جستجوی مقالات مرتبط با کلیدواژه "طبقه بندی" در نشریات گروه "جنگلداری"

Earth's surface reflection is significantly affected by atmospheric conditions such as water vapor and particulate matter; therefore, atmospheric correction is needed to minimize these effects and convert digital number values to surface reflection. Therefore, as a remote sensing approach, atmospheric correction is required to minimize these effects and convert digital number (DN) values ​​to surface reflectance. The main objective of this research was to study of four atmospheric correction models, including (1) dark object subtraction (DOS), (2) fast line-of-sight atmospheric analysis of spectral hypercube (FLAASH), (3) the second simulation of the satellite signal in solar spectrum (6SV), (4) atmospheric topographic correction (ATCOR) model and their comparison with the OLI original image for estimation of the aboveground biomass (AGB) in the forests of Avardim of the Shafarood watershed, Guilan province.

In order to estimate of biomass, 246 plots (3600 m2) were established as systematic-random sampling pattern with 300m×300m dimensions and DBH, height of trees and shrubs were measured. A handheld GPS device (Garmin GPS MAP 64s with an accuracy of ±3 meters) was used for sampling and finding the sample pieces, and for this purpose, before starting the statistics, the latitude and longitude of the points (all sample pieces) were entered into the GPS device. And then, using the above device, the sample parts were determined in the field and the characteristics of the diameter at the chest, the height of the existing trees and shrubs (diameter more than 7.5 cm) were measured and then recorded in the relevant forms. The OLI sensor images of Landsat 8 satellite were extracted from the USGS global site. The selection of images was done according to the season, the amount of minimum cloud cover and also in the growing season close to the time of maximum greenness. These images are presented at L1T level and fully compatible with digital maps. In this research, 7 OLI sensor bands of Landsat 8 satellite related to pass/row number 166/34 have been used. Before the Landsat 8 satellite passes over Iran, Terra satellite prepares images with a time difference of about half an hour to local time. Due to the stability of the atmospheric conditions, it is possible to use MODIS information in step with Landsat 8. Also, in this research, three daily MODIS products were used for each of the Landsat 8 images with a spatial resolution of 500 meters, including: MOD04 (optical thickness of suspended particles in the range of 550 micrometers), MOD05 (water vapor) and MOD07 (total ozone). The DEM obtained from ASTER with a spatial resolution of 30 meters was obtained from the USGS global site. The DEM model was directly used in the SV6 atmospheric correction method. Also, DEM model was used for the atmospheric effect correction method of ATCOR to prepare the map of slope, direction, sky visibility.

The results showed that the atmospheric correction model based on 6SV radiative transfer code had a good performance in most of the plant indices obtained from the OLI sensor data of Landsat 8 satellite. The ARVI index obtained from the 6SV atmospheric correction model has the highest correlation analysis results with a correlation coefficient of 0.801. Also, in the case of using the FLAASH method, ARVI (0.779) and RVI (0.586) indices have the highest and lowest correlations, respectively. In the DOS or dark object atmospheric correction method, the highest and lowest correlations are related to GARI (0.762) and EVI (0.518) indices, respectively, and finally, in the ATCOR method, the highest and lowest correlations are respectively related to NDVI indices (732.0) and GNDVI (0.454). In general, in estimating forest biomass, 6SV atmospheric correction model showed the best performance with the lowest RMSE percentage (15.04%), followed by FLAASH, ATCOR and DOS models.

Estimating and monitoring the amount of biomass on land is necessary for climate change studies, carbon cycle production, food allocation and fuel accumulation, fire behavior studies, etc. in the ecosystem. Also, applying atmospheric corrections on the main bands of the images in the pre-processing process before classifying and extracting plant indices is necessary and unavoidable to remove the unwanted effects of the atmosphere and it improves the accuracy of the results. From the results obtained in the present study, it can be suggested that the 6SV atmospheric correction model, with the integration of water vapor and aerosol optical depth obtained from MODIS products, is more suitable for the estimation of terrestrial zinc based on remote sensing data, especially when using the data in are obtained in summer, when water vapor and temperature are both high and the forest canopy is in full development. Finally, it is suggested to use the 6SV atmospheric correction model to estimate of aboveground biomass based on remote sensing data.

