Télécharger le fichier original (Mémoire de fin d’études)

SPATIAL AND SEASONAL VARIABILITY OF RUNOFF AND RAINFALL

Mean annual rainfall values (1970 -2005) in 167 stations, interpolated using the Kriging method, are displayed in Figure 2. A strong opposition appears between the rainy eastern lowlands in the Amazon basin (more than 1500 mm/year – see also Table 2) and drier conditions in the Andes and in the basins of the Pacific coast (less than 1000 mm/year).
Weak rainfall along the Pacific coast is related to the subsidence in the South East Pacific subtropical anticyclone that prohibits rainfall. Though, in Pd, rainfall increases while latitude diminishes, from very low values in the southernmost basins (144 mm/year in PQ29, Table 2) to nearly 1000 mm/year in the North (Figure 3a). This can be related to the diminution of the strength of the southern Pacific anticyclone toward North and to the seasonal southward displacement of the Pacific ITCZ that reaches the northern coast of Peru. This South-North gradient is also observed in the specific runoff of the river (Figure 3b) that has been computed in order to avoid the problem of dependence between runoff and size of the basin. The specific runoff increases regularly toward North. More, along the Pacific coast rainfall increases with altitude with very low values near the coast (much lower than 100 mm/year in the South) and higher values in the higher parts of coastal basins, above the inversion layer that blocks the convection at about 500-800 meters (Dollfus, 1964 and Johnson, 1976).
On the eastern flank of the Andes, on the contrary, rainfall is abundant. Though, local maximum and minimum are noticeable on Figure 1. In particular, very high annual rainfall (above 6000 mm/year) is recorded in regions where the concave form of the Andes favours convergence (Borja region) or in valleys opened toward North and oriented toward the humid north-western wind (High center Ucayali valley) (Figueroa et Nobre, 1990; Laraque et al., 2007; Espinoza et al., 2009b). They contrast with less rainy regions as the high Huallaga valley that is sheltered by relief and less exposed to moist air advection. Rainfall tends to increase when altitude diminishes as cool temperatures in the Andes do not favour a high moist content of the air.
Figure 2. Mean annual rainfall values (mm) for the 1965-2007 period over Pacific (Pd), Amazonas (Ad) and Titicaca (Td) drainages.
Figure 3. Relationship between latitude and mean multiannual A) specific runoff in Pacific drainage, B) rainfall.
Labels and located of stations are described in Table 1.
Monthly rainfall and runoff values expressed as a percentage of the hydrological annual values are represented in Figure 4 for some selected gauging stations. There is a pronounced annual cycle along the Pacific coast and in the Altiplano, with a rainy season during austral summer and a dry season in winter. The rainy season on the coast corresponds to the southward displacement of the Pacific ITCZ. The rainfall seasonal coefficient of variation are often higher than 0.7 (Table 2) and increase from the northern to the southern Pacific coast, revealing a higher seasonal variability in the South. Moreover, many rivers in Pd, in the North as well as in the South, are non permanent and are characterized by null or near null low-level flow in austral winter; this feature, consistent with the strong seasonality of rainfall, may explain the very high seasonal discharge coefficient of variation in some basins (Table 2). On the Altiplano, the rainy season in austral summer is characterized by an intense convective activity combined with moisture advection from the Amazon basin (Montecinos et Aceituno, 2003; Garreaud, 1999). During austral winter, the eastern moist flux is replaced by westerly winds, which provide dry air related to the atmospheric stability over the Pacific Ocean. Consequently the seasonal coefficients of variation of rainfall and discharge are about 1. In contrast, rainfall is abundant all year round in the Amazon basin, with a maximum in summer during the South American Monsoon (Zhou et Lau, 1998) and the seasonal coefficients of variation are much lower than in Pd and Td, i.e. 0,2 in AQ1 and AQ2 (Table 2).
Figure 4. Mean monthly rainfall and runoff in percentage (ratio between monthly and mean annual values for runoff and ratio between monthly and total annual values for rainfall) in the selected basins in Peru.
In the small northern and center and probably southern Pd basins (e.g. Tumbes PQ-1, Santa PQ-10 and Cañete PQ-17) there is an immediate runoff response to rainfall with discharge peaks in March and low values in austral winter. Santa basin (PQ-10) shows singular characteristics (more runoff than rainfall in autumn and winter months, see Figure 3) due to glacier contribution (Pouyaud et al., 2005). In Southern Pd, there is one month lag between peaks in rainfall (January) and peaks in runoff (February) (e.g.). In greater basins there may be a month lag between the maximum rainfall and the high flow runoff (as in Majes PQ-24) In Td, there is one month lag between rainfall (January) and runoff (February) peaks in the large Ramis basin (TQ-1) . As the sizes of the Amazonas basins are huge, there are two months lag between peaks in rainfall (Mars) and runoff (May) as in Tabatinga gauging station (AQ-1) (Espinoza et al., 2006a).

