Bendix, J.* and Bendix, A**.
* Institute of Geography, University of Bonn,
Meckenheimer Allee 166, D-53115 Bonn-Germany
Telf: (++49-228) 73-2093, Fax: 73-7506,
e-mail: bendix@beauty.giub.uni-bonn.de
** Academy of Science and Literature,
Mainz, Dep. of Geoecology at University of Bonn,
Meckenheimer Allee 166, D-53115 Bonn-Germany,
e-mail: astrid@beauty.giub.uni-bonn.de
In the scope of the project ENPEX (El Niño Precipitation Experiment), the spatial distribution and temporal development of severe precipitation in Ecuador and northern Peru during the 1991/93 El Niño event have been investigated mainly based on satellite data (METEOSAT-3). The distribution of heavy rains is determined by the Convective-Stratiform-Technique (CST) which has been adjusted to METEOSAT-3 geometry as well as the KED interpolation method. The validation of the adjusted CST scheme reveals an encouraging accuracy also for extreme rainfall amounts up to 300 mm 3 days-1. The extraction of cloud motion winds (CMW) by means of cross correlation technique which is applied to sequences of METEOSAT imagery provides additional information on circulation patterns which are related to severe precipitation. The analysis have shown that the 1991/92 event is stronger than the 1972/73 but significantly weaker than the 1982/83 event. Severe precipitation occurs mainly in the coastal plain of Ecuador and north Peru up to the 1000 m contour line. Deep convection is originated by the land-sea-breeze phenomenon on most of the investigated days. In this case, heavy precipitation is locally confined and shows a clear diurnal cycle. Precipitation starts over land in the evening and is shifted to the coastal waters during night. A nocturnal center of heavy precipitation is the Gulf of Guayaquil due to the shape of the coastline which favours convergence and the frequently increased SSTís. in this area. Severe precipitation without any diurnal cycle which is characterised by great spatial extension occurs during main El Niño (Match-April) due to an extended labilisation of the lower troposphere. Deep convection is often organised in mesoscale convective complexes (MCC) which are related to extended areas of SSTís >28°C. During this situations, a strong meridional stream flow (Headly circulation) could be observed.
The spatial distribution and temporal formation of heavy precipitation in Ecuador and northern Peru during El Niño events is not finally known until today (Bendix 1994). Available studies which are mainly based on rain gauge data do not provide an exact overview over the patchy structure of the local rainfall distribution. Yet less known are the local circulation patterns which cause heavy precipitation within the mentioned area. Whereas some authors attribute anomalous El Niño precipitation in the arid environments of south Ecuador and north Peru to an extreme southward shift of the ITCZ due to increased SSTís (e.g. Miller and Laurs 1975), other investigations point out the importance of local circulation patterns (e.g. an extended land-sea-breeze phenomenon) for rainfall formation (e.g. Schütte 1968, Horel and Cornejo-Garrido 1986). The aim of the present investigation is to examine the interrelations of spatial rainfall distribution, circulation patterns and positive SST anomalies in Ecuador by means of a synergy of remote sensing techniques, ground based observations as well as spatial interpolation techniques and numerical modelling for the 1991/93 El Niño event.
Figure 1 gives an overview over the data sources and techniques which are used for the present investigation.
The spatial distribution of precipitation is estimated applying an adapted version of the Convective-Stratiform-Technique (hereafter CST) of Adler and Negri (1988) to Meteosat-3 data. Additional information concerning rainfall distribution is obtained by the application of a special interpolation technique (KED Kriging, Deutsch and Journel 1992) to rain gauge observations.
To derive the circulation patterns during single events of heavy precipitation, wind observations and radiosonde data are combined with cloud-motion winds (CMW) which are extracted from sequences of Meteosat-3 IR imagery by means of cross-correlation technique as described by Schmetz et al. (1993). This cloud tracking method provides the wind field of three atmospheric layers (850-700, 500-400, 200-100 hPa). Weekly data of multichannel sea surface temperatures (MCSST) derived from NOAA-AVHRR data which are available from the JPL-PO.ODAC archive (McClain et al. 1985) complete the investigations.