The crisis of decline or drying of oak trees in the Zagros forests is one of the problems we have faced in recent years. The first step in managing this crisis is to map the affected areas and classify the severity of deforestation. The present study aimed to evaluate the Landsat OLI capability to map oak stands dieback in the Koohdasht city of Lorestan province. In addition to the main bands and fused bands with the 15-meter panchromatic band of the OLI sensor, suitable vegetation indices and the first components from PCA were also applied in the claasification. 150 square sample plots with dimensions of 30 × 30 meters were recorded to produce ground truth map. Data classification was done using minimum distance to mean, maximum likelihood and artificial neural networks classifiers in 5 classes of dieback and accuracy assessment was done using ground truth map. Because of the low separability of some classes, these classes were merged. Finally, classification with three and two classes (healthy, dieback) was performed. The highest overall accuracy of 53%, 75% and 91% and Kappa coefficient of 0.43, 0.64 and 0.71 was obtained using fused bands and artificial neural networks classifier for five, three and two dieback classes, respectively. The results demonstrated high performance of Landsat 8-OLI for mapping of healthy and oak dieback areas, but low to moderate functionality for identification of the intensity of oak decline in the study area.

The aim of this study was to investigate characteristics and amount of biomass chages of gap makers in beech forests of Sourdar-Anarestan Forestry Plan, Noor. For this purpose, an area of 100 hectares was selected and physiographic characteristics, area and age of the gaps, also the diameter and height of the gap maker trees were measured by a full inventory whithin the gaps. The results showed that there is not a significant difference between the gap maker aboveground biomass changes in the studied area with physiographic characteristics and the canopy area of the stand, but there is a significant difference between gap maker species, gap age and the type of disturbance (α=0.01). The mean rank of gap maker trees biomass in the wind created gaps was the lowest value, and in the forest harvesting created gaps was the highest rank. The result of this study have shown, most type of disturbances only remove a small amount of biomass on the aboveground biomass in forests. The effect of disturbance on regeneration seems small at first, but then it has a significant effect on it and forest continuity. Therefore, by better understanding the existing disturbances, the most important factors of which in the region were wind, longevity and exploitation, it is possible to help the continuation of forest production and sustainability.

Monitoring of land use changes in forest areas provides acceptable information for better planning and management on forest resources. In this study, to assess the changes of the forest cover in the central part of Boyer-Ahmad County in Kohgiluyeh and Boyer-Ahmad Province in southwest of Iran, satellite imageries of two different Landsat time series including Landsat 5, TM satellite data (30.6.1987), and Landsat 8, OLI-TIRS data (23.7.2013) were used. Supervised classification was carried out by applying the Maximum Likelihood Algorithm and LMM model (Matrix Multiplication pictures). Moreover, Logistic regression and Geomod model were used to provide validation map and prediction of forest cover changes in 2039, respectively. Land cover maps achieved from the study area showed that there is a reduction of about 8395 hectares in the forest area during the past 26 years (1987-2013). Moreover, it is predicted that the forest area will decrease around 9938 hectares during the coming 26 years. The results showed that the kappa coefficient obtained from the supervised classification on the satellite imageries in 1987 and 2013 were 0.89 and 0.88, respectively. Moreover, the results of Logistic regression with pseudo R2 and ROC were 0.24 and 0.73, respectively, which indicate that the obtained model is relatively adapted to the real changes and there is an appropriate ability for the model to estimate the forest changes in the last 26 years. Results of simulation of the forest cover changes in 2013 reveal that the Geomod model is a good tool for forecasting the forest cover changes. The accuracy and precision of the obtained forest cover maps is about 90 percent.