INTERANNUAL VARIABILITY OF RAINFALL AND DISCHARGE

Annual hydrological runoff and rainfall during the 1970-2005 period are plotted in Figure 5 for the same stations than in Figure 4. On average, mean annual rainfall and runoff present similar variability and are significantly correlated (Table 2). In some coastal basins (PQ10, PQ14, PQ15, from PQ18 to PQ23, PQ27), the correlations are low. This is due to the presence of underground water that sustain low flow runoff, to the melting of glaciers in upstream valleys as in the case of the Santa River (PQ10), to barrages or water derivation (for irrigation or city water supply, as in the case of the Rimac basin (PQ15) that drives water towards the city of Lima. Moreover, Tumbes river (PQ-1) shows huge rainfall and runoff anomalies during the last two ENSO extreme events (1982-1983 and 1997-1998) (Figure 4). This results in high rainfall and runoff interannual coefficients of variation (Table 2). Though, Qmin interannual coefficients of variation that are generally higher than Qmax CVs, experience a large set of values, from 0,3 in some basins until 3 in PQ3 and PQ21. Correlation between rainfall and runoff are generally good in the Altiplano and in the Amazonas where the rivers are not equipped by man, except in AQ2. Interannual rainfall and runoff coefficients of variation (CVs) are very low in the Ad (see Tabatinga in Figure 5), indicating a low interannual variability, while they are a little higher in the Altiplano (see Ramis in Figure 5 and Table 2).
In summary, the Pd shows more rainfall and runoff variations than Td and Ad at seasonal and interannual time scale. This is particularly visible in Figures 6a and 6c where interannual coefficients of variation (MACV) of runoff and rainfall respectively are plotted against seasonal coefficients (SVC). Td and Ad stations are located near the origin of the graph while Pd basins are located higher, but occupying a wide range of values. Figures 6b and 6d show the ratio MACV / SVC in runoff and rainfall in Pd versus latitude. An opposition is evidenced between northern and southern stations, the northern Pd basins showing the highest interannual variation in rainfall and runoff, when compared to seasonal variation, probably in relation with the strong hydrological impacts of El Nino events in the northern coast. Another set of basins in the southernmost coastal region experiences a strong interannual runoff variability that is not observed in rainfall.
Figure 5. Annual series (September-August) for distinct basins over Peruvian drainages: the black lines are the mean annual runoff, the blue lines blue are the total annual rainfall, the black dotted lines are the maximum runoff and the black dashed lines are teh minimum runoff. Values are standardized and are corrected by coefficient in order to avoid confusion between the different lines. The coefficients are 4 for maximum runoff, 0 for mean runoff and -4 for minimum runoff. Figure 6. Relationship between multi-annual variation coefficients (MAVC) and the seasonal variation coefficients (SVC) in A) runoff and C) rainfall. Ratio between MAVC and SVC plotted in function of latitude in stations located in the Pacific drainage for B) runoff and D) rainfall. Circle black: stations located over Pacific drainage. Red plus sign: stations located over Amazonas drainage. Blue upward triangle: stations located over Titicaca drainage.

