Do climate changes influence dispersal and population dynamics of dragonflies in the western Peruvian Andes ?

For nine dragonfl y species (fi ve aeshnids and four libellulids) all previous and verifi able data are related to the vertical climate zones and nature regions of the western Peruvian Andes and the Peruvian Pacifi c coast. Climate changes due to the El Niño and La Niña phenomena, as well as the global climate change have an infl uence on the diff erent natural regions and also restrict aquatic biotopes. Th ese changes infl uence the dispersal and behavior of some dragonfl ies and concern also loss of habitats as well as alterations of biotic and abiotic factors at and in water. However new waters and habitats also are formed in most nature regions. Specialists like Rhionaeschna peralta, a species of high mountain regions and the Puna, are not able to react to habitat losses by adaptation, while other species such as R. maita and R. marchali do colonize new habitats also in higher altitudes. While the here represented aeshnids change their distribution ranges within the vertical nature regions of the west Andes, this is suspected for three of the four libellulids (Orthemis ferruginea, O. discolor and Pantala fl avescens) as latitudinally respectively longitudinally immigrations and expansions of their areals. For all species discussed, a seasonally earlier fl ight beginning is detectable, but for no species an extension of their fl ight time. Altogether, the above named three libellulid do react more fl exibly and faster to the alterations by climate changes than the majority of the fi ve aeshnid species. Th e infl uence of increased UV-B and UV-A radiation possibly aff ects also the site occurrence of some species in high altitudes of the Andes. BioRisk 5: 47–72 (2010) doi: 10.3897/biorisk.5.842 http://biorisk-journal.com/ Copyright Joachim Hoffmann. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. RESEARCH ARTICLE BioRisk A peer-reviewed open-access journal


Introduction
Until a few years ago, climatology was in Peru exclusively limited to examinations of the El Niño events and the consequences for agriculture and fi shery.With signing the Kyoto Protocol measurements and supervision of the CO 2 -emmissions were also included as part of the responsibility (SENAMHI 2006).In 1994 the national environmental council CONAM (Consejo Nacional del Ambiente) was brought into being (Congreso de la República 1994), which in 2001 with the submission of the fi rst report to the UN also fi xed the subject of climate change as the main focus of its work (CONAM 2001).It was not until 2003 that studies of the eff ects of climate changes on the biodiversity were incorporated in the conceptual formulations (CO-NAM 2004) -though the term "biodiversity" referred exclusively to cultivated plants and economically useful animal species.
To the present day there are only a few studies to the biodiversity of the western Peruvian Andes, and these concern botanical recordings and research on a few zoological taxa such as mammals, birds and butterfl ies which allow some statements to species diversity.Within the scope of such investigations by North American and Peruvian biologists since the beginning of the 90s of the last century, Odonata were also increasingly collected and deposited with the Natural History Museum (Museo de Historia Natural de la Universidad San Marcos) in Lima.Wider samplings in the coastal regions to 1400-m height and the high andean ranges between 3000 and 5000 ms were carried out since 1989 in irregular intervals by the author (Hoff mann 1991, in prep).However these investigations are up to now limited, however, to a pure sampling activity without studies to Odonata coenosis and its ecology.Th ese data and also the odonatological collections of the Natural History Museum in Lima, the University La Molina in Lima and the University of Cajamarca reveal for some species of the suborder Anisoptera changes in the dispersal behavior which is possibly caused of climatic changes within their distribution ranges.
Out of 481 dragonfl ies recorded in Peru (Hoff mann 2009), 46 species were sampled up to the present in the western Andes and the coastal desert.Some species from the families Aeshnidae and Libellulidae should be looked at more closely here, because at least enough data are available to be able to discuss changes due to climate change.

Ecological Regions
According to diff erent authors, the division of the Andes into ecological regions is variably handled.For example Brack (1985) distinguishes four terrestrial ecological regions in the West Andes and coast.Th e Instituto Nacional de Geografía de Peru defi nes for the same area of Peru fi ve regions (INGP 1989) and later with 14 subregions (in INRENA 1995).All three works are primarily based on precipitation and temperature data as well as plant communities.However, the subdivision into partly very small-fl at subregions (exclusively guided by plant communities) is at least not partially when viewed objectively comprehensible.
A fi rst division of the Peruvian Andes into nature regions was undertaken by Pulgar Vidal (1967) and was revised completely in the ninth edition in 1987.Pulgar Vidal divides Peru primarily in nature regions in accordance with: -altitudinale zones -relief -climate -vegetation -fauna -agriculture -(historic) folklore Th e nomenclature adapts the Indian population's original Quechua-names.Th e division of the West Andes (incl.coast desert) takes place accordingly in six regions (Table 1 and Fig. 1).
Temperature decreases with rising altitude, the precipitations increase at fi rst and then decrease also with sinking temperatures.Th is is very similar to the meridional course of the values of these climate elements in direction towards the pole (Table 2).
Long-term weather records in Peru before are only from few places and generally as from the 1980er years.Secular data series are only available from Lima.
Th e peculiarity of the mainland climate in the western part of the Andes is dictated by the Humboldt Current and El Niño events.Th e absence of the cold Humboldt Current is created by the El Niño event, which in turn causes a "natural" climate alteration of irregular intervals about Christmas time for several weeks occasionally.Th is ENSO-mechanism (El Niño / Southern Oscillation) stands in competition to the anthropogenic "greenhouse eff ect" (Arntz and Fahrbach 1991;Schönwiese 1995).
Th e El Niño event has the eff ect of producing hard showers in the coastal belt up to an altitude of 2.500 m (Quechua), while temperatures and precipitations at greater heights are infl uenced more weakly.