This study aims to determine the floristic composition and investigate the distribution of communities along an altitude gradient in oak forests of Taf in Khoramabad. The study area was classified into 100m elevation classes, and in each class two 500m2 plots were chosen randomly for measuring trees and shrubs layer. Also in each 500 m2 plot three 4m2 plots were distributed for measuring herbaceous layer composition. Changes in vegetation along an altitudinal gradient were studied using slope, altitude and some edaphic data by CCA, DCA and TWINSPAN ordination and classification methods. Based on the results, 116 species from 35 genera were observed along the altitude gradient. Asteraceae, Poaceae and Fabaceae families had the most species respectively. Therophytes, Hemicryptophytes and Cryptophytes were the most dominant life forms and Iran-Toranian and Iran-Torani-Mediterranean were the most chorotypes in this area.

Vegetation classification is an essential tool to describe, understand, predict and manage ecosystems. The aim of this study was to compare different types of hierarchical clustering. Three forest patches with similar slope and altitude gradients located on the southern slopes of Chahar Zebar forests, Kermanshah province, were selected. Vegetation sampling in each patch was conducted at 0, 25, 50, 100 and 150-meter distances along three transects that were 200 m apart. Cluster analysis was used for the classification of samples. Amongst the applied methods, Gower’s distance (or similarity) initially computes distances between pairs of variables over data sets and then merges those distances with the nearest neighbor, complete neighbor, average neighbor, and Ward’s method. The optimal number and quality of clusters were evaluated with silhouette criteria. In addition, the Cophenetic correlation coefficient was computed for evaluating the correlation between the dendrogram and the distance matrix. Results showed that two was the optimal number of clustering for oak stands. Moreover, the Cophenetic correlation coefficient between the distance matrix and the nearest neighbor and average method was higher than that returned between complete neighbor and Ward’s method. Based on silhouette criteria, the nearest neighbor and average methods were associated with higher cluster quality compared with two other methods. However, the mean value of the silhouette index was low for the second cluster of the nearest neighbor method. Considering the disadvantages of the nearest neighbor, the average method is suggested for clustering categorical data.

The forest type map is one of the most important thematic maps for forest ecosystem management. Forest mapping using field methods or aerial photos is labor-intensive and time consuming. In contrast, satellite data with its own characteristics like large and repetitive coverage, update and useful information in various wavelengths provides a good opportunity in this regard. This research was carried out with the aim of providing forest type map of central Zagros forests (Chahartagh forest reservoir), of Iran, using the Landsat 8 Operational Land Imager (OLI) data, in August 2016.Two ground-truth maps based on tree density and tree crown area were prepared by field surveying. Moreover, ancillary data such as tree species, location and crown area was taken. In order to increase the spatial resolution of multispectral bands, various image fusion techniques were applied. The best result obtained by the maximum likelihood algorithm with kappa coefficient and overall accuracy values of 57 and 63%, respectively. Due to high species diversity in this area the results showed that the OLI images have a moderate capability to produce forest type maps in Zagros forest.

Nowadays, detection of land cover changes and their effects on natural resources and the environment has been feasible through the technology of remote sensing and GIS. The present study aimed to assess the coverage change in Kiasar area over a period of 46 years (from 1966 to 2012) using remote sensing technology and satellite images. To this end, we prepared the land use maps for the years 1966, 1987 and 2012 using topographic maps, image of TM sensor and image of LISS III sensor, respectively. At first, the required pre-processing workflow comprising geometric and atmospheric corrections, and the creation of spectral and aspect ratio indices was performed on the bands of both sensors. After determination of current land uses in the study area, the training samples were recorded using GPS from each land uses and resolution of training samples was obtained using transformed divergence severability index. Classification was performed using maximum likelihood, parallel piped and minimum distance classifiers. The changes were compared after classification. The analysis revealed that over a period of 46-year the forest areas declined from 98.16 km2 (53.12% of land area in 1966) to 18.39 km2 (9.95% of land area in 2012). In other words, over this period of time 43.17% of the forest areas were replaced by uncovered land, agricultural lands, dry farm lands and rangelands. These changed areas increased by 14.15, 13.61, 10.05 and 5.37%, respectively. In conclusion, RS has a high capability in detection of the land cover changes.

