**2.2.3 Steps of the modeling**

The first step aims to convert all input data into erosive pressure levels (Fig. 3). 6 levels of pressure had been specified, from 0 to 5. Level 0 indicates absence of pressure and level 5 refers to a very high pressure level.

Some types of input data have only 5 levels. In this case, there is no level 0. This applies to "intercrop period", "daily intensity of rainfalls", "intercrop management" and, "yearly

SCALES: An Original Model to Diagnose Soil Erosion Hazard

Fig. 4. Defining hazard levels using SCALES model

**2.2.4 Integration and aggregation of SCALES data** 

800 ha) or as hydrologic units (approximately 2,000 ha).

154,400 ha).

and Assess the Impact of Climate Change on Its Evolution 233

The SCALES model incorporates specific cases that don't follow the general rule. The level 0 of combined data, referring to agricultural practices for permanent grassland and orchards and the level 0 of "slope" type when gradient is lower than 1 %, are automatically excluded from the treatments. These levels are directly converted into level 0 of erosion hazard. This

SCALES is a large-scale assessment model intended for mapping soil erosion hazard at the finest level of organization of the agricultural area. It supposes an integration of input data at parcel unit scale. Their area can exceed a few hectares but are generally lower than this unit of reference. This scale seems to be the better scale to determine the spatial context of soil erosion in Basse-Normandie. The layer of parcel units exists in the form of a vectorized and geo-referred database and such data came from the Inventory of Common Agricultural Politics given by the agricultural administration. Data inform on the land cover types in 2006 (source: Rpg\_anonyme\_014\_AUP\_2006). The parcels are classified in the three following: Grassland and arboriculture parcels (26,500 parcels, 111,000 ha) Temporary grassland parcels (13,200 parcels, 99,600 ha) and Crop parcels (25,800 parcels,

The integration of data in the parcel units requires beforehand a mapping of input data under raster format. We used the principle of allocating for each parcel unit and for each input data type, only one mode. When parcel unit holds several modes, we chose using a decision rule, to select the spatially most dominant mode. The application of the dominance rule has been carried out using the module *Spatial Analyst* in ESRI ® ArcView Gis 9.2. Obtained maps concerning the rainfall erosivity, the potential sensibility of areas and the soil erosion hazard are vector maps. Those are transformed to raster format for incorporating the output data in larger vector units as administrative units (approximately

conversion does not require the agricultural areas to fulfill these two conditions.

positive hydrological balance" types. Regardless of their characteristics, these four types of data generate favorable conditions for triggering soil erosion. Therefore, even with very short intercrop periods, soils will always be exposed to the erosive effect of the rain. Also, in spite of protecting practices during intercrop period such as implantation of plant cover, the time required by the plants to grow gives a period during which the bare soil stays unprotected from erosion agents. Concerning rainfall erosivity, the weather conditions of a mild maritime climate rule out the absence of rainy events during the intercrop period, that is to say between September and April.

Level 0 has been affected to "grassland" and "slopes" input data. The presence of a permanent plant cover such as grassland always protects soil from erosion, even if it is common to observe some runoff on permanent grassland. Parcels dedicated to orchard culture are also included in this category. For slopes, the level 0 refers to topography with surface gradient lower than 1%. In that case, the slope does not cause the surface water to flow, preventing all possibilities of soil erosion by water.

The second step consists of combining the pressure levels of input data following an additive approach (Fig. 3). Between two types of input data, every combination is conceivable. Summations are included in an interval from 0 to 10. These values are subsequently classed in the following categories: ≤ 2, 3-4, 5-6, 7-8 and 9-10. Each category is then reclassified into simple value equivalent to a combined level of erosive pressure. Combined levels can later be combined again with other input data or other combined levels. In any case, the combination and simplification process remains the same.

Fig. 3. Levels of erosive pressure and combination procedure in the SCALES model

The third step leads to the estimation of hazard levels. It implies to know the structure of the model (Fig. 4). SCALES is a tree form model, which means that input data are organized in a hierarchy according to their influence on the genesis of erosive runoff. Arguments in favor of this organization are the same as those previously exposed about the choice of input data. Therefore we can notice that the "intercrop period" type for example has a lower impact on triggering erosive runoff than the "slope" type, which has itself a lower impact than "structural instability".

Even if this organization differs from a weighting using coefficients, the classification of input data associated with additive approach make good case for this idea. The weighting occurs at every combination until the final hazard level. The weight of input data is always divided by two during the first combination, and then divided by two again with the next combination. According to this way of operating, we notice that the impact of pressure level of input data on final hazard is decreasing significantly when input data are lowered in the proposed hierarchy.

positive hydrological balance" types. Regardless of their characteristics, these four types of data generate favorable conditions for triggering soil erosion. Therefore, even with very short intercrop periods, soils will always be exposed to the erosive effect of the rain. Also, in spite of protecting practices during intercrop period such as implantation of plant cover, the time required by the plants to grow gives a period during which the bare soil stays unprotected from erosion agents. Concerning rainfall erosivity, the weather conditions of a mild maritime climate rule out the absence of rainy events during the intercrop period, that

Level 0 has been affected to "grassland" and "slopes" input data. The presence of a permanent plant cover such as grassland always protects soil from erosion, even if it is common to observe some runoff on permanent grassland. Parcels dedicated to orchard culture are also included in this category. For slopes, the level 0 refers to topography with surface gradient lower than 1%. In that case, the slope does not cause the surface water to

The second step consists of combining the pressure levels of input data following an additive approach (Fig. 3). Between two types of input data, every combination is conceivable. Summations are included in an interval from 0 to 10. These values are subsequently classed in the following categories: ≤ 2, 3-4, 5-6, 7-8 and 9-10. Each category is then reclassified into simple value equivalent to a combined level of erosive pressure. Combined levels can later be combined again with other input data or other combined

levels. In any case, the combination and simplification process remains the same.

Fig. 3. Levels of erosive pressure and combination procedure in the SCALES model

The third step leads to the estimation of hazard levels. It implies to know the structure of the model (Fig. 4). SCALES is a tree form model, which means that input data are organized in a hierarchy according to their influence on the genesis of erosive runoff. Arguments in favor of this organization are the same as those previously exposed about the choice of input data. Therefore we can notice that the "intercrop period" type for example has a lower impact on triggering erosive runoff than the "slope" type, which has itself a lower impact than

Even if this organization differs from a weighting using coefficients, the classification of input data associated with additive approach make good case for this idea. The weighting occurs at every combination until the final hazard level. The weight of input data is always divided by two during the first combination, and then divided by two again with the next combination. According to this way of operating, we notice that the impact of pressure level of input data on final hazard is decreasing significantly when input data are lowered in the

is to say between September and April.

"structural instability".

proposed hierarchy.

flow, preventing all possibilities of soil erosion by water.

Fig. 4. Defining hazard levels using SCALES model

The SCALES model incorporates specific cases that don't follow the general rule. The level 0 of combined data, referring to agricultural practices for permanent grassland and orchards and the level 0 of "slope" type when gradient is lower than 1 %, are automatically excluded from the treatments. These levels are directly converted into level 0 of erosion hazard. This conversion does not require the agricultural areas to fulfill these two conditions.