The CST rain retrieval technique was originally developed for the calculation of convective rainfall in Florida by means of GOES-6 IR-radiometer. The divergent viewing geometry of Meteosat-3 IR which observed Ecuador during the 1991/93 El Niño as well as the dissimilar climate of Ecuador require some adjustments to the original CST parameterisation (for further details of the CST procedure and the adjustment scheme refer to Bendix, J. 1997 or Bendix, J. and Bendix, A. 1996):
Validation of the adjusted CST scheme has been performed by comparing rain gauge data of 50 meteorological stations to rainfall calculated from 101 Meteosat images (Omega h interval) for the corresponding pixel locations (Figure 2, right). The rain retrieval by means of the adjusted CST reveals a good accuracy (r2 = 0.87, SD = 20.9 mm). It should be noted that also great amounts of rainfall up to 300 mm can be calculated by the adjusted scheme. However, a slight underestimation of heavy rainfall > 200 mm by the CST has been found. The outlier Na represents the station Naranjal (Ecuador) which has been proved as unreliable during several rainfall events by means of plausibility tests based on surrounding rain gauge stations.
A well known method for the spatial interpolation of precipitation maps from rain gauge data is the Kriging technique. However, the common ordinary Kriging technique often provides only unsatisfactory results because rainfall extension in the Tropics depends on various factors as e.g. the altitude. To account for topographical effects, the Kriging method with External Drift (KED, Deutsch and Journel 1992) has been chosen for the current study. The KED technique comprises three overall steps:
Table 1 shows the result of sensitivity test for different Kriging settings using the complete data set of 84 rain gauge stations for the 1982/83 El Niño event. Precipitation maps have been interpolated using KED with only 71 stations. Subsequently, the results for 13 independent stations have been compared to the real observations to determine accuaracy of the technique as well as the sensitivity in relation to single settings.
|
Kriging settings: Variogram adjustment Anisotropy external drift |
on on on |
on off on |
on on off |
on off off |
|
r2 |
0.94 |
0.92 |
0.88 |
0.84 |
|
r |
0.97 |
0.96 |
0.94 |
0.91 |
The results reveal clearly a major importance of the drift variable altitude with which the correlation between observed and interpolated precipitation can be significantly increased. The importance of external drift additionally increases if only marginal stations are considered for sensitivity test. The test results an improvement of correlation from r2=0.74 to r2=0.94 for 7 marginal of the 13 independent stations.
Wind field maps at several atmospheric layers could be derived for every day with heavy precipitation by applying the cross-correlation technique after Schmetz et al. (1993.
|
upper level |
middle level |
lower level |
|
|
wind speed [m sec-1] |
±2 |
±1 |
±0.5 |
|
wind direction [°] |
±15 |
±16 |
±23 |
The CMWís have been calculated for the low (850-700 hPa), middle (700-300 hPa) and upper atmospheric level (<300 hPa). For details of this technique refer to Schmetz et al. (1993) and Bendix and Bendix (1996). Information concerning the accuracy of this technique within the study area are presented in table 2.
The interpolated maps of precipitation as well as the corresponding rainfall anomalies for the 1991/92 and the 1992/93 event are presented in figure 3. The greatest rainfall amounts have been observed in the coastal plains of north and central Ecuador but a secondary maximum is found in the arid region of south Ecuador close to the 1000 m contour line. Obviously, precipitation during the 1991/92 event extends more to the south especially in the coastal area of northern Peru (Sechura desert). On the other hand, the inter-Andean basin as well as the Amazon region of Ecuador had been significantly dryer during the 1991/92 event as in 1992/93, especially in the northern part of Ecuador.