Different physiographic conditions and soil properties have resulted in the establishment of different plant communities in central Zagros forests. This study aimed to differentiate these plant communities by means of detecting the species similarity and relations between understory plant vegetation, physiographic conditions and soil properties. Sampling was conducted using 21 modified multiscale Whittaker plots that were randomly distributed amongst the vegetation types. In each plot, species name and canopy cover as well as physiographic data were recorded for each individual tree, and soil specimens were sampled. The relationships between understory vegetation and environmental variables were investigated by means of two way indicator species analysis (TWINSPAN) classification method as well as two ordination approaches, including detrended correspondence analysis (DCA) and canonical correspondence analysis (CCA). The results of ANOVA, ordination, and classification in segregation of vegetation types showed that the distribution of vegetation types, species and also ecological species groups in central Zagros is affected by altitude, slope, geographical aspects, and the amount of soil organic C and N. Moreover, the results of vegetation classification segregated local vegetation into three ecological groups, including 12 indicator species. These results were in line with the ordination results. This study showed a high resemblance between the results of vegetation classification and ordination in accordance to the extent of study area (landscape scale) and random selection of sampling units at the physiognomic types and the ecotones.

The present study aimed to evaluate the capability of IRS-P6-LISS III imagery to map a forest canopy density by employing the traditional hard classification and the Artificial Neural Networks (ANNs) in Marivan city، Kurdistan province. Geometric correction procedure was performed with less than 1 pixel root mean square error (RMSE). Various vegetation indices and artificially bands generated by principal component analysis (PCA) were used in the classification procedure. A ground truth map was produced based on a randomized-systematic method with a grid size of 250×400 meters and 50×50 meters sample size strata. The suitable band combinations for classification were selected through the training area using the Transformed Divergence index. Supervised classification methods i. e.، parallelepiped، minimum distance to mean، maximum likelihood، and ANNs algorithms were applied to generate the canopy density map with 4 classes (very sparse، sparse، semi-dense and dense). The accuracy assessment of the generated canopy density maps was implemented using the ground truth map. Some classes were also merged because of the low spectral separation between these classes. Finally the classification was performed to produce the canopy density map with 3 classes (sparse، semi-dense and dense). The highest overall accuracy and the Kappa coefficient were achieved by maximum likelihood method with 78. 47 percent and 0. 66، respectively. Our results indicated the high capability of the IRS-P6 LISS III imagery compared to other satellite images، for example، Landsat and Aster data، which already tested in the previous work to map the canopy density in Zagros forests.

To investigate the capability of satellite imagery to map poplar plantations, IRS-P6-LISSIII and IRS-P6-LISSIV scenes from 2007 as well as Landsat TM scenes from 2009 and 2011 were analyzed over Sowme`eh Sara region in Guilan province. The IRS-P6-LISSIII and IRS-P6-LISSIV scenes were geometrically corrected with root mean square error (RMSE) of ca. 11m and 60cm, respectively. Informative auxiliary bands ratios were calculated. In addition, the scenes were spatially enhanced by resolution merge of panchromatic and multispectral bands. Principal component analysis (PCA) and inverse PCA were also conducted for the spatially-fused data. Ground reference data was samples via field survey and included two classes of “maximum 5-year medium density canopy” and “dense canopy >5-years of age” poplar plantations. The ssatellite images were classified by crisp (supervised) and soft (fuzzy) classifiers using three different strategies of 1) classification into three classes including poplar plantation with “maximum 5-year medium density canopy”, “dense canopy >5-years of age” and others, 2) classification into eight classes including the two above-mentioned classes, rice, alder, conifer, sea, urban area and others, in which all non-poplar classes were aggregated into one class after the classification, and 3) independent classification of three sub-regions on the images. Results of accuracy assessment for the first and second strategy showed very low overall accuracy and kappa coefficient, whereas the third strategy showed a higher performance. This way, the highest rates of overall accuracy and kappa coefficient of the TM scene 2009 in the first sub-region were reported to be %80 and 0.58, respectively.