Meteorological data
It is very diffi cult to obtain weather records in Peru.Th e National Meteorological Institute (SENHAMI) functions under the auspices of the ministry of defense and weather data are provided only under certain prerequisites and then only against very high fees.Furthermore long-term series of measurements exist only from a few places.Th ese are often interrupted for longer periods of time due to faulty measuring instruments.Th e data used here were gathered predominantly from publications of SENHAMI, Internet presences and series of measurements by mining companies in the Andes.Older data were partly published in the "Atlas Del Peru" (Inst.Geogr.Nac.1989).Rare weather informations which are also taken into account can be found in older odonatological publications (e.g. Calvert 1956).
With the beginning of the CCA project in Peru (Hoff mann 2006) diurnal data at the sample waters were recorded by a mobile weather station 1 .Th ese however always represent only at hourly intervals values at a research locality for the duration of the stay.In addition to the weather data, the UV radiation (UV-A and UV-B) was also included.

Hydrological data
In 2006 all waters examined were analysed (abiotic factors), conductivity, pH, water hardness, temperature and oxygen content being measured.Unfortunately, the device for measuring oxygen already ceased functioning at the beginning of the examinations.For future measurments water level gauges were installed at the waters.
Water temperatures of some larger waters were often taken in the course of hydro geological examinations (e.g.CONAM 2005) and were taken from relevant publications.Other sources are found in publications of Peruvian limnologist and also among mine companies which from time to time measure water-levels of large lakes.Such data were provided by the national park offi ce Huascaran.

Dragonfl y data
In the present consideration literature data (Calvert 1956, Schmidt 1952, Ellenrieder 2003 and others) was used although only a very few publications take into account the odonata fauna of the western Peruvian Andes.Th e verifi cation of localities proved 1 Th e weather station were entrusted to the administration of the National Park Huascaran and remained for the purpose of long-term series of measurement at Pampacocha, a lake in the Puna (4,068 m of height).to be very diffi cult.Th e description of species and the publication of species lists as a rule were not made by the collectors themselves or by scientists on the spot, but by odonatologists in Europe or the USA.Locality labels in many cases have been assigned, interpreted or read wrongly 2 .Records which could not be verifi ed were not taken into consideration.Data from the entomological collections of the universities of Cajamarca and La Molina (Lima), as well as the Natural History's museum in Lima formed an essential basis.Th ese however required a thorough revision since more than 30% of the specimen from the western Andes had been wrongly identifi ed.Since 1994 Günther Peters (Berlin) thankfully accepted the examination of the aeshnids.
Furthermore, since 1989 the author regularly collected in the Peruvian western Andes, predominantly in the departments of Ica, Lima and Ancash (Hoff mann 1991 and in prep).Current data on odonata exist from the "Climate Change and Aeshnids -Project" in January and February 2006, August and September 2006 as well as February andMarch 2007 (CCA, Hoff mann 2006).

Climate
Since there is a historically natural infl uence on climate by the ENSO mechanism (El Niño / Southern Oscillation), this phenomenon will fi rst be described.

