peru, http://www.senamhi.gob.pe/). The rainy season is defined here as lasting from October until end of March and the dry season from April until end of September.

Remote sensing data

Digital elevation data (SRTM 90 m, from http://srtm.csi.cgiar.org) were used to represent regional topography. In order to illustrate annual precipitation extremes at a 5 × 5 km2 resolution, we used satellite-borne estimates of precipitation (“TRMM PR product 2A25 version 7”, Iguchi et al. 2000, 2009) merged with ground-based rain gauge measurements by a linear model from Manz et al. (2016). For linear regressions of time-series with ground-based rain gauge records we used daily “TRMM PR product 3B42 version 7” data, which are globally gridded estimates of precipitation compiled from passive-microwave instruments and geosynchronous infrared instruments at a spatial resolution of 0.25° (≈ 28 km for the study area) (Huffman et al. 2007, 2010). GIS data were analyzed and visualized using the open source software “QGIS 2.18.3” and daily rainfall estimates for rain gauge locations were extracted using the “Point Sampling Tool” plugin.

Selection of trees and measurement of radial stem growth

Within each site, 11–15 vital trees of the most frequent species (excluding palms) were selected for observation (Table 2). Identifications and scientific names of species are taken from Montegudo et al. (2014), and complete tree census data are accessible at http://www.forestplots.net/. All selected individuals were equipped with spring-loaded linear displacement potentiometers (MMR 10_11 R5 K, MEGATRON Elektronik AG and Co., Munich, Germany) with a resolution of < 10 µm to detect diurnal and seasonal stem radius changes. Potentiometers were mounted in the center of stainless-steel frames and then fixed to the stem at 1.3 m height with two screws anchored at least 5 cm deep in the xylem, at least 6 cm apart from the measuring point (Güney et al. 2017). They were placed with their tips onto the innermost bark, after partly removing the outer parts without infringing living cortex and cambium, covered and shielded with a reflecting and a plastic foil. Radial changes were recorded every 1-h using a Dl2e datalogger (Delta-T Devices, Cambridge, UK).

Table 2 Overview of monitored trees
Full size table

Data analysis

Daily stem radius change (src_d) was calculated after Köcher et al. (2012) as the difference between the maximum stem radii in the morning of two subsequent days. Statistical analyses and visualization were carried out in python v.3.5 using the open source packages “pandas”, “numpy”, “scipy” and “matplotlib”. To test for significant differences between src_d during rainy and dry season a Kruskal–Wallis-Test was used. Pearson correlation and linear regression were used to compare daily, monthly and seasonal data of TRMM PR estimates and rain gauge data. Air vapor pressure deficit (vpd) was calculated according to Goff and Gratch (1946). Data gaps are due to sensor or logger failure. Obvious outliers due to malfunctions were deleted.

Results

Regional climate, seasonal patterns and altitudinal gradients

Seasonal precipitation patterns of South America are controlled by the interaction of large-scale atmospheric currents and the complex Andean topography. Along the eastern flanks of the Andes the easterly trade winds and strong topographic gradients generate pronounced orographic effects (Bookhagen and Strecker 2008; Espinoza et al. 2009, 2015) resulting in strong convection (Romatschke and Houze 2010) and thereby spatio-temporally highly intermittent precipitation patterns with precipitation gradients of up to 190 mm km−1 (Espinoza et al. 2015). These orographic effects are illustrated in Fig. 1, showing an instant increase in annual rainfall from about 1800 mm at Iparia up to 11,000 mm over the El Sira uplands. Uplands receive up to four times more precipitation during rainy seasons than during dry seasons.

Occasionally cold fronts, advancing from the south along the Andean Mountain Range into the western area of the Amazon basin, let temperatures drop within hours by 10–20 °C (locally called “Friajes”), reaching values of 5–10 °C and generally associated with high wind speeds (Marengo 1983).

The altitudinal gradients in climatic parameters are shown in Fig. 3. From May 2011 until September 2015, temperature (Fig. 3a) ranged from 15.0 °C (July 28, 2013) to 36.1 °C (22 September 2013) in the lowland at Puerto Inca (200 m asl), from 14.6 °C (22 August 2011) to 34.1 °C (22 September 2013) in the submontane forest (860 m asl), and from 7.7 °C (18 July 2012) to 30.3 °C (22 September 2013) in the elfin forest (2230 m asl). Annual mean temperature decreased from 26.5 °C at 200 m asl to 15.1 °C at 2230 m asl (− 0.56 °C per 100 m).

