The input data for the first script consist in a vector map containing the bankfull channel limits and the limits of longitudinal and lateral bars. The area between the bankfull limits, with the exclusion of the areas occupied by bars, defines the active subchannels (Figure 1).

In order to have fixed reference points, in space and time, which the channel characteristics and changes from date to date refer to, a set of ground points along both borders of the channel must also be collected. These points, which can be digitized on a reliable map or acquired by GPS on the field, are projected perpendicularly to the channel centreline and reported along the x-axis of the graphics with their distances from one another and their cumulative distances from the origin.

In most works, the channel centreline, i.e. the line whose points are equidistant from the two limits of the channel, is traced manually. In a few works, specific algorithms have been proposed to automatically derive this feature by thinning a channel raster map pixel by pixel [4] [5] or to extract it from the bank lines in vector form

using Delaunay triangulation principle [6] . In [7] the centreline has been defined by joining a set of evenly- spaced points along the river channel, whose equidistance from the two banks has been determined by successive approximations. In GRASS, a specific module (v. centerline) is proposed to define a line that represents an approximation of the central tendency of a series of input lines, all of which with similar trajectories and, therefore, that can be used also to define the centreline of a channel represented by its two sides. However, the module gives good results when the lines are approximately parallel to each other. Moreover, it may give unreliable results when two reaches of the channel are very close to each other, as can happen at the neck of a meander.

In this work, a different approach is adopted and the centreline is extracted from the vector map containing the bankfull channel limits, manly by using the GRASS module that traces lines parallel to existing features. Starting from the bankfull channel margins, pairs of lines parallel to the channel banks are repeatedly traced out toward the channel centre. Since each of the two parallels traced out at each iteration is equidistant from the pertaining bank line, the intersection of the two parallels defines points that are equidistant from the banks (Figure 2(a1)-(a8)). The line joining these intersection points defines the channel centreline (Figure 2(b)). The length of the centreline is assumed as the length of the channel.

The channel width is commonly defined as the length of the line from bank to bank orthogonal to the channel centreline. In the above mentioned algorithms [4] [5] , channel width is automatically computed along the lines orthogonal to the pixels forming the computed raster centreline. In [8] , a Python script is described for the calculation of channel width at regularly-spaced transects orthogonal to the channel centreline. The same procedure is adopted in the script described in this work (Figure 2(c), Figure 2(d)).

Many different indexes have been proposed by various authors to express the degree of braiding of a river [9] . Among them, the so-called Channel count index [10] is the most commonly used one, since it is simple to calculate and is the least sensitive to river-stage effects [11] . It is simply expressed as the number of active subchannels along a channel transect.

Similarly to what is reported in the above cited paper [8] , in the script described herein this index is computed by simply counting the number of active channels along the same transects where the channel width is defined.

In its classical formulation, the sinuosity index is defined as the ratio of channel length to valley length [12] . It requires prior subdivision of the river valley into rectilinear segments. Then, the length of the channel centreline encompassed in the valley segment is divided by the length of the valley segment. The valley segmentation introduces a certain degree of subjectivity, especially where the valley shows a curvilinear path. To reduce the degree of subjectivity, in [8] a fixed length value is used for the valley segmentation. In this work, to further reduce subjectivity and obtain a more detailed assessment of sinuosity along the entire channel course, the river valley is ignored and only the channel centreline is used. More precisely, sinuosity is calculated by considering a portion of the channel centreline with fixed length, progressively shifted downstream by a constant distance. Sinuosity is obtained by dividing the fixed length of the centreline tract by the straight-line between its endpoints. The computed sinuosity value is assigned to the midpoint of the centreline tract (Figure 3).

The output of the first script consists of:

1) A screenshot of three simple graphs which allow a first rough check of the results (Figures 4(a)-(c)).

2) A vector map containing the channel centreline.

3) A report containing the main computed statistics (Table 1).

4) Three text files containing the values of width, braiding and sinuosity. In each file, the first column reports the progressive number of each measure, the second reports the distance from the origin and the third reports the parameter value (Table 2).

The second script computes the channel centreline lateral shifting between two dates. As it requires the pre- emptive construction of the channel centrelines as of the two dates by the previously described shell script, a specific shell script has been constructed. The script, starting from a vector map containing the channel centrelines as of the two dates, computes the lateral shifting using a procedure similar to the previously described one adopted to define the centreline between the two channel banks. Indeed, the parallels to the two centrelines are traced out and their intersection joined by a line that represents the line equidistant from the two centrelines. The distances between the two centrelines are then computed along lines orthogonal to the computed axis at regular intervals and are assumed as the channel centreline shifting between the two dates (Figure 5).

The output consists of a report (Table 3) and of a text file with the shifting values at regular interval along the computed centreline axis (Table 2). A rough simple graph is also displayed on the screen (Figure 4(d)).