COST 718 New technologies for agrometeorological model applications CNR Research Area Polo Scientifico di Sesto Fiorentino 14th february 2005 Integration of GIS and RS techniques for canopy variability evaluation in vineyards Niccolò Dainelli Consultant. dainelliniccolo@virgilio.it Claudio Conese - CNR Ibimet. C.conese@ibimet.cnr.it Bernardo Rapi CNR Ibimet. B.Rapi@ibimet.cnr.it Maurizio Romani CNR Ibimet. M.Romani@ibimet.cnr.it 1
Bacchus project, a brief summary BACCHUS is a Research and Technological Development (RTD) project cofunded by the European Community (EC) within the Energy, Environment and Sustainable Development (EESD). BACCHUS main scientific aim is to provide vineyard management organisations with an integrated and comprehensive solution to meet their information requirements, based on the use of Very High Resolution (VHR), Remote Sensing data, Geographical Information Systems (GIS) and modern software programming languages, improving current methodologies for: - vine areas location; - parcels identification; - vine characteristics specifications; - vineyard inventory - vineyard management Implemented and tested through a pilot system addressed to vineyard managers. Bacchus should be an instrument for land control and management, planning support and production evaluation in Controlled Origin Denomination areas. 2
Bacchus Partnership 14 companies, institutes, public agencies and regulating organisations belonging to some of the main wine producers regions in Europe: Spain, France, Portugal, Italy. European SpaceAgency (ESA) Project promoter Research and Development GEOSYS, Instituto de Desarrollo Regional Universidad de Castilla La Mancha, Instituto Nacionalde Técnica Aeroespacial, CEMAGREF, Instituto Geográfico Portugues Istituto di Biometeorologia CNR Commercial Instituto da Vinha e do Vinho Junta de Comunidades de Castilla La Mancha ONIVINS GéoDASEAFRCA Fédération Régionale de la Coopération Agricole Consorzio per la Tutela del Vino Prosecco di Conegliano-Valdobbiadene Consorzio Tutela Denominazione Frascati 3
VISAVES: a tool for Vineyard Suitability and Varability Evaluation Basic data: Geographic Meteorological Edaphic Agronomic Cadastral High Res RS image VISAVES: general scheme Geographic DB Variability Analysis Simulation Model GIS ENVIRONMENT Land Analysis Morphological factror System Setup General Settings General Scenario Local Scenario Analysis Scenario Management Factor Soil Factor Vegetational Factor Climate Factor Land Factor Land Suitability MAPS of spatial and temporal variation of LAI (Leaf Area Index) Plant Vigour (biomass) Maps of final variability of Grape Production Grape Quality Land suitability MAPS 5 class of land suitability on the base of specific vine exigencies. 4
VISAVES: main components LAND ANALYSIS MODULE 5
VARIABILITY ANALYSIS MODULE 6
The test site: The Prosecco DOC area 7
Boundaries of the DOC area 8
VISAVES Land Analysis Elevation Factor Soil Factor Slope Factor Climate Factor Solar Radiation Factor 9
VISAVES Land Analysis: Final result 10
VISAVES MODEL OUTPUTS: POTENTIAL VALUES Green biomass (g/m 2 ) Grape yield date Yield for each vine (kg) Quality index 11
VISAVES MODEL OUTPUTS: POTENTIAL VALUES CORRECTED BY MORPHOLOGY 12
13
VISAVES MODEL OUTPUTS: VALUES CORRECTED BY MORPHOLOGY Detail at the parcel scale 14
Objectives of the study Main objectives of the study were: To check the potentialities of Very High Resolution Remote Sensing Images for the detection of canopy variability within vine parcels To integrate the information carried with Remote Sensing images with the results of the VISAVES agrometeorological model 15
Reference works For the present study, published works from australian and american researchers have been taken into account as guidelines for the methodology. Below the most significant of these works are listed: Airborne remote sensing of vines for canopy variability and productivity (2002?) Lamb, Hall & Louis A method for extracting detailed information from high resolution multispectral images of vineyard Hall, Louis & Lamb Image-based decision tools for vineyard management (2003) Johnson, Pierce, DeMartino, Youkhana, Nemani & Bosch. 2003 ASAE Annual International Meeting, Las Vegas, USA 16