The rainfall anomalies also points out to a more intense event of 1992. Up to 1000% more precipitation could be observed in the arid coastal parts of north Peru especially between Tumbes (940%) and Talara (944%). On the other hand, only up to 300% surplus of rainfall could be measured during 1993 close to the 1000 m contour line at the border between Ecuador and Peru. Generally, the 1992 event was weaker than the 1982/83 Niño (e.g. Talara = 8071%) but stronger than 1972/73 (Bendix 1994). The 1993 continuation only is of local importance.
A more detailed overview over the spatial extension of heavy precipitation also including the marine environment provides the CST map which was calculated for the 45 days (i.e. 1087 Meteosat-3 images) with the heaviest rainfall during 1992 (figure 4). The CST map generally affirms the precipitation maximum as seen in the interpolated map. Greatest amounts are also observed in the coastal plains up to the 1000 m contour line but the calculated spatial rainfall distribution is more patchy. It can be noted that local rainfall maximum are often related to the weak elevations of the coastal cordillera (<500 m asl) and hence, forced convection seems to be involved in rainfall formation. A remarkable local maximum can be observed over the Gulf of Guayaquil where convection is expected to be intensified due to more frequently increased SSTís >28°C.
Further interesting features are generally weaker precipitation over the adjacent Pacific as far as the Galapagos Islands as well as the dry area at the eastern slope of the Ecuadorian Andes which is also present during a normal rainy season (Bendix and Lauer 1992).
Some theories which accounts for the formation of severe precipitation in Ecuador and north Peru have been derived from preceding EN events. :
The most of the 45 days with severe precipitation are characterised by a weak sea-breeze phenomenon which mainly causes local precipitation. Land and sea breeze in the arid coastal plains of south Ecuador and north Peru are only developed during increased SST during Niño situations. In normal years, the grate thermal contrast between cold up-welling and heated desert causes a divergent coats parallel stream flow (thermo-tidal wind, see Horel and Cornejo-Garrido 1986.) The typical course of deep convection is presented in figure 5.
The sea-breeze is developed at 10 LT (Manta) and initiates convection within the coastal plain until 19:00 LT especially in northern and central Ecuador. Cloud free is the area in the vicinity of the Gulf of Guayaquil because the shape of the coastline favours a divergent stream flow. Later at night, the areas of deep convection are shifted off the coastline due to the well developed land-breeze which can be observed e.g. at the station Manta (6:00 LT). Especially convection over the Gulf of Guayaquil is again influenced by the shape of the coastline which favours a convergent land breeze. The CST map shows that the areas of precipitation >80 mm 12 hours-1 are locally confined to just a few places with its maximum over the Gulf of Guayaquil.
A typical weather situation for the central El Niño period during March-April is presented in figure 6.
Centers of heavy precipitation are the eastern Amazon, the Pacific area north and south of the Galapagos Islands as well as the coastal plains of Ecuador and north Peru. The severest rainfall >200 mm 12 hours-1 occurs within the Sechura desert close to Piura as well as on the peninsula Sta. Elena and the adjacent coastal waters. Deep convection is organised in mesoscale convective complexes (MCC) which are related to an extended labilisation of the lower troposphere over areas of SSTís 28°C mainly due to the suppression of coastal up-welling off the coast of Ecuador and Peru. The MCCís show no well-defined diurnal cycle as it is observed for precipitation formation due to the land-sea-breeze phenomenon.
The circulation patterns in the upper level are characterised by easterly winds along the equator but a meridional wind direction (from north) over the precipitation field south of Galapagos. This meridional stream flow replaces the trades of the south Pacific anticyclon which are normally well developed in this area. It represents a weak Walker in relation to an intensification of the Headly circulation due to a great meridional contrast in SST which is also typical for the main El Niño phase (Kousky and Ropelewski 1989). Convection in this area is sometimes originated from convective cloud lines which indicate an extended labilisation and are only observed during El Niño situations (Goldberg et al. 1987).
