Project Topics
Get Complete Project Material File(s) Now! »
Mean climatic conditions in the South East Pacific
On the one hand, under normal conditions, the Peruvian Pacific drainage region (hereafter Pd) is influenced by the Southern Pacific Anticyclone (hereafter SPA) in combination with the Humboldt current characterized by cold Sea Surface Temperatures (hereafter SST) which produces dry and stable conditions to the western central Andes, with moist air trapped below the inversion zone at about 900 hPa ~ 1000 m.asl (Vuille et al., 2000; Garreaud et al., 2002), conditions that produce extreme aridity until about that altitude (Lavado et al., 2012). Over this altitudinal limit, it is known that there is an influence of the southward displacement of the Inter Tropical Convergence Zone (hereafter ITCZ) and is supposed that other mechanisms influencing over the Peruvian Andes, also influence over the Peruvian Pacific slope (i.e. humidity transport from the Amazon, Bolivian High, etc) (Nickl, 2007; Lagos et al., 2008), nevertheless this has not be studied to date. On the other hand, this region exhibits greater seasonal and interannual precipitation variability than the two main others hydrological regions of Peru: the Amazon and the endorheic Titicaca drainage areas (Lavado et al., 2012), mainly caused by the El Niño Southern Oscillation (hereafter ENSO) influence in the northern areas during the rainy season, with no clear evidence of the ENSO influence for central and southern areas (Lagos et al., 2008; Lavado et al., 2012; Lavado and Espinoza, 2014).
Precipitation along the Pacific drainage of South America is characterized by a complex pattern of spatial and seasonal variability related to its meridional extension and the prominent topography of the Andes Cordillera (Waylen and Poveda, 2002; Garreaud et al., 2009). The Pd region is located at tropical latitudes and precipitation is mainly influenced by orographic conditions, ocean and atmosphere. The region is characterized by a steep topography that inhibits cross-shore atmospheric flow and disrupt a geotropically balanced zonal wind, inducing a northward sea level pressure gradient along the coast that accelerate the wind northward (Muñoz and Garreaud, 2005). Such a low-level northward mean circulation is associated with cool SST through inducing upwelling and evaporation, which makes this region persistently free of convective precipitation year-around (Takahashi and Battisti, 2007). The Pacific coast of Peru is thus mostly a “dry zone” that only episodically experiences precipitation events. At interannual timescales, those precipitation events are associated with ENSO phenomenon that is the main climatic influence over precipitation over the Peruvian Pacific coast (Lagos et al., 2008). A rainy season can also be developed owed to a slight weakening of the southeast Pacific anticyclone and the southward displacement of the Pacific ITCZ (Lavado et al., 2012).
Modes of variability
The ocean-atmosphere interactions in the Pacific Ocean produce a large-scale phenomenon known as the El Niño Southern Oscillation (ENSO). It is attributed to irregular variations in atmospheric pressure and sea surface temperature of the tropical Pacific Ocean (see Figure 2.1). This phenomenon is decomposed into two phases: El Niño phase is a hot event (positive sea surface temperature anomalies of the ocean) while La Niña phase is a cold event (negative sea surface temperature anomalies of the ocean). These climatic changes can severely disrupt the climate of the tropical regions on both sides of the Pacific Ocean (see Figure 2.1) and in particular the Pacific coast of South America (Aceituno, 1993).