El Niño / La Niña
As an "El Niño" is called the appearance of an unusual, non-cyclic, change of currents in the oceanographic-meteorological system of the equatorial Pacifi c.
When it occurs, the cold Humboldt Current decreases along the Chilean and Peruvian coast until it succumbs.Th is happens due to a shift of the wind zones through which the normal west fl ow of near-surface (warm) sea water fl ows back to the east.Th e Walker-circulation has reversed itself, with the consequence that the east Pacifi c heats up.Its roll in the global warming heating is discussed controversially.While 2 For example: for R. elsia is given by Calvert (1956) and later by Ellenrieder (2003; assumed from Calvert) a locality "Huanuco".Today this place (see distribution map in Ellenrieder) is the capital of the same named department in the eastern slope of the Andes at a height of barely 2000 m.An unusual locality for a strong "coastal dragonfl y", which only occurs on the west side of the Andes.It must be assumed that there is a mistake in the later localization.Th us there is a small place called Huanaco on the coast in the department of Ancash and in the department of Arequipa likewise near to the coast a place called Huancan.From both places are collection records of R. elsia, as well as of other species.
el Niño is a natural climate phenomenon it is, nevertheless, supposed that it will be strengthened by the anthropogenic greenhouse eff ect.
A current study points out, that El Niño events, particularly big ones, can be predicted more exactly than was previously assumed (Chen et al. 2004).
La Niña is an extremely cold current in the equatorial Pacifi c in contrast to El Niño, the naming being based on what, so to speak, is an anti-El Niño.
La Niña usually follows an El Niño event being triggered by exceptionally high diff erences of atmospheric pressure between South America and Indonesia.Th is leads to stronger trade winds.Th e warm water is driven at the surface in the Pacifi c Ocean by the trade wind to South-East Asia.As a result cold water fl ows up from the depth of the Pacifi c on the coast of Peru.Th is upwelling water lies up to 3°C below the average temperature.
Th e consequences are not as strong as in the case of the El Niño, but La Niña nevertheless has a considerable infl uence: -in the west Pacifi c the water at the surface is warmer, -in South-East Asia La Niña brings rains and colder temperatures, -in South America it rains less and the deserts dry up.
Altogether, less natural disasters occur, however, than during an El Niño.
Th e number of La Niña events has decreased during the last decades, that of the El Niño events has increased (fi g. 2).Th e assumption exsists that the anthropogenic greenhouse eff ect is responsible for this situation.Up to now, this could however not be proved.
An El Niño event lasts for about twelve to eighteen months and has its acme between December and June.Indeed, there were -dependent on defi nition -two recent exceptions: one event lasted from the middle of 1990 to the middle of 1995 while the Niño of 1997/98 was clearly shorter (NOAA 2005).An El Niño infl uences the weather predominantly on the coast up to the upper Yunga, has however, also an infl uence up to the uppermost elevated regions of the western Andes.Changes can only be perceived and measured during the events, but also infl uence the climate before and after event, though only to a slight degree.For example the El Niño event from November 1997 to April 1998 caused in Pisco (Pacifi c coast) maximal precipitations of 212 mm in December, while in January the amount of precipitation sank to 38.7 mm.On the other hand, the amount of precipitation in the Chala lay at about the level of the previous years while in the Puna however it lay under those of previous years (fi g. 6).In May 1998 the amounts of precipitation lay in Pisco with 12 mm again in the range of the long-term averages.
Eff ects on the temperature development in El Niño years are noticeable up to the high regions (s.fi g. 5, year 1998).1997/1998 (fi g. 5).Th e decrease of temperatures of 1999 and 2001 were accompanied by a La Niña event within these years (cf.fi g. 2).

According to
Th e amount of precipitation decreased continuously in Chala and Quechua since 1997 / 1998.Th e amount in the Puna is very irregularly, with a not interpretable peak in the year 2001 (fi g. 6).
Th e relative atmospheric humidity in January is relatively constant in the Chala with an increase of only 3% on average over the last eleven years.In the upper altitudes of the Puna a continuous decline of more than 20% is remarkable, in the Quechua the values show great variations of up to 16% (Fig. 3-6).

Dragonfl ies (Aeshnidae and Libellulidae)
Interpretable data are available for the following species.It should be noted that Rhionaeschna maita is thought of as a good species in contrast to Ellenrieder ( 2003) who classifi ed it as a synonym of R. brevifrons 3 :

3
Rhionaeschna maita is represented here as a good species because there are some signifi cant diff erences to R. brevifrons.Possibly a "wrong" lectotype was synonymized by Ellenrieder (2003), then the actually published photos by Geller-Grimm (2005) of the putative holotype de facto show Rhionaeschna brevifrons and do not correspond to the fi rst description of R. maita by Förster (1909).Both Ris (1913) and Schmidt (1952) describe R. maita as closely related to Rhionaeschna marchali, to which resembles in many characteristics.All three species are sympatric in the upper Yunga and the lower Suni.Records and information before 1996: at least from 1982 to 1996 data were regularly collected in complete annual courses, also in the western Andes and the coastal region.
Records and information after 1996: samples in irregular time periods, with some years not sampled at all.
Current results from the CCA project begun in 2006 (Hoff mann 2006) are presented separately.

Aeshnidae
Rhionaeschna maita (Förster, 1909) Nature region: Until 2005 the species was recorded exclusively in Quechua and Suni at altitudes between 2,300 m and 3,900 m.Evidence is currently (2006) available from the Puna at 4,100 m.Some records from the Yunga and the lower Quechua must be ascribed to the sister species Rhionaeschna brevifrons on the basis of re-determinations of collection material.
Phenology: Obviously the fl ying time in the upper altitudes (more than 3,000 m) is from August to October, i.e. in the "andean spring" while at less than 3,000 m a second fl ight period from February to May ("andean autumn") is found.Habitat utilization: Till now the species was found only at mesotrophic, predominantly slow running and not too cold (> 15°C) waters.

Rhionaeschna absoluta (Calvert, 1952)
Nature region: Th e main distribution of R. absoluta is in Yunga and Quechua.As the only species of the genus Rhionaeschna it was recorded also from the eastern high Andes at 3,800 m.
Phenology: Th e species shows two fl ight periods which are similarly to those of R. maita in the "andean spring" and "andean autumn".Records from the Yunga in the months August to October are available only since 1999, but then regularly.
Habitat utilization: Except at strongly fl owing waters R. absoluta was found at all other types of waters.Eutrophic waters are preferred, oligotrophic waters avoided.