Fig. 3
figure 3

Altitudinal gradient in climatic parameters. Open boxes and symbols refer to dry seasons (April–September), filled boxes and symbols to rainy seasons (October to March). Boxes indicate the interquartile range (IQR) of daily averages, diamonds represent means, and whiskers extend 1.5 times the IQR from boxes, “+” indicate outliers and circles show absolute extremes. a Temperature gradient. Dashed and dotted lines indicate linear regression of average temperature during dry seasons (y = −0.578/100x + 26.95, R2 = 0.975, p = 0.013) and rainy seasons, respectively (y = − 0.58/100x + 27.43, R2 = 0.978, p = 0.011). Gradients in b vapor pressure deficit (vpd) of air, c solar radiation, and d sum of rainfall between April and August 2015 at the western slope (circles: rain gauges; triangles: TRMM PR estimates) and at the eastern slope (diamonds, Graham et al. 2016)

Full size image

Wind is predominantly blowing from N, NW or E, but during “Friajes” wind is blowing from the S or SE. Mean vpd of air also decreases along the transect (Fig. 3b, diamonds), as a consequence of lower temperatures and cloud formation. The latter leads to a decrease in solar radiation, although exposed upland ridges experience highly variable weather conditions throughout daily courses (Fig. 4) leading to large daily amplitudes in temperature, vpd and solar radiation.

Fig. 4
figure 4

Daily courses of climate data along the elevational climatic gradient during three selected days (“sunny”, “rainy/cloudy”, and “inversion”). Altitudes: 860 m (solid lines), 1540 m (dashed lines), and 2230 m (dotted lines). The inversion is predominantly recognizable in air temperature at midday. Daily sums of solar radiation are: 24.1, 19.2 and 13.9 MJ m−2 for the “sunny”, 2.1, 2.9 and 3.0 MJ m−2 for the “cloudy/rainy”, and 2.1, 15.7 and 24.4 MJ m−2 for the “inversion” day (P845, P1568 and P2230, respectively)

Full size image

Figure 4 shows three typical days for “sunny”, “cloudy/rainy” and “inversion” weather conditions. On the “sunny” day (observed on 1 February 2012), cloud formation was only little, solar radiation exceeded 1000 W m−2 at midday, while temperature and vpd exhibited similar daily courses. Dew point temperature (indicating absolute air humidity) varied only little at all sites but decreased with altitude. On the “cloudy/rainy” day (22 February 2012), solar radiation didn’t exceed 200 W m−2, air and dew point temperature stayed almost constant throughout the day, vpd was below 4 hPa, with about 122 mm rain measured at P1568 (33 mm at P2230). During the “inversion” day (10 June 2012), the temperature gradient along the transect is inverted (caused by a “Friaje”) with cloud formation at lower altitudes. This situation was observed on 42 days (35 of them during dry seasons) and usually lasted for 2–5 days. The temperature drop during these days is predominantly noticeable in lower areas, whereas it is hardly recognizable in the uplands, likely because cold air flows near the ground.

Between April and August 2015, total sum of rainfall increased from 655 mm at Puerto Inca and 849 mm at P251 to 1001 mm at P845 and 1661 mm at P1568 (Fig. 3d, open circles). Towards the peak of the mountain range at P2230, total precipitation appeared to have decreased (1339 mm). The actual amount of rain at the exposed elfin forest site was probably higher but not collected by the rain gauge, as strong winds might have carried rain drops horizontally over the opening of the gauge (Nešpor and Sevruk 1999). In general, precipitation was higher at the eastern slope compared to the western slope (Fig. 3d). Seasonality is particularly noticeable with respect to precipitation, as during rainy seasons total precipitation accounted on average 2.6 times the precipitation of dry seasons (Puerto Inca, 200 m asl, 1993–2009). Mean daily temperatures differed only slightly between seasons, but lowest temperatures were predominantly observed during dry seasons when “Friajes” occurred. During dry seasons, relative air humidity is lower and mean vpd higher. Mean solar radiation at elfin forest ridges is slightly larger during dry seasons likely because there is less water vapor in the air.