El Niño in Peru is often associated with heavy rain along the coastal fringe of the country due to the intrusion of warm tropical waters along the coast that allows deep convection in a region where cold upwelled waters and semi-arid to arid climate usually prevails (Tapley, 1990; Lagos and Buizer, 1992). Over the last decades, two extreme El Niño events (1982/1983 and 1997/1998) took place and led to large socio-economical consequences because of the disasters caused by the floods and droughts. Considering the societal concern, recent studies have thus been devoted to inferring precipitation in Peru based on seasonal forecast products from climate models (Lagos et al., 2008). The approach consists of building a statistical model between local precipitation as inferred from observations and climate indices as derived from SST of the tropical Pacific predicted by the seasonal forecast systems (ex. NCEP). The skill of the forecast system is also highly dependent on the selected predictors and statistical method. For instance, the linear assumption, often used for the statistical approach, is certainly not the most appropriate for predicting precipitation in Peru, because ENSO has a strong positive asymmetry (evidenced by the extreme warm events) reflecting the non-linearity in the system (An and Jin, 2004). Recent studies have revealed that there is also different types of El Niño events that have distinct characteristics in terms of atmospheric teleconnections (Yeh et al., 2009), frequency (Kim and Yu, 2012), and oceanographic manifestations off Peru (Dewitte et al., 2012). Two types of El Niño events have been documented so far in the literature (Ashock et al., 2007; Kug et al., 2009): the so-called Cold Tongue El Niño or Eastern Pacific El Niño (hereafter EP El Niño) that corresponds to extreme warm events developing strong SST anomalies in the eastern equatorial Pacific, and the Warm Pool El Niño or Central Pacific El Niño (hereafter CP El Niño) that corresponds to standing warm SST anomaly development in the central equatorial Pacific, within the so-called Warm Pool region (see Figure 2.2). In a recent study, Takahashi et al. (2011) suggest that the dominant mode of variability in the equatorial Pacific is in fact associated with a regime that encompasses the CP El Niño and that the EP El Niño events are extreme events, which are by definition much rarer than CP El Niño events. This view has of course implications for the study of the ENSO teleconnection pattern over Peru for precipitation because CP El Niño are characterized by weak anomalous SST conditions off Peru compared to EP El Niño (Dewitte et al., 2012).
Temperature and evapotranspiration
Temperature series were obtained from 59 meteorological stations (see Figure 2.6) managed by the SENAMHI for the 1970‒2008 available period. A careful quality check of these data was performed. Mean monthly temperature data were homogenized and validated following Lavado et al. (2013). Missing values were filled by monthly average and by a multiple correlation method based on nearby geographical stations data.
Also lumped temperature time series were computed at basin-scale for water and energy balance analysis over the Pd catchments. Temperature data were interpolated to a 5 x 5 km grid using the inverse distance weighting technique. Orographic effects on temperature are notorious and they were accounted for using the SRTM digital elevation model in a similar way as described in Ruelland et al. (2014). These effects on precipitation were considered using the approach proposed by Valéry et al. (2010) with a constant lapse rate of -6.5°C/km (estimated from the observed data). Finally, since the only data available for calculating potential evapotranspiration (PET) were temperature data, a formula relying on clear monthly sky solar radiation and mean monthly air temperature was selected (Oudin et al. 2005): = R T+K if T + K > 0 λρ K = 0, (2.1)
where PET is the rate of potential evapotranspiration (mm/d), Re is the extraterrestrial radiation (MJ/m2/d), λ is the latent heat flux (2.45 MJ/kg), ρ is the density of water (kg/m3), T is the mean daily air temperature (°C) and K1 and K2 are fitted parameters (for a general case: K1~100 and K2~5). Equation 2.1 was applied at monthly time step. The Oudin formula is a temperature-based evapotranspiration model, which is adapted to arid and semi-arid regions limited by scarcity of in-situ climate data (see e.g. Hublart et al. 2015; 2016). Indeed, Oudin et al. (2005) showed that from an operational point of view, this model is as efficient as more complex models such as the Penman model (Penman, 1948) and its variants.
K-means clustering technique
K-means clustering is a statistical technique designed to assign objects to a fixed number of groups (clusters) based on a set of specified variables. One of the principal advantages of k-means technique consists in its cluster’s identifying performance which allows ranking the obtained clusters as a function of their representativeness. The process involves a partitioning schema into k different clusters previously defined. Objects that are within those k clusters must be as similar as possible to those that belongs to its own group and completely dissimilar to the objects that are in the other clusters. Similarity depends on correlation, average difference or another type of metrics. By definition each cluster is characterized by its own centroid with the cluster members located all around it. The algorithm used at annual precipitation timescale was the Hartigan-Wong which adopts the squared Euclidean distance as a dissimilarity measurement. See more details of this method in Hartigan and Wong (1979).
A key part of the k-means application is to define an optimum number of clusters. In order to succeed in the definition of partitioning groups, an estimation of the silhouette number must be performed for each desired number of groups. The silhouette width is used to evaluate the statistical significance of each identified cluster (Rousseeuw, 1987). The silhouette value is obtained following Rousseeuw (1987) as: () = (, ) () (3.2) ( ), (, ).
