Spatial variation in mean annual rainfall

Input data included mean monthly and annual precipitation and temperature data from 110 rain gauge stations. Figure 2 shows the distribution of the rain gauge stations in the area. The correlation between precipitation and elevation is extremely low (0.1), as expected in such a large area. Indeed, it has been demonstrated (Guida et al. 1980) that the correlation P-h is quite high only for small homogeneous areas (subzones), but is very low for a whole morphologically complex region (e.g. the Cilento area or Campania). The correlation between the annual rainfall at each station is extremely low (0.1), and between the annual rainfall of each station the standard deviation varies between 240 and 510 mm/a. The Pearson correlation values between the stations show the similarity of stations within a climatic zone rather than within the same mean elevation, according to the considerations previously expounded.

In this study to evaluate the variability of annual rainfall data, the rainfall rate was calculated for each year by interpolating the rain gauge data, using kriging interpolation techniques. The maps were constructed in a GIS environment, allowing the digital rainfall model (DRM) to be defined for each year (pixel: 100 m x 100 m).

Rainfall variability was observed by comparing the 49 yearly DRM raster maps (1951-1999) instead of operating as usual with point yearly data (related to each rain gauge station). The rainfall maps constructed by this method highlight the differences between areas. Moreover, the applied technique allows the problem of data paucity to

be overcome and minimizes the effect of occasionally rainy or dry years on the general rainfall pattern (Ducci 1999).

Statistical comparison between the 49 DRM highlighted the differences between the period 1951-1980 and the drier period 1981-1999. Figures 5 and 6 show the mean annual rainfall maps for the years 1951-1980 and 1981-1999, respectively. The rainfall maps were obtained by the average of the annual DRM raster maps for 30 years and 19 years, respectively. The rainfall maps constructed according to this method differ markedly in some sectors from the average rainfall maps obtained by interpolation of mean annual data.

For the whole region, the annual volume of rainfall is 16 000 Mm3 for the period 1951 to 1980, and 13 500 Mm3 for the period 1981-1999, with a decrease of about 15% (Fig. 7).

The selection of the method to compute the evapotranspiration was based on data availability at regional scale. Detailed methods such as the Penman-Monteith equation, recommended by the Food and Agriculture Organization of the United Nations, were not feasible to apply due to lack of data (especially crop-soil data).

Therefore, annual evapotranspiration (Er, mm) was computed by using Turc's empirical formula (Turc 1961), as several studies have confirmed the reliability of this formula for central and southern Italy (Bono 1993; De Felice et al. 1993; Dragoni & Valigi 1991, 1993):

1600 1500 ~ 1400

J 1300

.g 1200

< 1000 900 800

1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Fig. 7. Mean annual rainfall in Campania (1951-1999) with confidence bands for mean.

1 |
(i | ||||

I - |
V T |
\ |

where: P = yearly rainfall (mm/a); L = 300 + 25 T + 0.05 T3 (the capacity of the atmosphere to evaporate water); T = mean annual temperature (°C).

In the whole region the number of stations with reliable and continuous data is limited (about ten). Fortunately, the strong correlation between temperature and elevation allows the values of T to be combined with the digital elevation model (DEM) by linear regression. Hence the DEM is the best spatial descriptor for this climatic variable. The temperature maps were constructed in a GIS environment, for each year, from 1951 to 1999, on the basis of the DEM. The strong correlation between altitude and temperatures (0.9 in 62% and 0.8 in 93% of the years) emerges in Table 1,

Table 1. Regression equation for each year between elevation and temperature and correlation factor R2

Year Direction coefficient Intercept R2 Year Direction coefficient Intercept R2

Table 1. Regression equation for each year between elevation and temperature and correlation factor R2