Rhionaeschna elsia (Calvert, 1952)
Nature region: Until now this species was found only in the arid coastal area of Peru.All fi ndings until 1996 are up to 500 m above sea-level.In 2001 R. elsia was found for the fi rst time in the Yunga at 900 m and in 2006 even at 1500 m.
Phenology: Obviously the fl ight times also have changed.Calvert (1956) indicates February to May as the period; later records (until 1996) cover a longer time from Oc- o migrating specimens tober to July and more recent records (1996 onwards) cover the period from the end of August till the middle of July, indicating a phenological omnipresence.
Habitat utilization: Former records were taken exclusively in standing, mostly brackish waters (coastal lagoons and ponds).Th e records from the Yunga (900 m respectively 1,500 m) come from stagnant waters or temporary ponds in dry riverbeds.In 2006 imagos and larvae were also collected from a slowly fl owing brook in the Chala at an altitude of 285 m.

Rhionaeschna marchali (Rambur, 1842)
Nature region: Th e main distribution of the species is in Yunga and Quechua between 1,900 m and 2,900 m.Currently (2006) R. marchali was also stated from the Suni between 3,500 m to 3,800 m.According to Calvert (1956) obviously the species also occurs on the eastern Andean slope.Th is, the only record from the 1930s could, however, not be verifi ed.Phenology: In the Quechua the species fl ies all the year round, except for November and December, the fl ight records for May to July having been taken before 1956.In Yunga and Suni the fl ight time is probably limited to the "andean spring".
Habitat utilization: Ellenrieder (2003) stated as habitats wet soils, crop pastures, muddy bottom pools, creeks, running brooks with aquatic plants and streams.Also according to current records (2006) the species occurs euryoeciously even in the Suni at 3,850 m, where water temperature of all waters sampled was between 15°C and 23°C.(Ris, 1918) Nature region: A defi nite "mountain aeshnid" which occurs only above the Yunga (2,300 m) and which reproduces as the only aeshnid up to the lower margin of the  (Schmidt 1956) Habitat utilization: Th e species occurs at wide brooks, permanent stagnant waters as well as mesotrophical ponds in riverbeds in Quechua and Suni.Sampling localities in the Puna were lakes, bog ponds and fl ood areas of artesian springs.In the upper Puna and lower Janca R. peralta was also found at stagnant areas of open thermal springs (temperature of emerging water 35°C, cooled down to 20°C at the outer edges).Th e temperatures of the water sample were between 16°C and 20°C; all waters being mesotrophic to oligotrophic.
Nature region: E. Cleopatra occurs mainly in the arid coastal area, however, ascends infrequently also to altitudes of 1,000 m.Since 1996 records from more than 2,000 m are on the increase.Current records ( 2006) of E. cleopatra come from the lower Quechua at 2,500 m.
Phenology: In the Chala (arid coastal area) the species fl ies from October to May, i.e. with the exception of the winter month, for over eight months of the year.Th e + records before 1936 (Schmidt 1956) Janca (records at 5,000 m; see Corbet 1999).Th e main distribution of the species in the Peruvian Andes lies in Quechua and Suni, i.e. between 2,300 ms and 4,000 ms.
Phenology: R. peralta is phenologically omnipresent at heights between 2,300 m to 4,300 m (similar to R. elsia on the coast).
fl ight time at altitudes between 500 m to 2,500 m, i.e. in Yunga and Quechua, does not extend for more than fi ve months, but beside this in August and September -in these months the species is missing in the Chala.
Habitat utilization: Th e majority of records in the area of the Chala and the lower Yunga is from stagnant waters and temporarily fl ooded ponds in riverbeds.In contrast to R. elsia, brackish waters with high salinity are obviously avoided.Current records from the upper Yunga at 2,000 m were taken exclusively from temporarily fl ooded ponds in riverbeds.In 2006 imagines and larvae were also obtained from a slowly fl owing brook in the Chala at an altitude of 285 m.

Orthemis ferruginea (Fabricius, 1775)
Th is species was recorded for the fi rst time in Peru also as reproducing within the scope of the CCA project in the second half of September 2006, the sampling locality being situated in the Yunga at 1,200 m.Th e site was a mesotrophical, permanent pond in the bed of the Rio Santa.

Orthemis discolour (Burmeister, 1839)
Th e species has its main distribution in Peru to the east of the Andes.
Nature region: Most records from the western Andes are from the northern Chala up to 350 m.Th ere are a few records from the northern Yunga at 850 m and 1,900 m.Currently O. discolor was sampled also in the lower Quechua at 2,500-m, far to the south of older record localities in the Yunga.
Phenology: Th e fl ight time in the Chala extend for more than fi ve months from the middle of October to the end of February.In the Yunga records are from May as well as from September and October after 1996.Currently (2006) the species was collected in September from the lower Quechua.
Habitat utilization: In the Chala O. discolor was recorded exclusively at mesotrophical to slightly eurotrophical standing waters, brackish coastal lagoons being avoided.Th e locations in the Yunga are mesotrophical standing waters in river beds + records before 1936 (Schmidt 1956) and backwaters.Th e records in the Quechua were taken at an oligotrophical to mesotrophical backwater in the upper reaches of a brook.