Available data: Methodology Two QUICKBIRD scenes, orthorectified, with panchromatic and multispectral bands, acquired in late June and early August 2004 An AMDC airborne sensor image, orthorectified, with panchromatic and multispectral bands, acquired in early June 2004 GIS layers: vineyard cadastre, DEM, boundary of study area, Quickbird specifications AMDC specifications Bands 1. Blue 420-495 nm 2. Green 515-565 nm 3. Red 600-670 nm 4. Near IR 738-772 nm Pan 530-680 nm Spatial resolution: 0,8 m (resolution depends on flight altitude) 17
Data processing: Conversion to Top of Atmosphere Spectral Radiance, in order to compare images taken in different periods of the year or images of different sensors Calculation of Vegetation Index NDVI NDVI = NIR NIR + R R Tentative application of several image sharpening techniques: enhance spatial resolution of low-res multispectral images with hi-res panchromatic images NDVI subsetting with vineyard cadastre: image statistics are calculated only for vine areas, variability is enhanced 18
Results obtained QuickBird NDVI variability within each parcel and among different parcels NDVI calculated from QB images is able to put in evidence a canopy vigour variability within each parcel and among different parcels. BUT some issues have been identified: Confrontation among different parcels at a single date can frequently be influenced by: Morphology, especially aspect: parcels with different solar illumination show different NDVI values. Cloud cover: cloud shadows tend to drop NDVI values. Ground resolution of multispectral images (2.8 m) is not high enough to identify each single vine row. The pixel carries the spectral information of both the vine row and the interrow. 19
20
21
22
Consequences of the issues previously shown: Parcels in different morphological contexts, at the moment cannot be compared Only those parcels in a limited area with homogeneous morphology and illumination conditions can be analysed together Ground resolution of Quickbird multispectral images is not sufficient to permit an analysis of canopy variability at the vine row scale. Variability within each parcel is shown but with a low precision. To solve the problem of spatial resolution, some image sharpening techniques have been applied to Quickbird scenes (Principal Component, Brovey, Gram-Schmidt) In spite of a generally good-looking appearance, these techniques did not give reliable results, since pixel values were sistematically altered. 23
AMDC NDVI variability within each parcel and among different parcels NDVI calculated from AMDC images gives far better results than QuickBird. Due to its 0,8 m pixel, each vine row can be resolved. The problem of mixed pixels (vine/interrow) is severely reduced The issues concerning the negative effects of morphology still remain 24
25
AMDC SPECTRAL PROFILES THROUGH VINEYARDS 26 QUICKBIRD
Multi-temporal analysis of variabilty Using AMDC and Quickbird NDVI images, a multitemporal analys has been carried out. Zonal statistics has been applied to NDVI for each parcel, in order to extract the MEAN NDVI value and a COEFFICIENT OF VARIATION The application of this methodology can be of interest for farmers who want to know which are the parcels where variability is higher, usually those where intervention is most needed. 27
EARLY JUNE 28
LATE JUNE 29
EARLY AUGUST 30
Integration with the VISAVES model Potential values of green biomass calculated by the VISAVES agrometeorological model are corrected with the NDVI images, representing the actual vigour situation for a given date. 31
CONCLUSIONS Vineyard variability evaluation by Remote sensing is a field of study in continuous development, thanks to the advances in Very High Resolution multispectral sensors designing At present, however, commercial VHR satellite-borne sensors still cannot achieve a ground resolution which can allow a precise vine row canopy variability analysis Airborne multispectral sensors, allowing for variable ground resolution according to flight height, perform much better and probably represent a more suitable tool for wine farms or consortiums. In fact, besides a better spatial resolution, airborne sensors can be flown at will, both in time and space. Variability analysis of NDVI images at parcel scale can represent an useful tool for farmers to identify those vineyards where variability is higher and intervention is needed Some issues negatively affecting NDVI variability need further examination: The effect of morphology (mainly aspect) The effect of cloud shadows 32