Year Direction coefficient Intercept R2 Year Direction coefficient Intercept R2

1951 |
-0.0068 |
18.020 |
0.9488 |
1975 |
- 0.0068 |
17.090 |
0.8031 |

1952 |
-0.0070 |
18.393 |
0.9881 |
1976 |
- 0.0070 |
16.717 |
0.8026 |

1953 |
-0.0086 |
18.256 |
0.9803 |
1977 |
- 0.0065 |
17.448 |
0.8634 |

1954 |
-0.0085 |
17.625 |
0.9841 |
1978 |
- 0.0067 |
16.591 |
0.8636 |

1955 |
-0.0075 |
17.858 |
0.9767 |
1979 |
- 0.0075 |
17.717 |
0.9717 |

1956 |
- 0.0079 |
16.936 |
0.9718 |
1980 |
- 0.0077 |
17.168 |
0.9253 |

1957 |
- 0.0069 |
17.847 |
0.9779 |
1981 |
- 0.0074 |
17.694 |
0.9446 |

1958 |
- 0.0070 |
18.013 |
0.9360 |
1982 |
- 0.0068 |
17.806 |
0.8352 |

1959 |
- 0.0083 |
18.414 |
0.8287 |
1983 |
- 0.0061 |
16.807 |
0.7621 |

1960 |
- 0.0079 |
18.538 |
0.9670 |
1984 |
- 0.0066 |
16.747 |
0.8208 |

1961 |
- 0.0099 |
19.095 |
0.9188 |
1985 |
- 0.0069 |
17.918 |
0.8012 |

1962 |
- 0.0065 |
18.568 |
0.8919 |
1986 |
- 0.0073 |
17.794 |
0.7610 |

1963 |
- 0.0076 |
18.371 |
0.9678 |
1987 |
- 0.0069 |
17.884 |
0.6847 |

1964 |
- 0.0095 |
19.048 |
0.9801 |
1988 |
- 0.0068 |
18.093 |
0.6701 |

1965 |
- 0.0070 |
18.270 |
0.9340 |
1989 |
- 0.0075 |
17.681 |
0.9920 |

1966 |
- 0.0086 |
18.734 |
0.8751 |
1990 |
- 0.0074 |
18.835 |
0.9440 |

1967 |
- 0.0081 |
18.592 |
0.8529 |
1992 |
- 0.0077 |
19.097 |
0.9941 |

1968 |
- 0.0064 |
18.047 |
0.8788 |
1993 |
- 0.0079 |
19.218 |
0.9859 |

1969 |
- 0.0077 |
17.926 |
0.9848 |
1994 |
- 0.0084 |
20.301 |
0.9968 |

1970 |
- 0.0074 |
18.036 |
0.9910 |
1995 |
- 0.0082 |
18.136 |
0.8821 |

1971 |
- 0.0075 |
17.681 |
0.9920 |
1996 |
- 0.0085 |
18.357 |
0.8746 |

1972 |
- 0.0072 |
17.475 |
0.9366 |
1997 |
- 0.0083 |
18.907 |
0.9417 |

1973 |
- 0.0070 |
17.289 |
0.8094 |
1998 |
- 0.0073 |
17.604 |
0.9896 |

1974 |
- 0.0067 |
16.826 |
0.8608 |
1999 |
- 0.0074 |
18.849 |
0.9445 |

The year 1991 is not evaluated due to paucity of data.

The year 1991 is not evaluated due to paucity of data.

Fig. 8. Linear regression between elevation and temperature for the years 1951-1999.

Fig. 8. Linear regression between elevation and temperature for the years 1951-1999.

lower in the coastal plains. Statistical comparison between the 48 maps allowed its to detect the differences between the period 1951-1980 and the warmer period 1981-1999. Figure 9a and b show the mean annual temperature maps for the years 1951-1980 and 1981-1999, respectively. Evaluating the real evapotranspiration by the Turc method, in the last 20 years the percentage of evaporated water is 6% greater than in the 30 years before, increasing from 52% to 58% of the precipitation volume, due to the increase in temperature.

The map of the variation in total flow (runoff + recharge) obtained from the difference between precipitation and evapotranspiration is very significant (Fig. 10). Moreover the areas with a decrease of more than 200 mm are mountainous areas where permeable rocks (limestone, lavas) crop out. The role of lithology in the infiltration decrease is outlined below.

where the regression equation for each year and the corresponding R2 are reported. Figure 8 summarizes the trends of the linear regression with a medium gradient of about — 0.7°C each 100 m.

The temperature variability was observed by comparing the 48 yearly raster maps of the temperature (1951-1999, excluding the year 1991 with only two stations working). The temperature maps highlight the differences between areas according to elevation: the higher increases in temperature are in mountainous areas and the

Due to the difference in the position of the groundwater divide and the watershed divide, it is almost impossible to evaluate runoff directly, using data from river gauge stations. Recharge is estimated as a percentage of the surface and subsurface runoff as a function of the permeability of the outcropping rocks. The recharge coefficient cr ranges from 10-20% in impervious rocks to 100% in limestones with strong karstification. The lithological map of Campania (Fig. 1) was

used to evaluate the recharge coefficient on the basis of the permeability of the mapped units (Table 2). The coefficients are multiplied (pixel by pixel) by the difference between precipitation and evapotranspiration to compute the value of the groundwater recharge.

Lithologic unit |
Recharge |

coefficient (%) | |

Alluvial and pyroclastic deposits |
60 |

Lavas |
80 |

Pyroclastics |
40 |

Arenaceous-clayey-marly deposits |
20 |

Carbonate rocks (prev. Limestone) |
90 |

In the last 20 years the volume of infiltration has been 30% lower than in the previous 30 years, due to the decrease in rainfall and to the increase, in percentage terms, in evapotranspiration resulting from the temperature rise of about 0.3°C. This is in agreement with previous studies carried out in central Italy (Cambi & Dragoni 2001) and in a basin in Campania (Ducci et al. 2000).

The recharge maps (Fig. 11a and b) highlight the differences between areas. The areas most severely affected by the recharge reduction, where the decrease sometimes exceeds 50%, are the Matese carbonate area (in the northern part of the region), the area around Avellino, the Sorrentine Peninsula carbonate area (located west of Salerno) and the Cilento area (mountainous coastal area in the southern part of the region). The least affected area is the zone surrounding the town of Naples. In terms of differences, and hence the amount of

recharge, the zones most affected by the reduction are often in the Apennine chain, as these carbonate areas are the 'freshwater reservoirs' of Campania (Celico et al. 2004a). In these areas there are few rain gauge stations, but this lack of data has relatively little effect in determining recharge since the recharge reduction was verified on stations at high altitude with very long, continuous data series (such as the Montevergine rain gauge station, at 1287 m a.s.l.). In order to highlight this aspect, the recharge values (mm/a) were multiplied by the recharge area to obtain the total amount of infiltration. The difference (Fig. 12) is 1163 Mm3 (37 m3/s).

Variation in spring discharge and piezometric surfaces

The impact of precipitation on groundwater levels is more or less delayed and attenuated by porous media (Kresic 1997). To define the normal hydro-dynamic behaviour of an aquifer it is necessary to develop some stochastic models of the input-output type. Even the simplest stochastic model provides considerable information on the aquifer's structure and on the connections between hydro-logic variables.

In Campania there are several long series of piezometric and discharge data. For instance, the Acerra Capomazzo well (in the coastal alluvial Volturno river plain), which is a low-depth (about 15 m) well in fine-medium grained pyroclastic

Was this article helpful?

## Post a comment