Pantala fl avescens (Fabricius, 1798)
Nature region: West of the Andes the species is autochthonous only in the Chala up to 500 m.Nevertheless there were some records also from other nature regions above 4,000 m which before 1996 supported exclusively migrating specimens.Currently (2006) the species was sampled in the Yunga as well as in Quechua up to 2,800 m, in the Yunga already in 1999 autochthonous.
Phenology: Th e species fl ies in the Chala (arid coastal area) from October to May, i.e. with the exception of the peruvian winter about eight months of the year.Up to now records in the Yunga are only for November and December.In the Quechua P. fl avescens was found exclusively in August and September (including reproductive records), i.e. at a time when the species does not fl y in its main distribution area western of the Andes.
Habitat utilization: In the arid coastal area P. fl avescens is euryoecious and only avoids brackish waters with a high salinity and strongly running waters.A mass emergence in a swimming pool with several thousand specimens was stated from Tacna in February 1994 (Hoff mann 1998).In Yunga and Quechua the species could be found in mesotrophical ditches and temporary areas of standing waters in river beds.

Climate change and abiotic sequence eff ects
In what way global warming and the consequences of El Niño events overlap or mutually increase each other is not clear up to the present.It is noticable at least that the intervals between El Niño years become increasingly shorter, while the number of the Niña events is decreasing.Since the variations of the SST (sea surface temperature) have increased in the tropical Pacifi c during the last 100 years, and an accumulation of El Niño events appeared, the question is whether a connection exists between El Niños and the anthropogenically caused greenhouse eff ect.In order to fi nd this out, Timmermann et al. (1998) have calculated and analysed a greenhouse simulation with a globally coupled ocean atmosphere model.It turned out that the eastern Pacifi c will warm up until 2100 by about 3°C.On the other hand the west Pacifi c will warm up by only about 1°C.Th is would mean that in future, situations similar to El Niño would much more frequently appear if the emission of greenhouse gases (primarily of CO 2 ) is not limited or becomes reduced.Increasing interannuale variability overlaie the longterm warming trend.
Consequences of intensive El Niño events in Chala and Yunga (0-2,300 m) are documented adequately by state authorities in Peru, because these regularly impair the economy immediately and at short notice, particularly as far as the fi shing industry and agriculture are concerned.Such documentations do not exist for the higher altitudes of the western Andes, possibly because the consequences do not play an important part here.
Th e climatic change due to global warming manifests itself much more strongly, although in El Niño years the change is hardly recognizable at the lower altitudes.Documentations of the consequences of the climate change are available only with regards to the melting of andean glaciers.Th e consequences although already recognizable and measurable, are shown merely as sequence scenarios.
In the mid 2006, a French-Peruvian project was started to now examine more intensely factors, progress and consequences of the gradual disappearance of the glaciers.
Analyses of satellite photos show for the Cordillera Blanca only, the largest tropical glacier massif, a reduction of the ice masses of 15% (corresponding to 13.5 km³) in the period 1970 to 1997 (27 years).During the last 35 years before 2005 (1970 to 2005) the reduction had accelerated and now, lies by 22% (CONAM 2005), an increasement of 7% during eight years.In subareas such as the Huascaran massif, this decline amounted even to just below 30% (CONAM 2001).According to UNESCO (2006) some smaller glaciers below a height of 5,000 m have already completely disappeared and others will be completely gone the latest by 2010.
Th e melting of the glaciers in parts of the upper nature regions from 3,000 m to 4,800 m (upper Quechua, Suni and Puna) has at fi rst sight provided a greater available of surface water, however, on the other hand, also a loss of the water reservoir of 12% compared to present day.
Th e melt water are led away by the rivers to the Pacifi c, with the result that these show from May to September higher water levels in downstream regions, also during the dry times in the Andes, where 20 years ago these rivers still fell dry at this time of the year.
At the same time, the amount of precipitation and atmospheric humidity decrease continuously from 500 m to 4,000 m altitude, while on the long term in all nature regions a rise in temperature is to be registered.Beck et al. (2006) indicate changes in the extend of the climate zones defi ned by Köppen as a consequence of climatic warming.Th us positive trends are to be noted in South America for temperate climatic zones (average temperature of the coldest month between 3°C and 18°C) and tropical climates (the coldest month less than 18°C), while dry climates (no certain cycles nor summer nor winter rains) show a long-term decrease.Similar decreasing trends show the cold climates (average temperature of the warmest month below 10°C).