Comparison of satellite estimates and ground-based records at the Cordillera El Sira

In contrast to rain gauge observations, satellite based rainfall estimates from TRMM PR exist continuously since 1997 for the research area. Figure 5 shows the monthly precipitation between March 2000 and December 2015 at Puerto Inca. Rain gauge records and TRMM PR satellite estimates are highly correlated (linear regression: y = 0.83x − 1.43, R2 = 0.72, p < 0.001) and correlation is slightly higher during rainy seasons as compared to dry seasons. TRMM PR estimates mostly overestimate ground-based measurements of Puerto Inca, especially during dry seasons. Maximum seasonal precipitation for both time-series was recorded during the rainy season 2003–2004 (2060 mm gauge and 2084 mm TRMM PR) and minimum was recorded during dry season 2008 (453 mm gauge and 652 mm TRMM PR).

Fig. 5
figure 5

Comparison of satellite precipitation estimates (TRMM PR 3A42) and ground-based rain gauge measurements for Puerto Inca from March 2000 to December 2015 (left), during dry (middle) and rainy seasons separately (right). Grey shaded areas represent rainy seasons. Dashed lines indicate 1:1 relationships

Full size image

For the altitudinal transect there are only a few months of precipitation records available between 2011 and 2015. Although one grid cell of the TRMM PR data covers about 28 × 28 km2 and therefore large elevational differences (170–1945 m asl for Puerto Inca and P845, and 170–2443 m asl for P1568 and P2230), monthly precipitations are significantly correlated (R2 = 0.74, p < 0.01 at P845 and R2 = 0.78, p < 0.01 at P1568), but actual rainfall totals are generally underestimated by TRMM PR estimations (Fig. 6; Fig. 3d, open triangles), as they represent the average rainfall over an area of large elevational differences. Only at 2230 m asl correlation is weak (R2 = 0.21, p = 0.13).

Fig. 6
figure 6

Comparison of satellite precipitation estimates (TRMM 3A42) and ground-based rain gauge measurements along the elevational gradient. Dashed lines indicate 1:1 relationships

Full size image

A cross-correlation matrix of rain gauge records and TRMM PR estimates for daily, monthly and seasonal data (Table 3) shows that (1) quality of correlations increases from daily over monthly to seasonal records; (2) monthly and daily rain gauge records of Puerto Inca and along the altitudinal gradient are significantly correlated (p < 0.01); (3), TRMM PR estimates on seasonal (p < 0.001), monthly (p < 0.01) and daily (p < 0.001) basis are significantly correlated among each other; and (4) rain gauge records and TRMM PR estimates of Pucallpa, which is located about 140 km NE, are significantly correlated with data of our study area. Quantitative info, however, should be taken with caution.

Table 3 Cross-correlation (Pearson) of daily, monthly and seasonal precipitation sums of satellite estimates and rain gauge observations
Full size table

In order to identify rainfall anomalies between 2010 and 2015, deviations from the long-term precipitation pattern between 2000 and 2015 (Fig. 7) are shown in Fig. 8 for Puerto Inca. Deviations calculated on rain gauge and TRMM PR basis are highly significantly correlated (p < 0.001) and show excessive rainfall during December 2012 (+ 234 mm for TRMM PR estimates and + 309 mm for rain gauge records). The largest negative deviations were observed during the rainy season from October 2013 to March 2014. Particularly during November 2013 (− 105 mm for TRMM PR and − 70 mm rain gauge), December 2013 (− 176 mm for TRMM PR and − 211 mm rain gauge) and February 2014 (− 214 mm for TRMM PR and − 183 mm rain gauge) it rained unusually little.

Fig. 7
figure 7

Mean monthly precipitation at Puerto Inca (2000–2015, 200 m asl), calculated from rain gauge records (solid line) and from TRMM PR rainfall estimates (dashed line). Grey shaded areas represent rainy season

Full size image
Fig. 8
figure 8

Monthly rainfall deviations between 2010 and 2015 from the long-term precipitation regime at Puerto Inca (Fig. 7). Grey shaded areas represent rainy seasons. Correlation of time series: r2 = 0.63, p < 0.0001

Full size image

The rainy season of 2013/14 was extraordinarily dry, as only 79% according to TRMM PR estimates and only 73% according rain gauge records of long-term mean precipitation were received. Obviously, satellite based rainfall estimates can be used to identify precipitation anomalies, but rainfall totals for uplands with large elevational differences should be taken with caution.