Where: a(i) corresponds to the average similarity between the ith object and the other objects of the same group and b(i,k) is the average similarity between the ith object and the members of the kth clusters. The range of variation for this silhouette index is between -1 and +1, when the silhouette value is close to +1 means that there is a better member correspondence to its own cluster, while a negative value represents the object this is not well located in the appropriate cluster. Meanwhile the value of 0 means that objects could belong to any k cluster. We also computed an average silhouette width for the whole k clusters which represents the mean of S(i), and it can be used to choose the best number of clusters, by taking the value of k for which S(i) is maximal.
Regionalization Analysis
There are classical ways to predefine regions; it can be based on stations proximity and homogeneity, physiographic patterns or topographical constraints related to isohyets (Espinoza et al., 2009; Bourrel et al., 2015). Here, precipitation stations grouped by k-means clustering are set up as predefined regions. The criteria for using k-means clustering as first step of regionalization is based on the advantages in time solving and the preset number of groups at the beginning of the process whilst RVM requires defining the stations grouped into a predefined region, being a long and exhaustive methodology if it is not provided an accurate number of groups.
Regionalization was performed using the RVM, which is generally oriented to: a) assess precipitation data quality based on the homogeneity within a predetermined region (Espinoza et al., 2009) and b) achieve precipitation regionalization processes (establishment of representative vectors of homogeneous precipitation zones) to gather the stations exhibiting the same interannual variability. The process for regionalization is similar to the process explained in section 3.2.1 (item 2). It depends on the computation of a “mean station” or “vector” from all data involved in the study area that will be compared with each pluviometric station (Brunet-Moret, 1979). Prior to the use of the RVM, it is necessary to group stations into predefined regions.
Once calculated, the RV is compared iteratively with data station for discarding those stations whose data are not consistent with the RV and reprise the process. The rejection of a given station could mean that this station belongs to a neighbouring region that could present greater consistency. Therefore in many cases, stations or areas are re-grouped or divided in order to obtain regions that show homogeneous features. The main statistical criteria for regrouping stations into homogeneous regions are based on thresholds applied to the standard deviation of the differences between annual pluviometric indices of stations and the RV indices; and to the correlation coefficient between RV and annual pluviometric values of stations. These thresholds are fixed to the standard deviation lower than 0.4 and correlation coefficient greater than 0.7. Precipitation database management and RVM were carried out using the HYDRACCESS software (Vauchel, 2005).
Precipitation data interpolation
After regionalization based on punctual information (i.e. precipitation stations), we did a precipitation spatialization by isohyets allowing to delimit polygonal regions. Annual precipitation was interpolated incorporating elevation data using the co-kriging classical geostatistical approach, which is widely used in the hydrometeorological field (Goovaerts, 2000; Diodato, 2005; Buytaert et al., 2006). Co-kriging, which is a multivariate version of kriging technique, took into account the digital elevation model (DEM) SRTM – 90 m data as correlated secondary information based on a spherical variogram (Goovaerts, 2000; Mair et al., 2011). This precipitation interpolation map was used as a background raster guide for delineating polygonal regions involving the station points grouped with regionalization analysis. These polygons follow the isohyets shape with geometrical approach (perpendicular and bisector criteria of boundaries of regions traversing isohyets and stations) and a statistical approach (revalidation of new defined areas with the RVM with proper fit of stations inside each region).
Finally, representative monthly precipitation time series of each region were obtained with the co-kriging methodology because of better performance than other techniques (e.g. Thiessen Polygons, Inverse Distance Weighted and Kriging) over mountains areas (Hevesi et al., 1992a, 1992b; Goovaerts, 2000; Diodato et al., 2005). Time series were assigned to centroids as representative points for obtain mean latitude, longitude and altitude of each region.
Precipitation Classification
A cluster analysis of the annual precipitation data was performed by applying k-means technique on the 124 precipitation stations previously selected. The optimal value for the cluster numbers was determined by an average silhouette value and a negative silhouette number for a number of clusters varying from 3 to 10 (see Table 3.1).