Habitat losses and formations of new habitats
Th e climate changes cause in part considerable habitat losses both at upper altitudes and in the arid coastal zone.
Glaciers melting leads to more permanent wettings in the Puna as well as fl ooding of upland bogs as well as artesian and thermal springs.
Th us Rhionaeschna peralta could not be recorded any more at such waters and even disappeared from the highest up now recorded reproductive waters in the Jalca at 5,000 m (cf.Corbet 1999).Also at many Puna lakes the species was not found any more after 2001.Th e reasons for this could be a noticeable increase of the water-level which led (apart from an anthropogenic use) to the loss of the Totora reed (Scirpus sp. and Juncus sp.) as oviposition and larvae habitats.However, at other lakes in which the water levels have fallen during the last years due to the increased melting of the glaciers supplying them with water the abundances appears to have increased (e.g.Lake Titicaca).
A further reason for the disappearance of species is possibly also the increase in water temperatures of some lakes in the Puna (e.g. in the Cordillera Negra).Thus a surface temperature (at a depth of 20 cm) of 21°C was measured in the Laguna Ututo (at 4,464 m) in 2006 (Hoffmann 2006).Here these waters were not or no longer fed by glacier waters.The centre of the distribution of R. peralta therefore lies at present between 2,300 m and 4,000 m in Quechua and Suni, in which the species apparently could profit in these regions from the formation of new habitats in form permantly water-bearing ponds and stagnant waters, for reasons of the steady melt water run-off in river beds profited, as long as these were mesotrophical.
For Rhionaeshna maita, with regard to the vertical distribution, the same applies as for R. peralta, with respect to their center of distribution in Quechua and Suni.While the latter appears to retreat, however, slowly from the Puna (see above), R. maita conquers at present waters in the lower altitudes of this region (up to 4,100 m).Apart from slowly fl owing and mesotrophic run-off s of backwaters are preferred.As a rule, the temperatures of these waters do not lie below 15°C.Such waters appeared within the last few 10 -15 years.Th e species thus occupies a newly arisen niche, which R. peralta due to its ecological claims -up to now it has never been recorded from running waters in the Puna -obviously cannot taken up.With increasing warming and as long as a water surplus will be available, R. maita will possibly ascent to higher altitudes in future.
A further aeshnid which appears to profi t from climatic changes is Rhionaeschna marchali.It has expanded its area of distribution from the Yunga and Quechua into the Suni and appears likewise to be euryoecious.Conspicuous for this species are mass appearences of the population, above all at standing waters in the Quechua, something that can possibly be explained due to increasing water temperatures (see below).
Similar trends in vertical distribution can also be ascertained in arid coastal region.Both Rhionaeschna elsia and Erythrodiplax cleopatra have extended their distribution into higher natural regions with R. elsia up to the middle Yunga at 1,500 m and E. cleopatra even up to the lower Quechua at 2,500 m.Both coastal species appear to profi t from changed conditions, particularly with respect to suitable new habitats in the higher regions.On the other hand it is remarkable, that in particular R. elsia is no longer present at some coastal lagoons, but has recently appeared in slow running waters.Obviously the occurrence of this species is tied to the existence of slightly ionized waters (salts, conductivity values between 1.5 S/m and 3.9 S/m)4.Such condition of waters are at present increasingly full fi lled and areconverted into usable lands since particularly the need for agricultural products has increased and a permanent water regime of the rivers makes irrigation throughout the year possible as a result of the climate change.Furthermore it can be supposed that rising water temperatures and evaporation rates of the coastal lagoons also lead to habitat losses.
An example of a specie for which no changes in its vertical distribution can be ascertained is Rhionaeschna absoluta.Th e surprise is greater, because occuring exclusively in Yunga and Quechua-the species is considered as euryoecious and has been recorded up to now as the only species of the genus Rhionaeschna also in the eastern Andes.

Changes of water temperatures
Measurement series of the last 25 years of diff erent types of waters prove a general increase of water temperatures of around up to 18% (max.3.6°C) in the nature regions above the Chala and even up to 27.5% (max.9.1°C) in a few small waters of the Chala.Some newly formed bodies of water by increased water effl uents of the Quechua and Suni show temperatures of up to 25°C.In very small coastal lagoons of the Chala up to 29°C were measured.No dragonfl ies could be found in such waters.Some species, in particular aeshnids, give up waters in the upper altitudes with temperature rises of more than 21°C.Th e only aeshnid which is an exception of this, and which appears to settle such waters is Rhionaeschna marchali.
On the other hand the majority of the species appears to profi t from an increase in water temperatures in upper nature regions.Either they migrate due to changing conditions in their original reproduction waters to higher regions or they colonized new 4 For comparison purposes: seawater shows a conductivity of approx.5 S/m, natural incidences of dissolved copper salts up to 5800 S/m habitats without giving up their original waters in the lower regions, as for instance Orthemis discolor and Pantala fl avescens.

Phenological adaptation
For the majority of the species discussed here, a change of fl ight times is noticeable after 1996, compared with former data.All species have fl ight times during the southern summer season with a beginning in "late spring".According to the altitudinal zone this beginning is "brought forward" during the last years at the end of the winters.Species which started to fl y formerly only late in the seasonal course can already be found in spring (e.g.R. marchali in the Quechua and E. cleopatra in the Yunga).In connection with this it has to be pointed, that "winter" and "summer" -particularly with the exception of the Chala -have to be defi ned over the amounts of precipitation, dry and respectively rainy season.While temperatures rise above the Yunga in the longter mean, precipitations decrease over the same period of time (cf.fi g. 5 and fi g. 6).Other species such as R. elsia in the Chala and R. peralta between 2,300 m and 4,200 m (Quechua to Puna) fl y recently over the whole year and are phenologically therefore omnipresent.Most species above 1,000 m are not fl ight-active against this during the months in June and July, with the lowest precipitations of the year.Th ere was one exception, however in July 2006, in a really dry summer with precipitations of less than 5 mm, already started in May (normally long-standing monthly means in May with 75 mm).Th is resulted in a mass emergences of Rhionaeschna sp. in the Quechua, with thousands of dragonfl ies appearing in the streets of Huaraz (3,048 m) for several days.Inquiries through the local press and local inhabitants showed that it concerned without questions aeshnids, presumably Rhionaeschna marchali.Obviously there was a mass emergence briefl y before with following migration events, similar to described in Europe for Hemianax ephippiger (Peters and Günther 2000;Günther 2005) and currently for Aeshna mixta (Günther 2006).As a cause it is to supposed that early beginning dryness and relatively high temperatures (monthly mean of daily temperatures of 27°C instead of 24°C of a long-standing mean) led to a drying up of standing waters with a low water-level.Because R. marchali starts to fl y very early in the year (according to older data of Calvert 1956, the complete wintertime could not be confi rmed until, however), it possibly concerned the phenomenon of an emergency emergence of the last larvae instar.Moreover the species fl ies at least in two generations.
Currently, changes in the phenology have become known for some European species, even if they probably still represent the exceptions (e.g.Sympetrum fl aveolum, Kunz 2006).
On the other hand such "prolongations" of fl ight times as reported by Busch-Nowak (2002) about the very late emergence of Libellula depressa in October in Baden-Wurttemberg (Germany) and by Hoess (2006) about very late fl ight times of Coenagrion puella in Switzerland are not ascertainable from the western Peruvian Andes in an analogous way up to now.