The aim of this study was to investigate the capability of IRS-P6-LISS IV data for forest density mapping in the Pistachio forests. So LISS IV images (5 m)، dated 2007 from Khaje Kalat (500ha) in Khorasan Razavi were analyzed. The geometric correction of images was implemented using nine control points extracted from an orthorectified image of the study area. The RMSE was less than one pixel. In addition to original bands، different synthetic bands from principal component analysis and transformation methods were created and used. The suitable bands set were selected by training areas and divergence indices. In order to assess the accuracy of classification results، a ground truth map covering 7% of the total area was prepared through fieldwork using 34 sample areas and canopy percent was estimated. Satellite data were classified by supervised classification methods including minimum distance to mean (MD)، maximum likelihood (ML) and fuzzy. There were spectral interferences between medium density classes (5-10%، 10-15% and 15-20%). Therefore these classes were merged. In hard supervised classification method، the highest overall accuracy and kappa coefficient، 67% and 0. 40، respectively، were obtained by ML classifier with three classes (0-5%، 5-20% and > 20%). Using mode filter with a 7×7 pixel increased the accuracy up to 3%. The results of Fuzzy algorithm showed higher accuracy and kappa coefficient، 70% and 0. 44، respectively. In both methods، second density class (5-20%) represented highest kappa coefficient. It could be concluded that the result of classification was not desirable regarding to low kappa، even if reaching to pretty good overall accuracy. To obtain a better result، it is suggested to use higher spectral resolution data and preparing fieldwork in smaller sample area and determining canopy percentage quantitatively.

This research focused on determining the possibility of separating different percents of Beech and Hornbeam species based on two methods of spectral reflectance assessment and classification using landsat imagery in a northern forest of Iran. To do this, Landsat7 ETM+ imagery of the study area for 24 June, 2000 (middle of germination season) were provided. The 100×100 m (1 ha) sample plots with 90m distances from each other over 3 areas containing Beach and Hornbeam mixtures were transmitted from forest type map on a 1:25000 scale. Four groups were determined after field inventory, I) pure Beech, II) 80% Beech and 20% Hornbeam, III) 70% Beech and 30% Hornbeam and IV) 60% Beech and 40% Hornbeam. To avoid road reflectance interference with tree reflectance, sample plots were selected with 60 m distance from roads. Followed by digitizing the roads and plots, the images were geometrically corrected at the ortho level using 13 ground control points. Then, the digitized roads and sample plots were overlaid on 6 bands (1, 2, 3, 4, 5 and 7) of the ETM+ imagery and their pixel data were extracted. In the next step, the obtained data was statistically analyzed. The differences among the selected groups were compared via t-test method. The results showed the capability of ETM+ imagery in separating the pure Beech from the mixture of 70% Beech and 30% Hornbeam and 60% Beech and 40% Hornbeam. Then, the images were classified using a Maximum Likelihood Algorithm. Based on the classification results, the maximum likelihood were measured with 63% total accuracy and 44% kappa coefficient. The results of comparing four above mentioned groups using maximum likelihood classification method were in accordance with the results of applying their pixel values.

The purpose of this research is the possibility to recognize of trees canopy die back using Quickbird satellite images and mapping of its distribution in district one from Shastkolate forest, North-west of Iran. After geometric quality and radiometric evaluation of data, geometric correction of panchromatic image was carried out with 45 ground control points and RMSE of 0.65 at X axis direction and 0.78 at Y axis direction. Moreover, the multi-spectral images were registered with georeferenced panchromatic image with 60 ground control points and RMSE of 0.19 at X axis direction and 0.25 at Y axis direction. Using ratioing, principal component analysis and creation of suitable vegetation indices, some artificial bands were created and used as suitable bands for image analysis. In order to prepare the training area and to evaluate classification accuracy, sample ground truth were provided by recording the died back trees using DGPS on a 500 m×100 m systematic network and 360 sample plots with 1000 m² area. After selection of training area and suitable bands collection, data were classified with supervised method by using maximum likelihood and density slicing. Results of the classification accuracy evaluation on 4 main bands and also 7 selective bands by maximum likelihood algorithm and vegetation indexes by density slicing algorithm showed that overall accuracy amount and Kappa coefficient for 2 forest classes and died back for 4 main bands and 7 best selected bands by maximum likelihood algorithm were 77%, 0.56, 83% and 0.685, respectively. In addition, overall accuracy amount and Kappa coefficient for density slicing of NDVI and TNDVI vegetation indexes were 51%, 0.16, 56% and 0.19, respectively. Results showed that recognition of died back trees using Quickbird satellite image was not completely possible due to reflection of shrub and under storey plants, adjacent trees crowns and low ratio of reflection of dried branches in compare to rest green crown, which are registered as digital value of pixels.

