| 1. |
- Allard, Anna, et al.
(författare)
-
Fångst av vegetationsdata och Natura 2000-habitat i fjällen genom flygbildstolkning i IRF med punktgittermetodik
- 2007
-
Rapport (övrigt vetenskapligt/konstnärligt)abstract
- I denna rapport visas att flygbildstolkning med punktgittermetodik har stor potential för att ta fram tillförlitliga och kostnadseffektiva data på tillstånd och förändringar i fjällens vegetation. Projektet är finansierat av Naturvårdsverkets svenska miljöövervakning.Inom denna studie har följande delmål nåtts1) Utprovning av försöksdesign har gjorts med antal och fördelning av punkter, samt design och val av tolkningsvariabler.2) Ett operativt tolkningssystem för punktgittertolkning har tagits fram.3) Tolkningsprecisionen för olika variabler och Natura 2000-habitat är utvärderade mot fältdata.4) Metodens kvaliteter är utvärderade.5) Beräkning av tidsåtgång och kostnadsberäkningar är gjorda och ställda mot en polygonbaserad metod för översiktlig tolkning av NILS 5x5 km ruta.Förutom dessa delmål har även säkerheten ytterligare undersökts via ett personvariationstest av tolkade data.Resultaten från de tolkade fältpunkterna visar att överensstämmelsen med fältinventeringen är mycket god. Ett gott resultat fanns från testet med personvariation, med några få problemområden. Metoden har några begränsningar. För flygbildstolkning i punktgitter med den metodik som föreslås här förutsätts att det finns välutbildade tolkare. För en van vegetationstolkare krävs det i storleksordning en tilläggsutbildning på minst 1 månad i tolkning av fjällvegetation, inklusive fältbesök. En exempelsamling av bilder och ett antal nycklar för tolkning bör tas fram.Punktgittermetoden är operativt användbar, vissa kompletterande fältstudier behövs. Metoden har följande generella kvaliteter.Metoden är enkel att implementera.Punktgittermetoden är en kostnadseffektiv metod för att fånga landskapsdata. Den är snabbare än polygontolkning.Datafångst från en 5 x 5 km ruta kan göras på 1-3 dagar, beroende på antal punkter i gittret.Det är enkelt att bearbeta och analysera punktgitterdata.Precisionen i arealskattningarna kommer att kunna enkelt beräknas.Tolkningen i punktgitter kan verifieras genom jämförelse med NILS fältdata.Metoden är lämplig för förändringsstudier. Exempelvis kan den utgöra ett snabbt och effektivt sätt att statistiskt uppdatera information om area av vegetationstyper från vegetationskartorna över de svenska fjällen.Metoden kan fånga upp ovanligare naturtyper i 5 x 5 km ytan förutsatt att många punkter tolkas.Metoden kan användas för urval av objekt för riktade fältinventeringar till ovanligare habitat.Resultaten från metoden kan utgöra viktiga data för annan forskning, exempelvis som träningsdata till satellitbildsklassificeringar.
|
|
| 2. |
- Gilichinsky, Michael, et al.
(författare)
-
Application of national forest inventory for remote sensing classification of ground lichen in nothern Sweden
- 2010
-
Ingår i: ISPRS Archives. - : International Society for Photogrammetry and Remote Sensing. ; 38-4-8, s. 146-152
-
Konferensbidrag (refereegranskat)abstract
- Lichen is a major forage resource for reindeer and may constitute up to 80% of a reindeer's winter diet. The reindeer grazing area in Sweden covers almost half of the country, with reindeer using mountainous areas in the summer and forested areas in the winter. Knowledge about the spatial distribution of ground lichens is important for both practical and sustainable decisionmaking purposes. Since the early 1980s, remote sensing research of lichen cover in northern environments has focused on reindeer grazing issues. The objective of the present study was to use lichen information from the Swedish Forest Inventory (NFI) for classification of satellite data into ground lichen classes. The classification procedure was focused on using of NFI plots as training sets for supervised classification of the ground lichen cover in purpose to classify areas with different lichen coverage. The present research has shown the advantage of use forest inventory plot data by assessment of three methods: mahalanobis distance (MD) classification, maximum likelihood (ML) classification and spectral mixture analysis (SMA). The results of this study demonstrate high classification accuracy of SPOT imagery in distinction between lichenabundant and lichen-poor areas by mahalanobis distance classifier (overall accuracy 84.3%, kappa=0.68). The highest classification accuracy for Landsat scene was achieved by maximumlikelihood classification (overall accuracy 76.8%, kappa=0.53). The continuation research on more detailed fragmentation of lichen cover into fractions is proposed.
|
|
| 3. |
- Gilichinsky, Michael, et al.
