REMOTE SENSING AND GIS METHODS TO DETECT TREE HEDGEROWS IN AGROFORESTRY LANDSCAPES

Agroforestry denotes land use systems in which trees grow in combination with agricultural crops and/or livestock. The woody component usually consists of scattered or linear trees (planted or naturally growing), that can be located either inside the field or along the field boundaries, as tree hedgerows. This land use approach is aimed at the optimization of both ecological interactions and economical revenue. Agroforestry is increasingly perceived as providing ecosystem services, environmental benefits, and economic commodities as part of a multifunctional working landscape. These services and benefits occur over a range of spatial and temporal scales: from the farm/local scale, through the landscape/regional scale up to the global scale. The use of Remote Sensing and GIS spatial analysis is of the utmost importance for detecting landscape patterns, understanding the interactions between biological and physical components, and for assessing, mapping and quantifying the socio-economic values of the agroforestry systems services. Agroforestry systems have traditionally been used in different places of Europe employing several types of practices at different levels of intensity. However, a decline of this land use system occurred all through the 20th century due to agricultural intensification and mechanization. To slow down the decrease of these good practices, the European Common Agricultural Policy is currently supporting establishment and preservation of agroforestry systems, because of their higher ecological and socio-economic value. Most of the Italian territory is naturally suited for agroforestry systems due to its environmental setting, geomorphological and climatic conditions, as well as for the historical and cultural land management practices. Although that, few information is available concerning the current extent of agroforestry systems because tree detection in large agricultural areas is time consuming. Additionally, the magnitude of ecosystems services of agroforestry systems strongly depends on tree number and dimension. Thus, detecting these tree parameters by remote sensing is of dramatic importance. This study is focused on an agroforestry landscape located in the Umbria region (central Italy), where we investigated a farm managing over 600 ha of arable land and woods. The main land uses include herbaceous crops, tree hedgerows, shelterbelts and forest belts. In these systems, trees grow only at the edges of fields, within hedgerows, or on scarps and drainage ditches between fields; the trees provide established positive effects on soil erosion, wind shielding and ecological enrichment as well as an aesthetic enhancement of the landscape. We combined different methodologies comprising Remote Sensing, photo interpretation, GIS analysis and field survey to detect the spatial distribution of the land cover/use of the study area and to reveal the spatial interactions between the crop and tree components of the system. In particular, we used the hemispherical canopy ground photography technique to assess the shading effect of trees on crops. The aims of this study were: i) to map and estimate the extant of Tree Hedge Rows (THR) in the study area; ii) to quantify the influence of THRs on the yield of crops at the plot scale. Basing on the land use classification, performed by photo interpretation (Data source: AGEA 2011), we identified two experimental sites (ES) to study the continuous and discontinuous THRs along the margins of the cultivated fields. Each site contains a plot of annual crops and THRs along at least one of the borders consisting of oaks, mainly Quercus pubescens and Quercus cerris. Through the aerial photos (2011) and Google Earth images (2017), we identified two test areas (TA) of 100 ha (1km x 1km squares) each one containing one of the two ES. We tested a procedure for the GIS inventory of THR over the two TAs, consisting in: 1) survey of THRs with GPS device in the ES, for the proper georeferencing and actual measurement of the tree linear systems, measurement of height (H), diameter at breast height (DBH) for each tree of the THRs and of the distance between adjacent individuals; 2) recognition of THRs by photo interpretation of aerial and satellite images; 3) comparison of field measurements against estimates by photo interpretation for the error evaluation; 4) estimation of the incidence of THRs per hectare of cultivated area over the two TAs and over the whole farmland. The recognition of THRs was based on photo interpretation of high-resolution multispectral Sentinel2 (HRS2) images. In particular, evaluating the NDVI (Normalized Difference Vegetation Index, NDVI = (NIR-VIS) / (NIR + VIS)), we could easily discriminate between areas with dense vegetation coverage (0.6 <NDVI <0.9, tree covered) and areas with low/zero vegetation cover (cultivated areas or bare soil areas). Starting from the HRS2 images and using the raster algebra of the SNAP (Sentinel Application Platform), the NDVI was derived and the corresponding raster file was generated for the TAs. The vectorization of the raster file generated a vector file of polygonal elements that were classified according to the NDVI. The 10m spatial resolution of HRS2 scenes allowed the identification of long and narrow polygons corresponding to the crowns of trees. The THRs in the two TAs were identified and confirmed by the photo interpretation, subsequently the THRs have been validated by comparison with the field GPS surveys. This procedure was applied throughout the study area to estimate the incidence of THRs per hectare. We also collected samples for yield estimations of crops adjacent to THR, along four transects (25m long) for each ES, two of which being under the influence of tree crowns, at increasing distances from the tree rows. During the summer 2017, we collected five wheat samples from each transect, amounting to a total of 20 plots per site (each one measuring 1 m2). 24 hemispherical photos were also taken along the four transects of each ES. Using these photos, we estimated the light transmission during the growing season in relation to the canopy structure of the tree rows, indirectly calculating the effect of the trees' shade on crops. We used the Gap Light Analyzer software to analyze the digital hemispherical canopy photos. Our results show that, in the study area, the 14% of the total fields' perimeter is covered by THRs dominated by oaks, consisting mostly of adult trees with a high aesthetic added value. The linear density of THRs is variable, amounting to an average value of 67 m / ha. The existing THRs along the field boundaries also play an essential ecological function, connecting the otherwise fragmented forest patches. The effects of the trees on the yield of wheat as an adjacent crop were inconclusive, but they indicate that these effects were at least not entirely negative. To assess the conceivable effects of the THRs on crops in greater detail, further studies with an increased number of transects should be performed.