High-resolution imagery such as IKONOS satellite can produce useful information for many resource management applications. We investigated the capability of Pan and multi spectral IKONOS imagery of one region of Tehran dated from 2005 in urban vegetation management and mapping of green cover. After data quality of images for radiometric noises, the images were geometrically corrected by polynomial method using GCPS and DEM with less than 0.2 pixel RMSe. Fusion of multi spectral and panchromatic images was done using PCA method. Some suitable vegetation indices including NDVI, VI, GNDVI and TNDVI were generated by rationing transformations on the main bands. Based on the field checking of the study area, four classes of urban green cover types including non-forested area, broad-leaved forest, needle-leaved forest, and grass were considered. For each class, some points were sampled for training (20%) and accuracy assessment (80%) of classification using GPS and aerial photos. The best band selection was done using divergence index. The classification was done by maximum likelihood algorithm on the four best bands. The accuracy assessment of classification results was accomplished using 80 percent of rest sample points. The results of accuracy assessment showed that overall accuracy and kappa coefficient were 87% and 0.80, respectively. According to the results, the non-forested area, deciduous, coniferous vegetations and grasses covered 59%, 17%, 23% and 1% of the study area, respectively. Results of this research exposed that IKONOS imagery has capability for mapping the urban vegetation covers and can produce useful information for managing urban vegetation covers in different time periods.

Classifying age classes in a large area using remotely sensed data has considerable significance for forest sustainable management. In this research, Landsat ETM+ data from Loveh forest, dating July 2002, were analyzed to investigate the potential of this sensor for age class mapping. We applied a systematic cluster sampling method to collect field data. We used 99 plots so that contained 32 plot. In stands with 25-45 years, 33 plots in stand with 5-25 years and 34 plots in stands with >45 years. The quality of the image was first evaluated for radiometric noises. Separability of three age classes 5-25, 25-45 and >45 years, using a supervised classification and four algorithm of maximum likelihood, minimum distance, parallel piped and linear discriminate analysis (Fisher). The results showed that maximum likelihood in three and two age classes with overall accuracy and kappa coefficient were (79% and 94%) and (0.68 and 0.86), respectively. Signature separability, producer and user accuracies showed the highest spectral similarity between 5-25 and >45 age classes. By merging the two classes, the overall accuracy and kappa coefficient became equal to 94% and 0.86, respectively. These results demonstrate that the reflectance values recorded by ETM+ sensor are related to forest stands. This information could also be used to estimate forest biomass and carbon content, identify locations within the stands that might require treatment and plan other management activities.

In order to investigate the possibility of beech forest type (Fagetum) mapping using Landsat ETM+, its data from Chelir district (780 ha) in Khyrood forest, Caspian forests, Iran were analyzed. Geometric registration was applied using 14 ground control points based on digital topographic maps at 1:25000 scale. The RMS error obtained was less than half of an ETM+ pixel. In order to estimate the accuracy of the classified satellite images, a ground truth map covering 42% of the total area, was qualitatively prepared as strips after field inspection. Image classification was performed using original and synthetic bands (rationing, principal component analysis, tasseled cap transformation and fusion) for following four beech forest types: pure beech, dominant beech, mixed beech and non-beech types. Classification was performed using maximum likelihood, minimum distance to mean and KNN classifiers. The highest overall accuracy (35%) was obtained using KNN classifier. The main reason for low overall accuracy can be related to the kind of related classes and also spectral similarity between pure and dominant beech classes. Therefore, these two classes were merged and classification was done again. The highest overall accuracy, considering three classes increased the classification accuracy up to 51%. The results showed that the spectral data of ETM+ do not have a high potential for beech forest type mapping in heterogeneous and uneven-aged Hyrcanian forests, Because the type of considering classes in such a classification is based on one specie (here beech) whereas the abundance of spectral reflectance of other species is neglected.

