Logo Agrotecnio v2




This research line has been partially funded by:







Inici > English version > GRAP research > GRAP works > Spatial analysis of agricultural data

Spatial analysis of agricultural data

Yield mapping in vineyards

The use of sensors and yield monitors in vineyards has shown that grape production within a plot can be very variable. In fact, many growers are aware of this spatial variability, which affects not only production but also certain parameters related to grape quality. The consequence of this variability is a greater difficulty to predict the harvest and, in any case, the entry into cellar of a product that does not differentiate quality, limiting the opportunity to finally make and sell significantly different wine qualities.



Source: Spatial variability in grape yield and quality influenced by soil and crop nutrition characteristics, Precision Agriculture 13(3): 393-410; DOI: 10.1007/s11119-011-9254-1


Obtaining a yield map based on information provided by a yield monitor installed in a harvester is a relatively simple task. However, the proper interpretation of the map can be further complicated. In fact, before the specific pattern of distribution of the crop yield in their plots, the grower is more concerned about questions such as: Why the yield is so different within the same plot? What are the causes of this variability? If the production is variable, so will the grapes quality? Is it technically and economically justified, to perform a site-specific management of the plot according to the spatial variability contained?

The research started in the late 90s in countries such as Australia, South Africa, France, Chile, and later in Spain is enabling a new way to understand viticulture, where the classical idea of "plot" as the minimum territorial unit is replaced by the concept of "subplot" or management zone. Their limits are now dynamic and can be established each year after assessing different parameters and productive/qualitative based criteria.



Remote sensing and analysis of spatial variability of yield and quality parameters

The vegetation index PCD (Plant Cell Density) is used for classification of vigor in viticulture. The interest of estimating the vineyard vigor using remote sensing lies in its influence on the yield and grape quality. In Australia it has been shown (Lamb, 2001) that red vines with greater vigor and/or foliage density produce more grape but of lower quality due to the low concentration of phenolic compounds and pulp color. In the same vein, it has been proved the high correlation between the vegetation index PCD (obtained from aerial photography) and some qualitative properties of the juice, which expressed significantly in microvinification made from grapes harvested in different areas (Friedman test).


Source: Experiències en Viticultura de Precisió a Catalunya. Xarxa Temàtica de Viticultura de Precisió, 2009



Oportunity of selective vintage

Recent research in Precision Viticulture have shown that, in addition to the temporal stability of yield maps, the division of a plot into a small amount of different yield zones (2 or 3 zones) can also discriminate different grape qualities (generally, lower yield zones present higher grape quality than higher yield zones) (Bramley and Hamilton, 2004). Considering that usually the more productive the vines the more vigorous or the higher foliage development and considering the possibility of using satellite images to detect and classify different vigor zones within vineyard plots (based on radiometric indices such as NDVI or others), the later has been proved to be a reliable tool to discriminate within field quality zones.

The next figure shows a 5.01 ha Cabernet Sauvignon plot and the sampled vines for quality parameters. Previously, the NDVI image is classified into 2 levels using a cluster analysis. The analysis of variance for each quality parameter according to vigor clusters shows statistically significant differences between clusters validating the NDVI index as a parameter for zoning according to grape quality. The obtained results show that selective grape harvesting based on vigor zones is an option to be considered to discriminate grape quality and, hence, to make wines with statistically significant differences in quality.



Source: Review. Precision Viticulture. Research topics, challenges and opportunities in site-specific vineyard management, Spanish Journal of Agricultural Research (2009) 7(4): 779-790



Leaf area index mapping by using terrestrial laser scanners

An example of using LiDAR technology in agriculture is the creation of vegetation maps. Specifically, it is possible to obtain a leaf area index (LAI) map by scanning the vines in the field. The data acquisition is performed with a ground laser sensor (e.g. a SICK LMS 200 AG, Waldkirch, Germany), which is complemented by an inertial sensor and a GPS + RTK receiver. From the data obtained, using a specific algorithm for calculating (outlier filtering, calculation of vegetation surfaces) it is possible to estimate the LAI (for each meter of length along the row of vines) and its georeferenced locations in UTM coordinates. To obtain a final raster map a geostatistical interpolation (ordinary kriging) is required on a grid projection of 2 m. Finally, it is possible to zone the plot regarding the amount of vegetation (2 or 3 classes map) using a cluster analysis (fuzzy c-means algorithm) of interpolated LAI values.

  Source: Mapping the leaf area index in vineyard using a ground-based LiDAR scanner, 11th International Conference on Precision Agriculture (ICPA), 2012



Pest spatial distribution mapping for control and management

Another application of geostatistics in agriculture is the study of the spatial distribution of pests. The GRAP, along with IRTA and the Plant Protection Service of the Government of Catalonia, collects data on captures of certain pests in traps that are distributed throughout the area of Lleida fruit for over 10 years.

The technicians of Plant Protection Associations (ADV) are the ones maintaining the pheromone traps network. The data on captures are sent via email to the supervisory center for the production of digital maps of distribution of the pest (UdL) using geostatistical techniques. The distribution of the population in a given territory is characterized by a large number of variables that make its study complicated.

At the local and large-scale, geostatistics allows the study of most of these variables, such as the number of captures of Cydia pomonella (L.) males in pheromone traps in different flights at a given area. The study in 1996-97 was conducted in an area of ​​5,000 ha (Torregrossa ADV, with 55 traps). In 2001-2002 the study was expanded to 90,000 ha with 240 traps in 2001 and 500 in 2002, grouping different ADV of Pla d'Urgell . The result can the received by the producer or the technicians in three days.

Once the geostatistical study is done, it is possible to map the incidence of the pest in the region. To be able to relate and find correlations between different years, geographic information systems (GIS) were used. The inclusion of certain information layers (hydrology, moisture, roads, distribution of the pest, etc.) will allow the study of interrelations between them as well as the study of different pest dispersion indices. It is also important to take into consideration different landscape elements that can be incorporated and studied using high mapping accuracy.


Geostadistica_plagues_1    Geostadistica_plagues_2



Weed spatial Mapat de la distribució espacial de males herbes

Complementing the works on identification and classification of weeds, once the system is developed it is possible to georeference the data acquired in order to map the positions of the weed patches within the field. The analysis of this information will help the growers or assessors to better design the best estrategy to adopt:

  • No herbicide application
  • Application of an even dose of herbicide in the whole field
  • Selective application of herbicide, spraying only where required


The following figure shows a map of the spatial distribution of weed classified into monocods and dicods:


   Source: Weed discrimination using ultrasonic sensors. 2011. Weed Research 51 (6), 543-547. DOI: 10.1111/j.1365-3180.2011.00876.x





Last modified: 13/05/2015
Imprimir Enrera Pujar

Home | Accessibility | About the web | Site map | Legal notice | Contact
© 2014 Universitat de Lleida - Pl. de Víctor Siurana 1, 25003 Lleida - Tel. (+34) 973 702 000 - All rights reserved