RUNOFF AND RAINFALL CHANGES AND TRENDS

The years of significant mean changes in runoff and rainfall in the Peruvian basins have been computed using Pettitt test. They are presented between parentheses in Table 3. Runoff and rainfall trend analysis have been computed using Bravais Pearson and Mann Kendal tests. For each basin, the best result between both methods is given in Table 3.
The main result is a change in all variables in the middle of the eighties in the two Amazonian basins, with rainfall and runoff decreases that have already been described by Espinoza et al. 2009. Itrend values, ratio between the slope of the linear trend and the mean value of the series, indicate that the stronger runoff decrease is observed in Tamshiyacu where there is no significant rainfall decrease, while in Tabatinga a moderate runoff decrease can be related with an important diminution of rainfall. In both stations the stronger runoff decrease is observed in minimum runoff.
In the Altiplano, on the contrary, there is an increase in minimum runoff. It is significant in the large Ramis basin (TQ-1), with a change in 1986, but it is not related to a change in annual rainfall. Is it associated with changes in winter rainfall that are not detected with our data and analysis.
Along the Pacific coast, the main changes and trends are observed in minimal runoff. The signals are different from a station to another and they are not related to change in annual rainfall. As there is no regional signal (the trends are as well positive or negative, the changes occur at very different dates, from the seventies until the late nineties), we can hypothesize that atmospheric circulation and rainfall do not originate these changes. On the contrary, they may be related to climatic change (glacier melting) or to constructions dedicated to sustain low flow when the trend is positive, and to the increase of water pumping installations for agriculture or city use, when the trend is negative. Positive trends are found in the Chillon (PQ-14), Rimac (PQ-15), Yauca (PQ-22) and Sama (PQ-28). The runoff of Chillon (PQ-14) and Rimac (PQ-15) rivers are mostly dedicated to urban use for the Lima city that according the last census 2007 from the National Institute of Statistic and Informatics from Peru, INEI (www.inei.gob.pe ), bind 30.8% of total Peruvian population. Thus, the minimum runoff increase is due to runoff’s control in the high rainy zones of these basins, which involve many reservoirs systems and include water transfer from the Amazon basin. Negative trends are observed in rivers Viru (PQ-9), Pativilca (PQ-11), Huaura ( PQ-12), Acari ( PQ-21), Ocona (PQ-23). In these basins agricultural activity may limit low flow discharge by surface and underground water pumping that has considerably increased since the 1990s, according to the Peruvian Agriculture Ministry (www.minag.gob.pe).
There is no trend in the big Santa River (PQ-10); on the one hand it benefits from the melting of the White Cordillera glacier but, on another hand, its water is intensely exploited for export agriculture production.

IMPACT OF OCEAN-ATMOSPHERIC INDICES ON PERUVIAN HYDROLOGY

In order to search explanation for the interannual and long term variability of runoff and rainfall, the hydrological series have been correlated to ocean-atmosphere indices as the Southern Oscillation Index (SOI) and Atlantic indexes constructed using Sea Surface Temperature (SST) in the southern and northern tropical Atlantic ocean. These indexes have been chosen because on the one hand coastal Peru is a key region for the El Nino – Southern Oscillation (ENSO) phenomena and because the tropical Atlantic SST is related to the strength of the trade winds and consequently to the water vapour advection from the Atlantic to the South America tropics (Marengo 1992, Marengo 2004). Their time series are presented in Figure 7a, b, c.
Correlations values using Bravais Pearson test are reported in Table 4. They are very similar to those computed using Mann-Kendall test (not shown). Correlation values are often low, explaining generally 25 to 30% of the hydrological series variability. But in some cases the relationship is much higher and half of the variance may be represented.
In the Amazon basin, there is a large discrepancy between runoff and rainfall results. Rainfall is positively related to SOI, with less rain during El Nino while there is no runoff-SOI signal. On the contrary, there is a negative signal between NATL and runoff indicating a higher runoff when the northern Atlantic SST is cooler than usual and no signal in rainfall. This discrepancy may be due to the rainfall gauge stations sampling that has been used to compute rainfall in the Amazon basin. Many stations were available in the Andes while few stations in the lowlands that represent the largest and rainiest part of the basins. Consequently, rainfall mainly represents the tropical Andes, while runoff the lowlands. That is why an Andean usual ENSO signal (reference) is found in rainfall and no signal in runoff, consistently with the near-absence of ENSO signal in the western Amazon basin. NATL signal is strong in the Andes but is also present in the lowlands (Figure 7d, Lavado et al. Submitted); that is why it is present in runoff. The Amazonian basins downward runoff trend put in evidence in the former chapter may be related to the positive trend in NATL observed in Figure 7b, as high temperature in the northern tropical Atlantic inhibits water vapour transport toward tropical South America.
In the Titicaca basin, rainfall is usually lower during El Nino, but, surprisingly, this signal is observed in the Ilave basin (TQ-3) only. A better signal, with lower discharge during El Nino is observed in runoff, especially in Qmax in the Ramis and Ilave basins (TQ-1 and TQ-3). This is consistent with the timing of El Nino events that peak during austral summer. As in the Amazon basin, there is a negative relationship between NATL or the Atlantic SST gradient and hydrological series. Though it is rather weak and not all the basins or all the variables are affected.
Along the coast, as the basins have been largely modified by human activity or by climatic change, and as the runoff values may be null in austral winter, correlations between minimum runoff and ocean-atmospheric indexes are not considered as reliable. The relationship between SOI and rainfall changes along the Pacific coast, being negative (more rainfall during El Nino) in the North and positive (less rainfall during El Nino) in the Southern basins, largely influenced by the tropical Andes signal. A similar positive signal is observed in Qmean and Qmax values of the southern basins but surprisingly the northern runoff are not related to SOI. Is it because gauging scales are destroyed during El Nino flooding. A further examination of ENSO – hydrology relationships in Peru will highlight this feature.
Rainfall, Qmean and Qmax in all the southern basins (from PQ11 for rainfall and PQ-16 for runoff until PQ-29) are negatively related to the SST gradient in the tropical Atlantic (NATL-SATL). This means that runoff is higher when the southern tropical Atlantic is warmer than usual and/or the northern tropical Atlantic is cooler than usual. In certain cases the relationship is rather strong (near 50% of the variance explained). The physical mechanism explaining this feature needs to be investigated.