Altitudinal gradient of biotic parameters

Altitudinal gradients in climatological parameters lead to floristic, physiognomic, structural and ecological gradients in vegetation (Fig. 9 and Table 1). Tree taxa diversity, diameters at breast height (DBH) and total tree heights decrease with increasing altitude (Fig. 9 a, b, see also Valenzuela et al. 2015). Aboveground dry matter culminates at the pre-montane forest site P845 (Fig. 9c) and total basal area follows the same course, but peaks at the mid-successional cloud forest site P1568. Here the large portion of light wooded pioneer species, predominantly Heliocarpus americanus with 14.3%, Cecropia angustifolia with 11.3%, Hyptidendron arboreum with 4.7% and Sapium glandulosum with 1.2% of trees, contributes more to total basal area than to dry matter.

Fig. 9
figure 9

Altitudinal gradient in vegetation parameters. a Number of tree taxa (DBH ≥ 10 cm). Triangles: species; circles: genera; and diamonds: families. b Boxplots of DBH (open boxes and circles) and tree heights (filled boxes and dots). Boxes indicate the interquartile range (IQR), dots and circles averages, whiskers extend 1.5 times the IQR from boxes, “+” indicate outliers. c Above-ground dry biomass (filled squares) and total basal area at breast height (open diamonds). d Mean seasonal stem radius change of individual trees during dry seasons (open circles) and rainy seasons (dots). Squares indicate the respective average. Dotted line represents linear regression for dry season (y = − 0.015/100x + 0.866, r2 = 0.77, p = 0.12) and dashed line for rainy season (y = − 0.046/100x + 1.464, r2 = 0.96, p = 0.02). a–c refer to whole inventory RAINFOR plots of 1 ha

Full size image

Mean annual radial stem growth decreased with rising elevation from 2.02 mm a−1 at the lowland forest site (max = 5.47 mm a−1 in Pachira insignis, Moraceae), to 1.72 mm·a−1 at the submontane forest site (max = 4.43 mm a−1 in Dacryodes nitens, Burseraceae), 1.44 mm·a−1 at the late successional cloud forest site (max = 3.47 mm a−1 in Anthodiscus amazonicus, Caryocaraceae) and to 0.97 mm a−1 at the elfin forest site (max = 2.97 mm a−1 in Cinchona hirsuta, Rubiaceae). At the mid successional cloud forest site, however, an exceptionally high mean annual radial growth rate of 3.15 mm a−1 (max = 6.5 mm a−1 in Hyptidendron arboreum, Lamiaceae) was observed due to the high portion of pioneer species.

During dry seasons stems still gain in radius at all sites, but differences between altitudes were not detectable, whereas they were clearly found during rainy seasons as mean radial stem growth decreased by 46 μm per 100 m in altitude (Fig. 9d).

Tree growth patterns in response to season and precipitation anomalies along the altitudinal gradient

Continuous monitoring of radial stem growth permits the identification of different intra-annual growth patterns, shown in Fig. 10 and listed in Table 2. “Seasonal” are trees showing annually repeated periods of radial increment predominantly during rainy seasons (“sr”, e.g. Hieronyma duquei) or during dry seasons (“sd”, e.g. Tachigali sp.) followed by a period of stagnation in growth. “Continuous” radial increment (“c”) was predominantly, but not exclusively, observed for elfin forest trees, where no stagnation in radial increment was detected (e.g. Anthodiscus amazonicus). If radial increment appeared to be continuous but was significantly larger during rainy or dry season the letter “r” or “d” was appended, respectively. “Indifferent”, “i”, also called “supra-annual” in Newstrom et al. (1994) is defined for trees exhibiting periods of enhanced growth and periods of less or no growth, regardless of seasons (e.g. Licania heteromorpha).