(författare)
-
Mapping ground lichens using forest inventory and optical satellite data
- 2011
-
Ingår i: International Journal of Remote Sensing. - : Informa UK Limited. - 0143-1161 .- 1366-5901. ; 32:2, s. 455-472
-
Tidskriftsartikel (refereegranskat)abstract
- Lichen is a major forage resource for reindeer and may constitute up to 80% of areindeer’s winter diet. The reindeer grazing area in Sweden covers almost half of thecountry, with reindeer using mountainous areas in the summer and forested areas inthe winter. Knowledge about the spatial distribution of ground lichens is importantfor both practical and decision-making purposes. Since the early 1980s, remotesensing research of lichen cover in northern environments has focused on reindeergrazing issues. The objective of this study was to use lichen information collected inthe Swedish National Forest Inventory (NFI) as training data to classify opticalsatellite images into ground lichen cover classes. The study site was located within thereindeer husbandry area in northern Sweden and consisted of the common areabetween two contiguous Satellite Pour l’Observation de la Terre (SPOT)-5 scenesand one Landsat-7 Enhanced Thematic Mapper Plus (ETMþ) scene. Three classificationmethods were tested: Mahalanobis distance, maximum likelihood andspectral mixture analysis. Post-classification calibration was applied using a membershipprobability threshold in order to match the NFI-measured proportions oflichen coverage classes. The classification results were assessed using an independentlycollected field dataset (229 validation areas). The results demonstrated highclassification accuracy of SPOT imagery for the classification of lichen-abundantand lichen-poor areas when using theMahalanobis distance classifier (overall accuracy84.3%, kappa ¼ 0.68). The highest classification accuracy for Landsat wasachieved using a maximum likelihood classification (overall accuracy 76.8%, kappa¼ 0.53). These results provided an initial indication of the utility of NFI data astraining data in the process of mapping lichen classes over large areas.
|
|
| 4. |
- Nilsson, Mats, et al.
(författare)
-
Estimating annual cuttings using multi-temporal satellite data and field data from the Swedish NFI
- 2009
-
Ingår i: International Journal of Remote Sensing. - : Informa UK Limited. - 0143-1161 .- 1366-5901. ; 30, s. 5109-5116
-
Tidskriftsartikel (refereegranskat)abstract
- Many countries have ongoing national forest inventories (NFIs) that provide reliable information on current forest conditions and changes in the forest landscape. These inventories are often based on data collected using field inventory procedures and the results are presented in terms of forest statistics for different geographical areas. The Swedish NFI has decided to combine their field data with optical satellite data by using post-stratification to obtain improved and unbiased estimates of forest variables. The method has been shown to reduce the sampling error (standard error) by 10-35% for variables such as stem volume and forest area. The objective of this study is to investigate the effect on sampling error for the estimated annual clear-felled area when the NFI plots are post-stratified by cuttings mapped from multi-temporal satellite images. Clear-felled areas mapped by the Swedish Forest Agency using image pairs (SPOT and Landsat) from the years 2001/2002, 2002/2003, 2003/2004, and 2004/2005 were used to post-stratify the NFI plots. The study area covers approximately a 1.3 Mha forest land area in Coastal Vasterbotten. It was found that the sampling error (standard error) for the annually clear-felled area was reduced by 31% using post-stratification compared to use of field data alone.
|
|
| 5. |
- Reese, Heather, et al.
(författare)
-
Comparison of Resourcesat-1 AWiFS and SPOT-5 data over managed boreal forest stands
- 2009
-
Ingår i: International Journal of Remote Sensing. - : Informa UK Limited. - 0143-1161 .- 1366-5901. ; 30, s. 4957-4978
-
Tidskriftsartikel (refereegranskat)abstract
- In this study, the utility of Advanced Wide Field Sensor (AWiFS) data in relation to stem volume estimation for managed boreal forest stands was investigated. Multiple linear regressions were used to predict stem volume (m(3) ha(-1)) with standwise mean spectral values as the independent variables. For comparison, two SPOT-5 images were used, one with nearly simultaneous acquisition. The adjusted coefficient of determination (R(adj)(2)) using AWiFS data to predict stem volume was 0.573 (SE = 56.9%) while SPOT had an R(adj)(2) of 0.598 (SE = 55.2%). All bands were negatively correlated, with the shortwave infrared (SWIR) band having the single strongest correlation with stem volume. When stem volume was predicted based on stand size, AWiFS and SPOT produced R(adj)(2) values of 0.310 and 0.293, respectively, for stands less than 2 ha in size. Predictive ability increased with stand size, with the highest R(adj)(2) at 20 ha (R(adj)(2) = 0.677 AWiFS, R(adj)(2) = 0.692 SPOT). For stands of 20 ha and larger, the correlation between stem volume and near-infrared (NIR) reflectance increased while decreasing for the visible bands. The explanation behind the trends observed may be due to the management practices in the area. The best two-band predictor of stem volume was the NIR and red band combination for AWiFS, and the NIR and SWIR bands for SPOT. Discriminant analysis of basic forest types showed similar results for AWiFS (65.6% correct) and SPOT (66.4% correct).
|
|
