Materials and Methods

Study area

Chikkamagalur district is situated in the south­ western part of Karnataka state. District has large hilly regions that are forested with heavy rainfall. District is situated between 12°54'42" and 13°53'53" north latitude and between 75°04'46" and 76°21'50" east longitude (Fig. 1), bounded on the east by the Tumkur district, on the south by the Hassan district, on the west by the Western Ghats which separate it from the Dakshina Kannada district, on the north east by the Chitradurga district and on the north by the Shimoga district (Chikkamagalur District, 2018 - http://www.chickmagalur.nic.in/). Population in 2001 was 1.14 million with a population density of 158 persons per km2 and current population in the district is 1.13 million, and there has been a decrease of 0.28% during the last decade (Census

of India, 2011 - http ://censusindia.gov.in). Large parts of Chikmaga lur district are mountainous. The highest point in the district is Mullaiyanagiri at 1926 m amsl) (Karnataka State Gazetteer, Chikkamagalur District, 1981; http://gazateer.kar.nic.in). The principal rivers of the district are the twin streams - Tunga and Bhadra. Fig. 2 shows the location of major rivers in the district. This is the wettest district in the state having an annual average rainfall of 1772 mm. Mean Annual distribution of Rainfall in the district is as shown in Fig. 3.

Forests of Chikkamagalur district consists of evergreen and semi-evergreen climax forests and their degradation types and deciduous climax forests and their degradation types (Pascal, 1988). The evergreen and semi-evergreen climax forests and degradation type consists following categories: Dipterocarpus indicus - Humboldia brunonis -Poeciloneuron indicum type Dipterocarpus indicus - Diospyrus candol/eana -

Fig. 1: Study region - Chikkamagalur district, Karnataka State, India.

Fig. 2: Major Rivers of Chikkamagalur district

Diospyros oocarpa type, Dipterocarpus indicus - Persea macrantha type, Persea macrantha - Diospyros spp.- Holigarna spp. type, Diospyros spp.­ Dysoxylum malabaricum - Persea macrantha Kan forest type of low elevation (0-850m) Mesuaferrea - Palaquium ellipticum type, Palaquium ellipticum - Poeciloneuron indicum - Hopea canarensis type of medium elevation (800-1400 m) and Scheff/era spp. - Gordonia obtuse - Meliosma arnottiana type. The secondary or degraded type contains secondary

Evergreen, Semi-evergreen and moist deciduous forests. Deciduous climax forests consist of moist deciduous type - Lagerstoemia microcarpa -Tectona grandis - Dillenia pentagyne type and dry deciduous - Anogeissus /atifolia - Tectona grandis - Terminalia tomentosa type. The vegetation broadly falls into 4 types i). Dry deciduous hill type ii). Moist deciduous type, iii) the Evergreen type and iv). The Sholas and Grassland type (Fig. 4)

Fig. 3: Mean Annual rainfall distribution from 1901-2010 for Chikkamagalur district

Fig . 4: Vegetation distribution (classification as per Champion and Seth, 1968).

Materials

The satellite data of Landsat series Multispectral sensor, thematic mapper and IRS LISS Ill sensors for four-decade period (1976 to 2009) were acquired from Global Land Cover Facility (GLCF), (United States Geological Survey (USGS)-Earth Explorer, GLOVIS and Bhuvan website as listed in Table 1.

Method

Assessment of landscape dynamics involved (i) temporal analysis of land cover and land use using remote sensing data, (ii) quantification of natural forests, (iii) assessment of extent of forest fragmentation (due to encroachment and subsequent changes in land uses). The procedure followed to assess landscape dynamics is outlined in Figure 5. Spatio-temporal changes of land cover and land use (LULC) were studied using temporal remote sensing data with geospatial techniques. The remote sensing data obtained were geo-referenced, rectified and cropped pertaining to the study area. Geo-registration of remote sensing data (Landsat data) has been done using ground control points collected from the field using GPS and also from known points (such as road intersections, etc.) collected from geo-referenced topographic maps published by the Survey of India. Ground control points (GCP's) were collected from the Survey of India (SOI) topographic maps as well as from field using hand held pre-calibrated GPS. This helped in geometrically correcting the distorted remote sensing data. Landsat satellite 1976 data have a spatial resolution of 57.5 X 57.5 m (nominal resolution) were resampled to 28.5 m comparable to the 1991- 2009 data which are 28.5 X 28.5 m (nominal resolution).

