Põllumajanduslikus kasutuses olnud maa metsastumise kaardistamine satelliidi Sentinel-2 talvistelt piltidelt
Laen...
Kuupäev
2024
Kättesaadavus
05.09.2024
Autorid
Ajakirja pealkiri
Ajakirja ISSN
Köite pealkiri
Kirjastaja
Eesti Maaülikool
Abstrakt
Eesti territooriumi kogupindalast moodustab statistilise metsainventuuri (SMI) valikuuringu põhjal tänapäeval 51 % metsamaa ning 27 % põllumaa. Tuginedes Hannes Palangi tehtud doktoritöö tulemustele on 1918 aastast alates põllumaa osakaal vähenenud ning metsade osakaal tõusnud. Mõneks põhjuseks on soode kuivendamine ja kasutuseta jäänud põllumaade metsastumise.
Bakalaureusetöö eesmärkideks oli 1) leida talvel lausalise lumikattega oludes pildistatud sateliidipilte töödeldes varasemalt põllumajanduslikus kasutuses olnud metsastunud alade kogupindala ning kirjeldada metsana klassifitseeritud kujundite paiknemist 23 x 20 km testalal Valgamaa, Viljandimaa ja Tartumaa kokkupuutealal ja 2) hinnata võimalike vale-positiivsetena klassifitseeritud metsa-kujundite ilmumist tulemustesse ning nende osakaalu metsana klassifitseeritud kujundite seasendisel põllumajandusmaal. Töös kasutati riikliku satelliidiandmete keskuse ESTHub lehe vahendusel vabavarana allalaetud Sentinel-2 skanneri MultiSpectral Instrument (MSI) kahte pilti pildistatuna 18.02.2021.a ja 10.03.2023.a. Lisaks satelliidipiltidele kasutati põllumaal kasvama hakanud metsalappide tuvastamiseks põllumajandusmaa vektorandmestikku Eesti Baaskaardilt, mis esindab 1980- ndate teise poole ja 1990-ndate alguse põllumajandusmaa seisu. Tänapäevase põllumaa situatsiooni arvestamiseks kasutati Eesti Topograafia Andmekogu (ETAK) ja Põllumajanduse Registrite ja Informatsiooni Ameti (PRIA) põllumaa kaarte. Andmetöötluseks kasutati kahte programmipaketti: Idrisi Taiga ja Mapinfo Pro. Satelliidipiltide pikslid klassifitseeriti eristusnivooga klassifitseerimisena kahte klassi: „mets“ ja „mitte-mets“. Leitud metsana klassifitseeritud pikslitest jäeti järeltöötlusel alles vaid need, mis paiknesid 1990-ndate aastate Baaskaardi põllumajandusmaal. Sateliidipildile omastest kujundite hägusatest servadest tulenevate klassifitseerimisvigade hindamiseks loodi põllumaa piirialale naabruses oleva metsaga ühe- ja kahepikslised servavööndid.
Töö tulemusena selgus, et endistest Baaskaardil esindatud põllumaadest, testala kogupindalaga 16 910 hektarit, on 2023.a satelliidipildi klassifitseerimistulemusena metsastunud 14 %. Põllumaa servadega on seotud 95 % metsastunud põllumaid, ühe pikslises servavööndis tuvastati 24% kõigist metsastunud lappidest.
Nowadays, 51 per cent of Estonian land area is covered with forest and 27 per cent is used as agricultural land according to National Forest Inventory. The relative share of forests and of agricultural land has been changed since 1918, since the very first years of the Republic of Estonia, the general tendency since that time has been a decrease of the agriculturally used land, while forested area has been increasing according to doctoral thesis by Hannes Palang. Some of the reasons are the draining of swamps as well as afforestation of formerly agriculturally used land. The aim of the thesis was to 1) to find the total area of forested patches on formerly agriculturally used land and to describe the spatial pattern of the respective patches. 2) The second study task was to isolate possible false-positive forest patches that appear in a narrow (mostly 1-pixel wide) boundary area of agricultural land with neighboring forests. The study area was a transitional area in a boundary region of Valga, Viljandi and Tartu counties covering an area of 23 by 20 kilometers. Two images by Sentinel-2 scanner MultiSpectral Instrument (MSI) were downloaded for free from National Satellite Data Centre ESTHub. The respective image dates were February 18.2021 and March 10.2023. The images were selected as substantially cloud-free accepting that the ground was covered with snow at image dates. Agriculturally used land vector polygons from Estonian Basemap were used depicting the situation of late 1980s and early 1990s in addition to satellite images. Agricultural land datasets from Estonian Topographic Database (ETAK) and from Agricultural Registers and Information Board (PRIA) were used to describe the current situation. Two program packages were used for data processing: Idrisi Taiga and Mapinfo Pro. Pixels in satellite images were classified into two classes: “forest” and “nonforest”. Finally, only those pixels of forest patches were retained that occurred within the area of agricultural land of Basemap data set. To avoid blurred edges coming from satellite image, one- and two-pixel edge strips were made to the edge of agricultural lands. As a result of the classification of satellite photo from 2023, 14 per cent of the formerly agriculturally used land was covered with forest cover from total area of 16 910 hectare. 95 per cent of the total forested area was related to the border of farmland. 24 per cent of possibly false-positive forest areas were found in the one-pixel zone at the edge of farmland.
Nowadays, 51 per cent of Estonian land area is covered with forest and 27 per cent is used as agricultural land according to National Forest Inventory. The relative share of forests and of agricultural land has been changed since 1918, since the very first years of the Republic of Estonia, the general tendency since that time has been a decrease of the agriculturally used land, while forested area has been increasing according to doctoral thesis by Hannes Palang. Some of the reasons are the draining of swamps as well as afforestation of formerly agriculturally used land. The aim of the thesis was to 1) to find the total area of forested patches on formerly agriculturally used land and to describe the spatial pattern of the respective patches. 2) The second study task was to isolate possible false-positive forest patches that appear in a narrow (mostly 1-pixel wide) boundary area of agricultural land with neighboring forests. The study area was a transitional area in a boundary region of Valga, Viljandi and Tartu counties covering an area of 23 by 20 kilometers. Two images by Sentinel-2 scanner MultiSpectral Instrument (MSI) were downloaded for free from National Satellite Data Centre ESTHub. The respective image dates were February 18.2021 and March 10.2023. The images were selected as substantially cloud-free accepting that the ground was covered with snow at image dates. Agriculturally used land vector polygons from Estonian Basemap were used depicting the situation of late 1980s and early 1990s in addition to satellite images. Agricultural land datasets from Estonian Topographic Database (ETAK) and from Agricultural Registers and Information Board (PRIA) were used to describe the current situation. Two program packages were used for data processing: Idrisi Taiga and Mapinfo Pro. Pixels in satellite images were classified into two classes: “forest” and “nonforest”. Finally, only those pixels of forest patches were retained that occurred within the area of agricultural land of Basemap data set. To avoid blurred edges coming from satellite image, one- and two-pixel edge strips were made to the edge of agricultural lands. As a result of the classification of satellite photo from 2023, 14 per cent of the formerly agriculturally used land was covered with forest cover from total area of 16 910 hectare. 95 per cent of the total forested area was related to the border of farmland. 24 per cent of possibly false-positive forest areas were found in the one-pixel zone at the edge of farmland.
Kirjeldus
Bakalaureusetöö
Metsanduse õppekaval
Märksõnad
bakalaureusetööd, Eesti Baaskaart, satelliidipildi klassifitseerimine, Sentinel-2 MultiSpectral Instrument