Fig. 10
figure 10

Examples of intra-annual growth patterns: Strictly “seasonal” growth (predominantly growing during rainy seasons), “continuous” growth throughout seasons without periods of stagnation, and “indifferent” increment, meaning that trees exhibit periods of growth and stagnation, regardless of season. 100% corresponds to the maximum radius measured at the end of observation period. Grey shaded areas represent rainy seasons

Full size image

The majority of monitored trees could be categorized as “sr” (Tables 2, 4) which were most abundant at the lower sites (P251, P845) and also well represented at the cloud forest (P1391, P1568) but completely absent at the elfin forest (P2230). The second largest group were trees growing “indifferently” (“i”) with largest proportions at the cloud and elfin forest sites. “Continuously” (“c” and “cr”) growing trees were found at all sites but especially at P2230. Only one tree at each P1391 and P2230 could be categorized as “sd”. Trees growing predominantly during rainy seasons (“sr” and “cr”) had on average larger stem radial growth rates (src_d) but more than half of them showed significantly (p < 0.05) reduced growth rates during the dry rainy season 2013/2014 (column [2] in Table 4). Trees categorized as “i” had on average the smallest src_d and only few were negatively affected (significantly less growth) by the dry rainy season 2013/2014. Most trees (13 out of 19) that were negatively affected by the dry rainy season 2013/2014 were predominantly growing during rainy seasons and present at all sites except for the elfin forest. Here, in contrast to other sites, most individuals (7 out of 11) showed larger radial growth rates during dry seasons, however, but significantly only in trees #2620 (Miconia sp.) and #2827 (Cinchona hirsuta). Only the latter was categorized as growing seasonally and predominantly during dry seasons (“sd”), all other trees were categorized as “c” or “i”.

Table 4 Overview of seasonal growth patterns and distribution along the transect
Full size table

Considering all trees along the altitudinal transect, respective site means of src_d during rainy seasons was significantly larger than during dry seasons except for the exposed elfin forest (Fig. 11, P2230). The dry rainy season of 2013/2014 significantly decreased src_d as compared to previous and following rainy seasons, although not significantly at P1568. Mean src_d at the elfin forest (P2230) was even larger than during other rainy seasons.

Fig. 11
figure 11

Stem radial change in µm day−1 (src_d) of all trees at the respective site during dry seasons, rainy seasons and the rainy season of 2013/2014 (October to March). Vertical bars: interquartile range of src_d; Cross: median; Diamond: mean. Different letters indicate significant differences at p < 0.01

Full size image

For some species it was possible to monitor more than one individual, as for Aspidosperma parvifolium (P251, #3501 and #3507), where only the larger and faster growing one (#3507) could be categorized as “c”; its growth was significantly (p < 0.01) reduced during rainy season 2013/2014. Individuals of the genus Eschweilera (P845, #223, #232 and #374) could all be categorized as “sr” and were significantly (p < 0.001) negatively affected by the dry rainy season 2013/2014. Both individuals of Dacryodes nitens (P845, #87 and #366) could be identified as “cr” and their growth was significantly (p < 0.01) reduced during rainy season 2013/2014. Licania heteromorpha (P1391, #1196, #1233, #1382 and #1395) could be categorized as “i”, except from the largest and fastest growing individual that had to be identified as “sr”. Dendrobangia boliviana (P1391, #1005, #1041 and #1250) was also heterogeneous in growth pattern (“i” and “sr”) and only growth of the largest individual (#1041) was significantly (p < 0.001) reduced during the dry rainy season. Similarly, Hyptidendron arboreum (P1568, #1713, #1866 and #1904) could be categorized as “sr” and “i” and, again, only the largest and fastest growing individual showed a significant (p < 0.001) growth reduction during rainy season 2013/2014.

Discussion

Applicability of satellite rainfall estimates for the El Sira uplands

Central Amazonian montane forests are poorly populated and accessibility is limited or in the case of El Sira extremely difficult. Therefore, networks of meteorological stations are sparse or absent, but space-borne precipitation measurements could overcome this lack of information.

Bookhagen and Strecker (2008) used the TRMM PR 2B31 product with higher spatial (5 × 5 km2) but lower temporal resolution in order to analyze rainfall gradients along the eastern flank of the Andes, although Espinoza et al. (2015) report that TRMM PR (2B25) underestimates rainfall by 35 percent in rainfall hotspot regions. Manz et al. (2016) reported low performance and high uncertainty for areas with low and intermittent precipitation regimes (e.g. the Peruvian Pacific coast).