Land cover analysis

Normalized Difference Vegetation Index (NOVI) was used to calculate the land cover types. The vegetation change analysis for multi-temporal data can be done using NOVI (Ramachandra et al., 2009; Ramachandra and Uttam Kumar, 2011). NOVI is calculated using Near Infrared and Red Bands data using equation (1),

NOVI= (NIR- IR )/(NIR + IR ) (1)

Table1: Details of the data used in the analysis.

Data and source	Resolution	Purpose
Landsat Series Multispectral sensor	57.5m	Land cover, Land use analysis and Fragmentation analysis
Landsat Series Thematic mapper
http://landsat.gsfc.nasa.gov	28.5m
IRS LISS III
http://nrsc.gov.in	23.5m
Survey of India (SOI) toposheets of 1:50000 and 1:250000 scales Generate boundary and Base layer maps
http://surveyofindia.gov.in		Generate Boundary and Base layer maps
Field visit data -captured using pre-calibrated GPS		Geo-correcting and generating validation dataset
Google earth and Bhuvan http://bhuvan.nrsc.gov.in		For digitizing various attribute data and as validation input

Fig. 5: Method used for Land use analysis.

NOVI values ranges from -1 to 1. The negative value indicates the non-vegetation and presence of built-up, water, sand etc. The increasing value from O indicates the presence of vegetation.

Land use analysis

Land use analysis was done for four-decade satellite data. The procedure involved (i) generation of False color composite (FCC) of the datasets using bands green, red and near infrared, mainly used in the identification of heterogeneous patches in the landscape, (ii) selection of training polygons (these correspond to heterogeneous patches in FCC) covering 15% of the study area and uniformly distributed over the entire study area, iii) loading these training polygons co-ordinates into pre­ calibrated GPS, iv) collection of the corresponding attribute data (land use types) for these polygons from the field. GPS helped in locating respective training polygons in the field, v) supplementing this information with Google Earth vi) 60% of the training data has been used for classification, while the balance is used for validation or accuracy assessment.The land use analysis was carried out with training data using supervised classification technique based on Gaussian Maximum Likelihood algorithm. The supervised classification approach preserves the basic land use characteristics through statistical classification techniques using a number of well-distributed training pixels. Gaussian Maximum Likelihood classifier (GMLC) is appropriate and efficient technique based on "ground truth" information for classifier learning. Supervised

training areas are located in regions of homogeneous cover type. All spectral classes in the scene are represented in the various subareas and then clustered independently to determine their identity. The following classes of land use were examined : built-up, water, cropland, open space or barren land, and forest. Such quantitative assessments, will lead to a deeper and more robust understanding of land-use changes for an appropriate policy intervention. GRASS G/S (Geographical Analysis Support System), a free and open source software having the robust support for processing both vector and raster files is used for the analysis. Accuracy assessment of classified data has been done through the computation of kappa (K) statistics and producer's and user's accuracies

Fragmentation analysis: Forest fragmentation analysis was performed to quantify the type of forest in the study area- patch, transitional, edge, perforated and interior based on the classified images of Chikkamagalur district. Forest fragmentation statistics and the total extent of forest (Pf) and its occurrence as adjacent pixels (Pff) is computed through fixed-area window (3x3) considering central pixel and its surrounding pixels (Ritters et al., 2000, Ramachandra and Uttam

Kumar, 2011, Ramachandra eta/., 2009; 2016). The result is stored at the location of the central pixel.

Thus, a pixel value in the derived map refers to between-pixel fragmentation around the corresponding forest location. Forest fragmentation category at pixel level is computed through Pf (the ratio of pixels that are forested to the total non-water pixels in the window) and Pff (the proportion of all adjacent (cardinal directions only) pixel pairs that include at least one forest pixel, for which both pixels are forested). Pff estimates the conditional probability that given a pixel of forest, its neighbor is also forest. Based on the knowledge of Pf and Pff, six fragmentation categories derived (Fig. 6) are (i) interior, when Pf= 1.0; (ii) Patch, when Pf< 0.4; (iii) transitional, when 0.4 < Pf< 0.6; (iv) edge, when Pf

> 0.6 and Pf - Pff> O; (v) perforated, when Pf> 0.6 and Pf - Pff< 0, and (vi) undetermined, when Pf>

0.6 and Pf= Pff.

Fig. 6: Forest Fragmentation using PF and PFF