Immigration
Immigrations from distribution areas in distinct latitudinal climatic zones such as described by Ott (1996 and2000) for Crocothemis erythrea are noticeable in the western Peruvian Andes at least for Orthemis ferruginea.Paulson (pers. comm. in Heckman 2006) considers Orthemis discolor as an own species and not conspecifi c to Orthemis ferruginea.He gives North and Central America as the distribution area of Orthemis ferruginea.Donnelly (1995) indicates that O. discolor has dispersed from South America along the Caribbean side to Florida whereas O. ferruginea coming from North America dispers along the pacifi c side to Costa Rica.Both species are sympatric in parts of Mexico and Costa Rica (fi g. 8) and sometimes even occur at the same waters.
On checking the collection material from Peru, some years ago, it was discovered that all specimens (with one exception) determined as O. ferruginea, had to be reassigned to O. discolor.Merely a male from the northern Yunga collected in August 2001 belonged to O. ferruginea.Later an ovipositing couple was caught in September 2006 about 400 km south of the fi rst location, also in the Yunga.Possibly the species is distributing along the Pacifi c coast in a southerly direction which means that it could also occur in Ecuador and Colombia.At present, the distribution of O. ferruginea still appears to be discontinuous.
Obviously the equatorial climatic zone extending for over 1,000 km from Costa Rica to Colombia has been up now a barrier which prevented dispersals to the south.Beck et al. (2006) show climatic changes in the pacifi c coastal areas, in that the equatorial climate is being taken over by a tropical climate.Th ese changes possibly favour the immigration of O. ferruginea into western South America.At least here are climatic conditions at the locations in the Yunga similar to the main area of the species in the western Central America.
Possibly Pantala fl avescens and Orthemis discolor take other ways of the immigration.Th e latter occurs in Peru predominantly east of the Andes in the equatorial climate zone and the changing tropical climate zones.O. discolor was, however, already recorded before 1936 in the northern and middle Chala, as well as after 1956 in the northern Yunga.Th e current records in Yunga and Quechua allow three options of dispersion: a longitudinal immigration from the northern distribution in the Yunga, a vertical dispersion from the Chala and thirdly immigration from its main distribution area east of the Andes.For an origin from the main distribution area speaks the fact, that O. discolor is stenoecious at the new locations, as it sometimes seems for some euryecious species at new settlements outside their eurytopic areas (Lattin 1967).Th is is not really to be expected with immigration from Chala and northern Yunga.
Pantala fl avescens shows a similar dispersion pattern.Th e area of distribution of this species extends in South America up to the 20°C isotherms (annual mean temperature).Accordingly it occurs autochthony up to now only in the Chala and after 1998 also in the lower Yunga.Findings at high elevation sites of the Andes concerned exclusively migrating specimens.Th e current fi ndings in the Yunga up to the Quechua at 2,800 m are from sheltered valleys (e.g.middle Santa and middle Pisco valley).Th e increase of the annual mean temperature, especially in higher altitudes was greater than the average for the whole country.Th us the annual mean for the temperature has increased in the Canyon del Pato from 18.6°C before 1980 to 19.7°C after 2000 (comm. of the data by Minera Huallanca S. A.C. 2006).Accordingly it must be supposed that the immigrations occur vertically.Th e species probably will be able to establish itself up within the next years up to the upper Quechua, particularly as it can reproduce successfully because it has a short larval development of three to fi ve months (cf.Corbet 1999) and because of the formation and warming of some waters to more than 18°C (see above).