On the other hand, rain gauge measurements represent only point source estimates of actual rainfall, as observations on small scales can differ largely within a landscape especially at low rainfall intensities (Habib et al. 2001; Ciach 2003). At exposed sites of high wind speeds, rain drops may be carried horizontally over the openings of rain gauges and may not be recorded (Nešpor and Sevruk 1999) thus actual rainfall is underestimated.

Our study shows that the TRMM PR 3B42 product generally overestimated total rainfall at Puerto Inca (200 m asl) east of the Cordillera El Sira but underestimated rainfall along the altitudinal gradient, likely as a consequence of its low spatial resolution. One grid cell covers an area of about 28 × 28 km2 and includes areas between 170 and 2443 m asl, therefore it represents the average over a large elevational gradient. However, temporal variations in rainfall on daily and monthly basis were well correlated along the transect and between rain gauge observations and satellite estimates (Table 4). It could be shown that the TRMM PR 3B42 product can be used to identify monthly and seasonal rainfall anomalies for the study area as was done for the unusually dry rainy season from October 2013 to March 2014.

Radial tree growth along tropical elevational gradients

Plant species diversity (Gentry 1988; Lieberman et al. 1996; Aiba and Kitayama 1999; Kessler 2002), tree stature (Grubb 1977; Cavelier 1996; Lieberman et al. 1996; Aiba and Kitayama 1999) and radial tree growth (Bräuning et al. 2008) generally decrease concomitant with higher elevation above sea level. Reduced tree growth can be partly explained by changes in species composition, since most fast growing pioneer species occur at lower altitudes (Homeier 2004), but it can also be a consequence of limited nutrient supply (Soethe et al. 2008), since lower temperatures, reduced transpiration (Küppers et al. 2008) and water-saturated soils inhibit mineralization. Furthermore, it is reported that leaf area index decreases with altitude (Moser et al. 2007; Unger et al. 2013), associated with lower canopy carbon gains and an increase in root to shoot ratio (Leuschner et al. 2007). The consequence is that less assimilated carbon is invested in aboveground biomass, leading to reduced stem growth. Additionally, Graham et al. (2003) reported that frequent cloud cover limits growth of rainforest trees, reducing radiation for photosynthesis. Motzer et al. (2005) and Küppers et al. (2008) demonstrated for montane tropical trees that stomatal conductance and transpiration are reduced which in turn lead to reduced sap flow.

The radial growth rates reported here are generally comparable with data reported for other tropical lowland and montane forests. For moist lowland forests in Panama, Lang and Knight (1983) reported annual diameter growth rates of 0–11.5 mm, Clark et al. (2003) noted 0.9–14.1 mm a−1 for Costa Rica and Laurance et al. (2004) measured between 0.25–7.01 mm a−1 in Brazil. Within the lowland forest site P251, we report a mean annual diameter growth between 4.04 and 10.94 mm.

Growth rates in tropical montane cloud forests are generally lower, e.g., Herwitz and Young (1994) reported diameter growth rates of 2.7 mm a−1 at montane rain forests in Australia, Bellingham and Tanner (2000), however, only measured rates between 0.4–0.9 mm a−1 in Jamaica. We observed means of 3.44 mm (submontane forest, P845) and 2.88 mm (montane cloud forest, P1391), although at the pioneer dominated cloud forest site (P1568) an exceptionally high average of 6.3 mm a−1 was measured.

Growth rates in tropical elfin forests are rather small, with rates of only 0.3 mm a−1 (Weaver et al. 1986) and 0.25 mm a−1 (Walker et al. 1996) both reported for elfin forests in Puerto Rico. The elfin forest site sampled here (P2230) showed growth rates of 1.84 mm a−1, which is comparable to those reported by Holder (2008) with 2.0 mm a−1 in Guatemala.