Maximum silhouette values are obtained for cluster-three (0.64), cluster-four (0.60) and cluster-six (0.55), considering as a reasonable structure a cluster having a silhouette value greater than 0.50 and as a weak structure a silhouette value less than 0.50 following Kononenko and Kukar (2007). The number of negative silhouette values is minimal for cluster-three (6), cluster-four (4) and cluster-six (6). After plotting the cluster groups into a map showing their spatial distribution, we selected the cluster-three and cluster-six among them, as these two clusters show certain arrangement of precipitation stations according to topographical and latitudinal variation (see Figures 3.2a and 3.2b). Cluster-four was an intermediate group that corresponds to one sub-region in the north.
Table of contents :
Chapter 1 Introduction
1.1. General context
1.2. Motivation
1.3. Main and specific objectives
1.4. Organization
Introduction (version française)
Chapter 2 Study area and data
2.1. Hydroclimatic context of the Peruvian Pacific drainage
2.1.1. Climate variability
a. Mean climatic conditions in the South East Pacific
b. Modes of variability
El Niño phenomenon (ENSO) and its diversity
Decadal variability
Intraseasonal variability (MJO and oceanic Kelvin waves)
2.1.2. Physical landscape
a. Topography
b. Geology
c. Vegetation
2.1.3. Hydrological context
a. Arid and semi-arid conditions
b. Anthropogenization
2.2. Data
2.2.1. Precipitation
2.2.2. Temperature and evapotranspiration
2.2.3. Streamflow
2.2.4. Climatic indices
a. ENSO indices
b. MJO index and Kelvin waves
Chapter 3 Precipitation regime
3.1. Theoretical background
3.2. Methods
3.2.1. Data homogenization and validation
3.2.2. Classification and Regionalization Process
a. K-means clustering technique
b. Regionalization Analysis
c. Precipitation data interpolation
3.3. Results and discussion
3.3.1. Precipitation Classification
3.3.2. Regionalization
3.3.3. Regions Characterization
3.4. Conclusions
Chapter 4 Hydroclimatic balance
4.1. Theoretical background
4.2. Methods
4.2.1. Catchment water balance disparity
4.3. Results and discussion
4.3.1. Hydroclimatic time series
4.3.2. Catchment water balance disparity
4.4. Conclusions
Chapter 5 Runoff regime
5.1. Theoretical background
5.1.1. Hydrological lumped conceptual modelling
5.1.2. Regional runoff .
5.2. Methods
5.2.1. Runoff simulation based on conceptual lumped models
5.2.2. Performance and efficiency of conceptual lumped models
5.2.3. Regional runoff model (RRM) and freshwater estimates
5.3. Results and discussion
5.3.1. Efficiency of the GR1A and GR2M models
5.3.2. Regional runoff model evaluation
5.3.3. Freshwater runoff estimation
5.4. Conclusions
Chapter 6 Impacts of climate variability and hydroclimatic change on precipitation and runoff
6.1. Precipitation and runoff variability associated with ENSO
6.1.1. Theoretical background
6.1.2. Methods
a. Principal Component Analysis (PCA)
b. The wavelets and coherence analysis
c. Correlation analysis
d. Covariance analysis
6.1.3. Results and discussion
a. PCA analysis of ENSO indices
b. Coherence between ENSO indices and precipitation series
c. Low frequency modulation of ENSO and precipitation regime
d. Precipitation variability and sea surface temperature anomalies
e. Low frequency modulation of ENSO and runoff regime
6.1.4. Conclusions
6.2. Trends and hydroclimatic change disparity over catchments
6.2.1. Theoretical background
6.2.2. Methods
a. Characterization of hydroclimatic time series
b. Hydroclimatic change disparity
6.2.3. Results and discussion
a. Characterization of hydroclimatic time series
b. Hydroclimatic change disparity
6.2.4. Conclusions
Chapter 7 General conclusions and perspectives
7.1. Conclusions
7.2. Perspectives
7.2.1. Impact of climate variability over seasonal hydrological regime as a forecasting tool
7.2.2. Impact of climate and catchment change over the hydrological regimes
Conclusions générales et perspectives (version française)
References