Conclusions and open questions
Th e present data provide merely indications for the infl uence of the change of climate on dispersion and ecological adaptation of dragonfl ies to changing environmental conditions in the western Peruvian Andes.In most cases these data do not suffi ce for a statistical verifi cation of the hypotheses.
Furthermore there is knowledge of only some few Anisoptera and also only in an area between 8° and 14° southern latitude.For rare species (as for example almost all gomphids) also, it will not be possible in future to collect the required data because of the necessary high personal fi nancial expenditure and because of the extensive, mostly inaccessible area of approx.200,000 kms ² in the western Peruvian Andes.Th erefore, suitable investigations should concentrate in future also on a few groups and species but of redundance without losing sight of the requirement to compile an inventory.Furthermore it is nessecary to collect the data on the ecology of each species, since this is incompletely, known.
Whether and how climate changes in the western Peruvian Andes have an eff ect on the Zygoptera fauna, cannot at present be answered, because there are too few data and many species have not been even described up to now.Similar eff ects as described above can be expect.For example this may be shown by Protallagma titicacae and Ischnura capreola.Within the last few years, both species could be collected in nature regions from where there were no previous records.Furthermore, for both species adaptations to higher respectively lower temperatures are noticeable.To what extent drift through altered wind directions and wind forces are responsible (possibly these do infl uence also the dispersal behaviour of diff erent Anisoptera), remains unanswered.
Changes in the UV radiation regime as a possible factor eff ecting the dispersal and diversity of dragonfl ies (and certainly also of other groups) should be taken into account in future and examined more closely.
During the last two study campaigns in August / September 2006 an Frebruary / March 2007 extremely high UV levels (UV-A and UV-B) could be measured primarily in the Puna at levels of more than 4,000 m.
So the level was 15 UVI at the lagoon Ututo (4,464 m), 12 UVI at the lagoon Querococha (3,980 m) and 14 UVI at the lagoon Pumapashimin (4,242 m).Th ese are extremely high values5 which are accompanied by low atmospheric humidity (day averages 36% resp.37% resp.20%).Conspicuous was that no dragonfl y larvae could be found in any of these lagoons, while only several adult Rhionaeschna peralta fl ew in the shadows of the Polylepis stands (rosacean trees) near the lagoon Querococha.Especially UV-B radiation (wavelength 280-320 nm) can also lead to considerable damage in insects -the smaller the wavelength the higher the energy and therefore the possibility of damage.Th us the genetic hereditary material is most sensitive to radiation of the above wavelengths (Tevini 1996).Studies on crustaceans which live just below the wa-ter surface showed heavy damages and high rates of mortality at low-frequency UV-B radiation levels (Siebeck 1996).
Furthermore it cannot be excluded, that the imagines, which have a visual valence in the UV band, are impaired by changes of refl ectivity.Th us may occur possible changes for the dragonfl ies originally visible species-specifi c colourings (Corbet in 1999) which could lead perhaps to irritations in the search for females and thus prevents matings.It is also conceivable that the colour of water surfaces and / or egglaying habitats (e.g.plants and plant groups) are diff erently perceived6.
Beside the inclusion of changed climate factors (wind and cloud cover included), it should be particularly important in future to examine the infl uence of UV radiation changes at high mountain levels.
Th e changed conditions considerably infl uence biotope and habitat structures and possibly also have an eff ect on the occurrence and the ecology of dragonfl ies in the western Peruvian Andes.Of the species discussed here, it can be suspected that the aeshnids react, with some few exceptions, predominantly to habitat losses, but probably are able to adapt fl ight times and reproduction times to the changed conditions.Th e libellulids seem to be much more fl exible and appear to be able to conquer and colonize new areas and nature regions much more quickly, as well as to adapt fl ight times and reproductive times.

Figure 1 .
Figure 1.Schematized profi les of the westerns Andes with its natural areas and vertical climate gradients

Figure 2 .
Figure 2. Sequence of El-Niño-and La Niña episodes in 1890-2000; temperature deviations (°C) are shown (acc.to Caviedes 2001) Hanson et al. (2006) the measured temperatures have increased in the annual global mean temperature since 1970 to 2005 by 0.6° C (fi g. 3).Th e increase in the global annual means of measured temperatures since 1985 till 2005 amounts altogether 0.5°C.Th is corresponds in the fi ve-year means since 1990: 1990-1995 = 0,1°C, 1995-2000 = 0,2°C and 2000-2005 = 0,3°C.Th e increase of the measured temperatures from 1985 to 2005 amounted to a maximum of 0.8°C in the western Andes (Puna) and a minimum of 0.2°C (Chala).With respect to all altitudes of the Peruvian western Andes the temperature increase since 1985 probably lies between 0.5°C and 0.8°C (no meteorological data are available from Yunga, Suni and Jalca).An increase twice as high as the global mean.Th e last fi ve-years showed in 2000-2005 a smaller increase in the Puna and a slight decline in the Chala (fi g. 4).

Figure 3 .
Figure 3. Global mean surface temperature anomaly (°C).Global annual surface temperature relative to 1951-1980 mean based on surface air measurements at meteorological station and ship and satellite measurements for sea surface temperature (from Hanson et al. 2006).

Figure 7 .
Figure 7. Average January atmospheric humidity in Puna (4,000 m), Quechua (3,050 m) and Chala (60 m) over a period of the past eleven years.Curves: gliding average for two periods; spotted line: occurrence of El Niño (source: SENAMHI 2006)

Figure 8 .
Figure 8. Distribution map of the Central American and Northwest South American regions of Orthemis ferruginea, O. discolor and the Antillean form.Arrows show the dispersion directions, question marks stand for open dispersal questions, the dots show the two places with records of O. ferruginea in the Peruvian Yunga.

Table 1 .
Relation of natural areas in the westerns Andes to altitudinal zones

Table 2 .
Relation of natural areas in the westerns Andes to altitudinal zones, precipitation and temperatures