Integrated spatial assessment of wind erosion risk in Hungary

Wind erosion susceptibility of Hungarian soils was mapped on the national level integrating three factors of the complex phenomenon of deflation (physical soil features, wind characteristics, and land use and land cover). Results of wind tunnel experiments on erodibility of representative soil samples were used for the parametrization of a countrywide map of soil texture compiled for the upper 5 cm layer of soil, which resulted in a map representing threshold wind velocity exceedance. Average wind velocity was spatially estimated with 0.5 resolution using the Meteorological Interpolation based on Surface Homogenised Data Basis (MISH) method elaborated for the spatial interpolation of surface meteorological elements. The probability of threshold wind velocity exceedance was determined based on values predicted by the soil texture map at the grid locations. Ratio values were further interpolated to a finer 1 ha resolution using sand and silt content of the uppermost (0–5 cm) layer of soil as spatial covariables. Land cover was also taken into account, excluding areas that are not relevant to wind erosion (forests, water bodies, settlements, etc.), to spatially assess the risk of wind erosion. According to the resulting map of wind erosion susceptibility, about 10 % of the total area of Hungary can be identified as susceptible to wind erosion. The map gives more detailed insight into the spatial distribution of wind-affected areas in Hungary compared to previous studies.


31
Wind erosion represents a serious problem worldwide: according to a report by the United Nations Environment 32 Programme in 1991, the phenomenon of wind erosion is responsible for more than 46% of the total degradation 33 of arid areas (Zheng, 2009). According to Lal (1994), the total agricultural area affected by wind erosion adds is richest in organic matter and nutrients (Funk & Reuter, 2006). A substantial consequence is the decline in the 40 productivity of endangered areas. Furthermore, the transport of nutrients and pre-sowing herbicides by wind 41 erosion can also be considered as a serious environmental problem (Funk et al., 2004). 42 Wind erosion is usually a natural, geological process which forms many eolian landforms (Lancaster 1995  The vegetation can also increase the soil moisture content through shading effect. Bare soil and arable lands are 52 wind erodibility measurements. They took into consideration wind velocity above 9 m/s to represent erosive 66 wind, however this value differs in the case of different soils. 67 Degradation of land caused by wind erosion strongly depends on the texture of topsoil, therefore mapping of 68 deflation requires the knowledge of soil texture of the uppermost soil layer in proper spatial detail (Borrelli et 69 al. 2014a, Mezősi et al. 2015). Soil texture can be assigned the most accurately by determining particle size 70 distribution (PSD). According to their size, different particles can be categorized as clay, silt, or sand. The size 71 intervals are defined by national or international textural classification systems. Soil textural classes are defined 72 by the numerical proportion (weight percentage) of the sand, silt, and clay separates in the fine-earth fraction (≤ 73 2 mm). The division is used to be depicted on a triangle diagram, the so-called 'texture triangle'. If the percentage 74 for any two of the soil separates are known, the correct textural class is determined; simultaneously, the sum of 75 the three percentages must total 100 percent. The most commonly used (also in wind erosion models) among

84
Furthermore the available auxiliary environmental information have been persistently widened. Unique digital 85 soil (related) map products can be compiled that were never mapped before, even nationally with relatively high 86 spatial resolution, taking also into consideration accuracy and reliability (Pásztor et al. 2015).

87
Our aim was to provide a nationwide, spatially detailed assessment of the susceptibility of the land of Hungary 88 to wind erosion integrating actual and representative wind tunnel measurements, the latest products provided by 89 both digital soil mapping and digital climate characteristic mapping, furthermore the most recent land cover map 90 provided by remote sensing.  In Hungary, large areas are covered by sandy and silty soils, which are mainly affected by wind erosion. More 98 than 60% of the relatively flat area is under agriculture cultivation which enlarges the exposure to wind erosion 99 causing serious soil degradation. The mean annual precipitation varies between 500 and 700 mm; the average 100 temperature is 10-11 ºC (1961-1990) (Péczely, 1998). The countrywide yearly average wind speed is 2-4 m/s.

101
The monthly average wind velocity increases continuously in the first months of the year, and the highest 102 monthly average wind velocity can be experienced in March and April (when the agricultural fields are bare).

103
The average wind velocity reaches its maximum in April and number of days on which the maximum wind 104 speed is over 10 m/s is also the highest in this month. The average wind velocity is 3.0-3.2 m/s in the months, 105 which are the most vulnerable to wind erosion (March-April). The main wind direction of winds above 5 m/s is    Wind tunnel measurement data 117 The upper 0-20 cm, ploughed layer was sampled at 215 sites collected from different parts of Hungary (Fig. 2a).

118
Soil texture was the primary selection criterion for the assignment of sampling locations. Another important 119 aspect was that each type should be represented by multiple samples; therefore, possible differences between 120 samples belonging to the same texture class became comparable.