The present investigation has shown that satellite data are an appropriate tool to examine the extension, formation and dynamics of severe precipitation in Ecuador and Peru during El Niño situations. The maximum of precipitation could be found within the coastal plains and the adjacent ocean as well as over the Gulf of Guayaquil with the greatest anomalies in the normally arid regions of Ecuador (peninsula Sta. Elena) and north Peru (Sechura desert). The formation of severe precipitation is either local due to the land-sea-breeze phenomenon or more extended due to deep convection which is frequently organised in MCCís. However, the overall spatial structure of rainfall reveals significant analogies to a normal rainy season even if its southward extension is much greater. On the other hand, satellite-based case studies of the 1982/83 Super Niño have shown some deviations to the situation of a so called Normal Niño. Therefore, the expected strong event of 1998 provides a good opportunity to compare a normal (1991/92) and an extreme Niño situation.
The project ENPEX ("El Niño Precipitation Experiment") is funded by the German Research Council (DFG; Grant-No Be 1780/1-1 & 1-2). The authors thank all the people at INAMHI, DAC (Ecuador) and CORPAC, SENAMHI (Peru) for providing the meteorological data.
Adler, R. F. and Negri, A. J., (1988). A satellite infrared technique to estimate tropical convective and stratiform rainfall, J. Appl. Meteorol. Vol. 27, 30-51.
Bendix, J., (1994). Zur Klimatologie des El Niño-Phänomens an der Westküste des tropischen Südamerika, Abh. d. Akad. d. Wiss. u. d. Lit. Vol. 2-94, 51-64.
Bendix, J., (1997). Adjustment of the Convective-Stratiform Technique to estimate 1991/93 El Niño rainfall distribution in Ecuador and Peru by means of Meteosat-3 IR data, Int. J. Rem. Sens. Vol. 18, 1387-1394, 2707.
Bendix, J. and Lauer, W. (1992). Die Niederschlagsjahreszeiten in Ecuador und ihre klimadynamische Interpretation, Erdkunde Vol. 46, 118-134.
Bendix, J. and Bendix, A., (1996) Investigation of precipitation dynamics during El Niño in Ecuador and Peru using Meteosat-3 and NOAA-AVHRR data, Proc. of the 1996 Meteorological Satellite Data Users' Conference, Wien 16.-20.9.1996, 453-458.
Deutsch, C. V. and Journel, A. G. (1992). GSLIB - Geostatistical Software Library and Userís Guide, New York, Oxford. (Oxford University Press).
Goldberg, R. A., Tisnado M. G. and Scofield, R. A. (1987). Characteristics of extreme rainfall events in northwestern Peru during the 1982-1983 El Niño period, Journal of Geophysical Research Vol. 92, 14225-14241.
Horel, J. D. and Cornejo-Garrido, A. G., (1986). Convection along the coast of Northern Peru during 1983: Spatial and temporal variation of clouds and rainfall, Month. Weath. Rev. Vol. 114, 2091-2105.
Kousky, V. E. and Ropelewski, C. F. (1989). Extremes in the Southern Oscillation and their relationsship to precipitation anomalies with emphasis on the South American region. Revista Brasileira de Meteorologia Vol. 4, 351-363.
McClain, E. P., Pichel, W. G., and Waltion, C. C., (1985). Comparative performance of AVHRR-based multichannel sea surface temperatures, J. Geophys. Res. Vol. 35, 11587-11601.
Miller, F. and Laurs, R. M., (1975). The El Niño of 1972/73 in the eastern tropical Pacific Ocean, Inter Am. Trop. Tuna Comm. Bull Vol. 16, 403-448.
Schmetz, J., Holmlund, K., Hoffmann, J., Strauss, B., Mason, B., Gaertner, V., Koch, A. and van de Berg, L., (1993). Operational cloud-motion winds from Meteosat infrared images. J. Appl. Meteorol. Vol. 32, 1206-1225.
Schütte, K., (1968). Untersuchungen zur Meteorologie und Klimatologie des El Niño Phänomens in Ecuador und Peru, Bonner Met. Abh. Vol. 9.