In order to investigate the capability of SPOT5-HRG data for forest density mapping in Caspian forests, the data of this sensor with 5 and 10 spatial resolutions dated 2002 were analyzed. The study area with 10000 ha is located in south western of Amlash city in Guilan province. In addition to original bands, some synthetic bands using ratio, fusion and transformation methods were created and used. In order to accuracy assessment of classification results, a ground truth map covering 26% of total area was prepared based on seven aerial photos (1:40000) dated 2001. The aerial photos were orthorectified and mosaiced. A total of 2520 circle sample plots with one ha area were selected on the digital orthophotomosaic. Canopy closure percent of each plot was interpreted using a 45 dot grid. Satellite data were classified by supervised classification methods including minimum distance to mean (MD) and maximum likelihood (ML). The highest overall accuracy and kappa coefficient equal to 74% and 0.33 were obtained by maximum likelihood classifier with four classes (1-10%, 10-50%, 50-100% and non-forest). Third density class (50-100%) represented highest producer and user accuracy, 95% and 82%, respectively. Lower producer and user accuracy were related to first density class 11% and 32%, respectively. It could be concluded that due to low kappa coefficient (0.33), even if reaching to pretty good overall accuracy (74%), the result of classification was not desirable. To obtain a better result, it is suggested to test other classification methods like object-based. Using higher spectral resolution data are also offered.

In order to determine the relationship of soil characteristics and topographic conditions with tree and shrub species, three compartments of district 16 of Shafaroud forest, Guilan, Iran were sampled by sixty 0.1 ha plots. In each plot, diameter of all trees and shrubs (DBH>10 cm), altitude, slope and aspect were recorded and soil samples were taken at 0-20 cm depth. Total of basal area for every species in each plot was calculated. Using TWINSPAN (Two-Way Indicator Species Analysis), the forest was classified in five types, and then 6 soil profiles were excavated in them. The relationship of forest types with soil characteristics and topographic factors were analyzed by CCA (Canonical Correspondence Analysis), ANOVA (Analysis of variance) and Tukey's method. Based on the statistical methods, elevation, slope, soil pH and Ca showed significant relationship with forest types. Alnus- Juglans and mixed forest types had a significant relationship with elevation, soil pH and Ca. High correlation in Fagus- Tilia and Fagus- Fraxinus types with slope was recognized.

This study was carried out in order to investigate the capability of digital data of ETM+ sensor in separation of forest types in Gazoo district of Lafoor area in Savadkooh. The bands were controlled according to radiometric and geometric errors, separately. Band 1, was omitted because of the existence of radiometric error and its less importance in vegetation cover study. Geometric correction was performed by 21 ground control points with DEM, up to ortho rectification level with precision of less than half pixel (0.3 pixel). The supervised classification was performed by using basic and synthetic bands to 6 classes, (pure beech type, mixed beech type, mixed hornbeam, road and non covered area, persimmon, mixed broad leaf). Ground truth map prepared through sampling in 24% of whole area. The highest overall accuracy was belong to maximum likelihood classification for 6 classes which was 38.29% and Kappa coefficient was 27.7%. Six vegetation types were merged because of radiometric mixing, therefore classification with 5 classes was performed again. Accuracy assessment of classification results indicated that the highest overall accuracy and Kappa coefficient were 53.22% and 34.71%, respectively. Results showed that the ML classification increases %15 of overall accuracy and %7 in Kappa coefficient. Overall, using ETM+ data is not so appropriate in the studies which the map type is considered as a base map with maximum number of existing type in the area. In order to increase the classification accuracy, using of other classification methods like object-base method and the other information and multitemporal data is suggestible.