GIS Modelling and Remote Sensing

Geographical Information System (GIS)

Defining a datum in space gives the datum a new meaning and value, because it enables spatial analyses and spatial presentation. The usefulness of spatial data in the field of plant protection was recognised relatively early on, since the first spatial analyses and presentations of the occurrence of plant diseases and pests were made in the second half of 1990s. The foundations of the organisation and cohesion of GIS in a comprehensive information system for the field of plant protection were established with the end of the CRP project “Building a Unified System for Monitoring and Analysing Pests in Agriculture” in 2001. With time, the scope of work at GIS increased to the level where it became an integral part of the organised support to the Phytosanitary Administration of the Republic of Slovenia. GIS is an irreplaceable tool in the management of special supervision of pests when issuing decisions and analysing the forecast effects of an individual measure. It enables (if required) the verification of effects and planning of the most rational solutions even before a decision on an individual phytosanitary measure is issued. With GIS analyses, we make lists of producers who have to be informed about an individual phytosanitary measure or help the phytosanitary inspection service in their supervision and notification of producers. Individual content of GIS is added to the phytosanitary spatial portal, where an external user can access the prepared content via a simple viewer.

Besides expert work, with our GIS content we also participate in research work, especially analyses of the spread and modelling the spread of individual pests (corn rootworm, pinewood nematode). (Link to the diseases)

Remote Sensing

Remote sensing could be defined as a science on acquiring data about an object, without direct contact, based on analysing the characteristics of electromagnetic waves. The term combines many techniques and methods, but primarily due to the fact that now there is a large presence of techniques of optical imaging of the objects with satellites or planes, the term remote sensing mainly denotes the above method. There are namely several techniques or methods. The method of optically capturing electromagnetic waves is also used as part of our work. By analysing the images, we can more or less reliably distinguish between the types of plants, their health condition and even various plant diseases.

Currently, we use high definition, multi-spectre satellite images from the WorldView 2 satellite in our work. In the analyses of images, we include complex statistical analyses or machine learning methods. In our research work, we focus on detecting the presence of quarantine form of the grapevine phytoplasma (grapevine flavescence dorée) in vineyards at test sites in the Primorska and Dolenjska regions.

Based on first results of analyses of high definition satellite images, we decided to intensively work on spectroscopic methods of detecting plant pathogens. Soon we will establish a laboratory for image spectroscopy that will enable even clearer demarcation between stress factors of plants, such as for example the presence of plant diseases and pests as well as detection of other stress factors (drought stress, nutritional disorders, etc.) In the laboratory, we will use two hyper-spectral cameras that will enable us to analyse electromagnetic waves in the practically continuous reflection in the area between 400–2500 nm wavelength.

Hyper-Spectral Imaging

Hyper-spectral imaging is a combination of digital recording and spectroscopy. It is a prospective non-invasive method that allows for differentiation of objects and certain states of objects based on the analysis of electromagnetic waves - the spectrum that is reflected from the object or, less frequently, when it is irradiated. A simple analogy can be used to better understand the functioning of hyper-spectral imaging. The human eye sees light in three bands, red, green and blue; spectral imaging can divide the visible part of the light spectrum into several dozen bands (so called spectral bands). Multispectral imaging involves up to ten spectral bands.

How does hyper-spectral imaging work?

Digital images are made of pixels. For each of these elements, the hyperspectral camera captures continuous information on the reflectance of light in every spectral band. Hyper-spectral sensors collect information from individual bands as a collection of images and then combine them into a three dimensional hyperspectral data cube. Every recorded light pixel thus contains a continuous light spectrum, the so called spectral signature. [Figure 1]

Slika 1: Enostavna multivariantna analiza spektralnih podpisov omogoča ločevanje plodov, listja (zdravega in bolnega ter mrtvega listja) in vejevja pri robidah

Poleg vidnega dela svetlobnega spektra (valovne dolžine od 390 do 700 nm) omogočajo določeni senzorji tudi zajem informacij v bližnjem infrardečem in kratkovalovnem infrardečem delu svetlobnega spektra (700 do 2500 nm).

Ob visoki spektralni ločljivosti (tj. število pasov) omogoča hiperspektralno snemanje tudi zajemanje podatkov z visoko prostorsko ločljivostjo (velikost svetlobnih celic). Ob premajhni prostorski ločljivosti pride do mešanja spektralnih podpisov iz več celic, kar je posebej nezaželeno na robovih predmetov oziroma na stičiščih več predmetov. V tem primeru govorimo o učinku roba, kjer zaradi mešanega spektralnega podpisa nastopijo težave s kvaliteto informacij in kasneje klasifikacijo. Hiperspektralno snemanje omogoča prostorske ločljivosti do 0,2 mm, kar zadostuje za ločevanje zrnc peska.

Za kaj je uporabno hiperspektralno snemanje?

S hiperspektralnim snemanjem zajamemo zelo natančne in podrobne informacije o proučevanih predmetih, kar nam omogoča zanesljivejše klasifikacije predmetov. Uporabnost te tehnologije je omejena samo s predmeti, ki nimajo spektralnega podpisa, torej takšnimi, ki ne odbijajo ali oddajajo svetlobe.

Ločimo lahko rastline (na primer ločevanje plevela ali invazivnih vrst od poljščin), posamezne organe rastlin (tudi določanje količine barvil, na primer klorofila) in določena stresna stanja rastlin; še pred izrazitimi vidnimi znaki (bolezni, prehrana, suša, poškodbe …). [Slika 2]

Slika 2: Določanje žarišč zlate trsne rumenice v vinogradu (zgoraj: bolne rastline so obarvane rdeče) in indeks Normalized difference vegetation indeks (NDVI, spodaj: višje vrednosti so obarvane rdeče)

Določimo lahko količino beljakovin, vlage, maščob in kosti v mesu. Prav tako lahko iščemo in ločujemo različne parazite, ki živijo v živalskem tkivu.

V določenih primerih lahko ločimo kemijsko oziroma mineralno sestavo predmetov, kar se s pridom uporablja v geologiji in tehnologiji živil. Metoda bi bila uporabna tudi za določanje mineralne sestave gline v arheoloških raziskavah.

Nenazadnje hiperspektralno snemanje uporabljajo tudi za analiziranje umetniških slik. Analiza spektralnih podpisov daje vpogled v način slikanja, uporabo metod in materialov ter iskanje napak in popravkov.