Table des matières

I. CHAPITRE 1. Introduction
1.1 Contexte général du Pérou
1.2 A basin-scale trends in rainfall and runoff in Peru (1969-2004): Pacific, Titicaca and Amazonas drainage.
1.3 Les programmes de recherche dans lesquels s’inscrit la thèse
1.3.1 Le programme HYBAM (Hydrogéodynamique du Bassin Amazonien)
1.3.2 Le SENAMHI (Service National de Météorologie et Hydrologie de Pérou)
1.4 Sélection de bassins versants
1.5 Objectifs de la thèse
1.6 Organisation de la thèse
II. CHAPITRE 2 .Les basins de l’Ucayali et du Huallaga
2.1 Aspects climatiques
2.1.1 Précipitations
2.1.2 Température et Evapotranspiration
2.2 Aspects géomorphologiques
2.2.1 Relief
2.2.2 Géologie
2.2.3 Sols
2.2.4 Végétation
2.2.5 Aspects socio-économiques
2.3 Réseau hydrologique du SENAMHI et lesmesures hydrologiques récentes acquises dans le cadre du programme HYBAM
2.4 Sous basins sélectionnés sur les bassins Ucayali et Huallaga.
2.5 Constitution de la banque de données
2.5.1 Données hydrologiques
2.5.2 Données pluviométriques
2.5.3 Données d’évapotranspiration
2.5.4 Données des sols FAO
2.5.5 Constitution des grilles
2.6 Conclusions
III. CHAPITRE 3 .Variabilité hydroclimatique sur les bassins des rios Ucayali et Huallaga
3.1 Recent trends in rainfall, temperature and evapotranspiration in the Peruvian Amazonas-Andes basin: Huallaga and Ucayali basins.
3.2 Variabilité spatiale des débits sur les bassins de l’Ucayali et de Huallaga
IV. CHAPITRE 4 .Modélisation de bilan hydrique au pas de temps mensuel et impact du changement climatique sur les régimes hydrologiques dans les basins des rios Ucayali et Huallaga
4.1 Premiers essais de modélisation en utilisant des modèles de bilan hydrique mensuel
4.1.2 Monthly water balance models in the Amazon drainage basin of Peru: Ucayali River basin
4.2 Modélisation au niveau mensuel sur sept sous-bassins des rios Ucayali et Huallaga, et évaluation des impacts du changement climatique sur le régime hydrologique
V. CHAPITRE 5. Conclusion générale et perspectives
5.1 Conclusion
5.2 Perspectives
5.2.1 Validation des autres sources de données de pluie sur le basin amazonien péruvien.
5.2.2 La modélisation hydrologique au pas de temps journalier
5.2.3 L’impact des changements d’usage du sol sur l’hydrologie
5.2.4 L’impact du changement hydrologique au niveau journalier sur l’hydrologie.
5.2.5 Autres perspectives
Bibliographie

Télécharger le rapport complet