Seasonal growth and drought sensitivity of radial increment along the altitudinal gradient

Over the last decades the Amazon basin has experienced several prolonged droughts (particularly in 2005 and 2010) which resulted in increased tree mortality (Phillips et al. 2009) and reduced carbon uptake (Feldpausch et al. 2016). Tropical montane rainforests, however, generally receive vast amounts of rainfall throughout the year and rarely experience severe soil water deficits (Lyford 1969; Herrmann 1971; Kapos and Tanner 1985; Bruijnzeel et al. 1993).

Our study indicates that both, trees of lowland and montane cloud forests predominantly grow during rainy seasons, and that the exceptionally dry rainy season from October 2013 to March 2014 reduced tree growth even in montane cloud forests. Seasonally growing trees tend to be more sensitive to droughts during rainy seasons and were more abundant at lower sites but completely absent from the elfin forest. These trees are presumably adapted to high rainfall amounts during rainy seasons and seasonal variability in growth is dependent on seasonal allocation of carbohydrates. We assume that most seasonal growing trees are mainly flowering and fruiting during dry seasons, being partly defoliated (reliably observed for Ocotea ovalifolia P845, #227). They concentrate their vegetative growth on rainy seasons when rainfalls are abundant and usually reliably frequent. The onset of growth may be synchronized with first heavy rainfalls and replenished soil water content; when, however, rainfalls stay absent radial growth will be delayed and prolonged dry periods lead to stem shrinkage and restrict stem growth.

At the elfin forest, in contrast, radial stem growth was comparatively larger during dry seasons and was particularly enhanced during the dry rainy season of 2013/2014. This is likely a consequence of reduced cloud immersion and, therefore, increased solar radiation available for photosynthesis (Graham et al. 2003). Soils of elfin forest formations in the uplands of the El Sira Communal Reserve are generally water-saturated, due to high annual rainfalls, and trees rarely experience severe droughts limiting radial increments.

Within lower altitudes and up to cloud forest formations at about 1500 m asl along the western slope of the northern El Sira uplift, however, prolonged droughts limit stem radial increments. Therefore, future shifts in species composition along the El Sira uplift may be a consequence of more frequent and/or intense droughts over the century, since many neotropical taxa are restricted to wet areas (Esquivel-Muelbert et al. 2016).

Within single taxa, detected growth pattern was not necessarily homogenous, but larger individuals tend to have larger growth rates and growth was more reduced during the dry rainy season 2013/2014. Additional information on wood anatomy could complement studies on drought sensitivity of trees as vessel diameter and frequency drives vulnerability to stem hydraulic dysfunction through embolized vessels (Tyree et al. 1994). Large vessels permit a high stem water flow and high canopy transpiration rates, allowing on the one hand high rates of carbon assimilation if sufficient soil water is available. On the other hand, the hydraulic system is more vulnerable to failure through water stress-induced embolism, leading to stem shrinkage and restricting stem radial growth. Small vessels can limit water supply to leaves during phases of high solar radiation and a high vpd, forcing stomata to reduce conductance, by this limiting carbon assimilation, however, vessels are less susceptible to water stress-induced embolism. Therefore, data on anatomy has been collected and will be presented in a future study.

Conclusion

Satellite estimates of precipitation with high temporal and low spatial resolution, like the TRMM PR 3B42 product, can be applied to identify monthly and seasonal rainfall anomalies. Quantitative estimates should be taken with caution at areas of large elevational differences within small distances, as for the Cordillera El Sira.

The rainy season from October 2013 to March 2014 at Puerto Inca was exceptionally dry with only 73% (79% for TRMM) of the mean seasonal precipitation.

Cold fronts from the south advancing into the western areas of the Amazon basin (locally called “Friajes”) do not affect uplands above 1500 m asl at the Cordillera El Sira.

Various radial tree growth patterns along the altitudinal gradient were identified (seasonal, continuous and indifferent growth). Stem growth during rainy seasons was generally larger than during dry season at all sites except from the elfin forest site. Trees within the elfin forest at about 2230 m asl predominantly exhibited a continuous increment throughout seasons.

The exceptionally dry rainy season of 2013/2014 resulted in reduced radial growth of trees at all, but the elfin forest site. Here, trees appear to profit of fewer cloud immersion and thus radial increment was enhanced and similar to growth rates during dry seasons. Seasonal growing trees tend to be more sensitive to droughts during rainy seasons.