121
Wind tunnel measurements were carried out to determine the erodibility of Hungarian soils with different soil 122 textures. The amount of the transported material by wind was calculated from the weight difference between the 123 samples before and after the experiment. Weight loss was normalized to an erosion modulus (ton hectare-1 min -124 1 ) to quantify the wind erosion transport rate. Erosion modulus for velocity was calculated by dividing average 125 accumulative soil loss by duration. The applied wind velocity was 16 m/s (which is the maximum wind velocity 126 available in wind tunnel) According to the measured data we created three erodibility categories on the basis of 127 the amount of transported material in an empirical way: 128  strongly erodible: 3200 -1500 gram/5 minutes (64 ton hectare-1 min -1 ); 129  moderately erodible: 1000-1500 gram/5 minutes (30 ton hectare-1 min -1 ); 130  slightly erodible: 0-1000 gram/5 minutes (20 ton hectare-1 min -1 ).

132
Threshold wind velocity data were evaluated according to the texture classification of the samples (Table 1), 133 whose result was applied in the parametrization of the countrywide soil texture map compiled for the uppermost 134 (0-5 cm) soil layer.    Table 2. categories were merged in order to increase the correlation with soil particle size distribution.   Due to the immense computation demand of gridding, the MISH method limits the number of predictor series 219 to 2,000, the length of series to 4,000 and the predictand locations to 10,000. The predictor series were the 72 220 station series, but speaking of hourly data they had much more than 4,000 values each and the predictands of a 221 0.5´grid were much more than the limit as well. The problem of the length of series was overcome by splitting 222 the series to fragments: gridding of hourly data was done separately for each month; moreover each month's 223 data were split into three parts to fit the 4,000 limit of length. In order to meet the limit of predictand locations 224 the grid had to be truncated. Considering that the spatial variance of wind speed is much less above flat terrain 225 an iteration formula was developed to determine the best grid network that has at most 10,000 grid points, but According to CLC50 about 56% of Hungary is under agricultural cultivation while about 44% of the country's 242 area is featured by land characteristics, which are more resistant to wind erosion. Since wind erosion does not 243 typically occur in forests, urbanized areas and over water surface, consequently these areas were masked out. hourly level during the observed 13 years in each point of the grid network (Fig. 5). According to the map, 250 spatial variability is relatively high throughout the country. Values in general range from 0% to above 2.5% in 251 relation to wind climatology, landscape, soil properties and land cover.

252
Most of Hungary has values smaller than 0.5%; moreover, a significant portion of the country presents ratios close to the north-eastern corner of the country. Wind speeds exceeding critical values are somewhat more 259 prevalent in these regions, but ratios higher than 2% are still exceptionally rare. In the western half of Hungary, 260 however, we can find more outstanding values. Around the Lake Balaton, especially in the Transdanubian

261
Mountains values higher than 1% are relatively common, and several smaller areas present ratios higher than 262 2.5%, indicating a significantly higher probability of critical wind stress. The ratio of hourly wind speed exceeding critical value can already be considered as a proper indicator of wind 268 erosion susceptibility. However basically, these values were inferred for the applied 0.5´ spatial resolution grid, 269 which cannot be actually considered as a real map. To create a spatially exhaustive map, the calculated values 270 were further interpolated using co-kriging to a 1 ha spatial resolution grid. Sand and silt content of the uppermost 271 (0-5 cm) soil layer, formerly used for the compilation of the reference soil texture map, were used as appropriate 272 numerical co-variables. The final map (Fig.7) was produced by masking out areas, which cannot be exposed to 273 wind erosion due to their land use/land cover characteristic.  The fine continuous scale of the wind erosion susceptibility map together with its high spatial resolution is not 286 necessarily applicable for decision making in land management and spatial planning. Consequently, we created 287 a simplified version of the map (Fig.8). We classified the ratio of hourly wind speed exceeding critical value 288 into three categories based on statistical properties of its distribution, which was supplemented with a fourth 289 category of non-erodible areas according to CLC50.    The applied climatic parameter, namely ratio of hourly wind speed exceeding critical values, proved to be a be the proper implication of these factors into process models and scenario based runs of the developed models. 351 We see some further possibilities for the improvement of the presented approach. It would be a major step 352 forward to functionally relate the resistance of soils to wind erosion with their erodibility factor (EF), which can 353 be calculated from basic soil properties (sand, silt, clay, organic matter and carbonate content) instead of leaning 354 on texture classes. In this case critical threshold values could be estimated directly by EF and also indirectly by 355 widely used soil data. Since nationwide soil property maps of these parameters have been very recently 356 compiled, they could support a new approximation for mapping wind erosion susceptibility on national level.

357
Eliminating the application of class averages the expected accuracy of both thematic and spatial prediction is 358 suggested to be improved.