Earth Belongs to all species

Loss of tree cover and diversity in private plantations

Assessment of Spatio-Temporal changes of Kodagu landscape using Geo-Spatial data

Introduction: Forests, agro-forests and other tree based land-uses play a vital role in regulating the climate and biodiversity conservation despite providing livelihood benefits to millions of people living in and around the forests and in rural areas. The forests of the Western Ghats region of peninsular India have undergone significant transformations over the past century. The nature, extent and causes of these transformations have been due to deforestation, over grazing, forest fire, rapid urbanization, encroachment for agriculture, etc. Hence the forest management requires an understanding of the spatial and temporal patterns using remote sensing data and Geographical Information system (GIS) techniques. Recently many researchers have used the remote sensing data to assess the forest cover and land use change associated with human activities. The Western Ghats of India is a biodiversity hotspot and many areas in Kodagu region have been recently declared as UNESCO sites. Kodagu region of the Western Ghats has mostly evergreen forests and is known for rich flora and fauna. Various studies have documented ecology and biodiversity of forest, agro-ecosystem and sacred groves of Kodagu. The landscape of Kodagu forms as an important catchment of the river Cavery which is a life line of peninsular India. As has been emphasized by us earlier, the river Cauvery sustains over 80 million people and about 600 major industries across south India. From the past decade, the landscape is undergoing rapid changes with respect to canopy densities and diversity due to changes in the production systems, due to urbanization and demographic pressures. These changes will have significant influence on biodiversity, hydrology, carbon sequestration potential and other processes that come together to affect over all climate and biosphere. Therefore, protection of Kodagu landscape is vital in order to ensure the food, water and economic stability of the region. This is why landscapes are increasingly considered to be appropriate levels to negotiate land use options. Within forest-agroforest landscape mosaics, various land use options such as agriculture, biodiversity, resource extraction etc. are competing for the space. Under these situations, negotiation and conflict mediation between stakeholders and their interests are therefore increasingly needed and as a result landscape level governance involving the representatives from government, NGO’s, corporate and planters associations is considered as an important approach in the recent years. IMFN also has envisaged landscape level governance as a key element in natural resource management of Model Forest sites. This requires critical examination of the landscape dynamics at various levels and scales based on the geospatial data in addition to biophysical and socio-economic aspects. Though the French Institute, Pondichery has geospatial data for Kodagu but it is up to 2007 and is based on the coarse resolution satellite data with scale of 1:250000. Therefore, the outcome of this project would provide the information to KMFT on what changes are occurring at the landscape level, the rates at which the changes are occurring and which habitats/land-use types are more prone for changes (for e.g. low land paddy cultivation is being changed into zinger, oil palm, and for construction). This would enable MF to focus its efforts on such habitats/areas within the landscape, strengthens the institutional capacity and thus could play a crucial role as a facilitator of landscape governance on the ground involving multi-stake holder. Furthermore, policy makers can be made aware of the extent of degradation of the landscape and the need for urgent measures to stabilize and restore the landscape.

Study area: The study area is extending from 110 56' North to120 52’ North and 750 22’ East to760 12’ East, with a total area of 4,102 Km2 and elevation varying from 900m to 1715 m (Fig 1). Mean temperature range from 20°C in the south to 24°C in the north with an average rainfall of 3000–4000 mm. It has a steep West to East climatic gradients especially, for temperature and rainfall from the edge of the ghats. The western slopes of the mountains experience heavy annual rainfall (with 80% during the southwest monsoon from June to September), while the Eastern slopes are drier; rainfall also decreases from south to north. The district is divided into the two forest divisions, namely Madikeri and Virajpet which include three wildlife sanctuaries (viz., Bramhagiri, Talakaveri, and Pushpagiri WLSs) and one national park (i.e., Nagarahole National Park). Two divisions of Coorg district are further divided into 14 ranges, 43 section and 79 beats. As per the census 2001 and 2011, the population is 548561 and 554762 respectively. The economy of district depends mainly on agriculture, plantations and forestry in addition to tourism. Primarily on coffee production and other plantation crops, characteristically and historically, paddy fields, ginger crops and meadows are found on the valley floors, with agro forestry in the surrounding hills. The coffee agro-forestry systems of Kodagu are one of the biologically diverse production systems in the world.

Fig.1 Location map of Kodagu district in Central Western Ghats

Soils are lateritic to red loamy, which have a mature profile and main rock formation belongs to the most ancient Archaean system with rock composed of peninsular gneiss, gneissic granites and gneiss.

2. Materials and methods:

2.1 Materials:

1. Survey of India Topographical maps (1:50,000 scale)

2. Ortho-rectified satellite data of the year 2005 and 2015 (IRS P6 LISS-III with 23.5 m resolution) were procured from National Remote Sensing Centre (NRSC), Hyderabad

3. Software’s used: Quantum GIS open source software.

4. Ancillary data

Topographical maps : Survey of India is the nodal agency that makes and manages maps in India. They produce topological maps from time to time and is made available for research use. However, these are available in hardcopy format. We procured 14 toposheets (Nos., 48PD43V6, 10, 11, 13, 14, 15, 16, 57DD43W2, 3, 48P7, 12, 49M/13, 57 D/4, 58A/1) with 1:50000 scale for the whole landscape of Kodagu. These were procured to obtain spatial information regarding administrative boundaries, topological boundary and other ancillary information that are controlled by Survey of India.

Satellite data sets: Data from multispectral satellite images from Indian Remote Sensing Satellite- IRS P6, LISS III sensor were procured from National Remote Sensing Centre (NRSC) Hyderabad India. These images were selected keeping in mind the area of the study landscape, the resolution and availability of four bands. The details of images procured for the study area are described in Table 1. To accommodate the study area completely, a total of two images per assessment year were used. Care was taken during the selection of imagery on visual quality, cloud cover and shadow. Images of both assessment years were selected around the same time of the year for consistency in signatures for classification. Use of multi-season images was not possible in this study due to the unavailability of good quality images during other time period of the year due to high cloud cover.

Assessment year

Path

Row

Shift

Date of pass

2005

98

64

50 % shift

10-01-2005

2005

99

64

0 % shift

15-01-2005

2015

98

64

50 % shift

23-02-2015

2015

99

65

0 % shift

04-02-2015

Details of satellite images used for the analysis

Imagery from IRS P6 LISS III has a resolution of 5.8 meters at nadir and captures data in four spectral bands viz., 0.52 to 0.59 microns (Green) 0.62 to 0.68 microns (Red) 0.76 to 0.86 microns (NIR) and 1.55 to 1.70 microns (SWIR). It has a swath of 141 Km. Additionally, Normalized Difference Vegetation Index (NDVI) was calculated within QGIS 2.2 using NIR and Red bands of the satellite images. For the classification satellite image we used Shortwave Infrared (SWIR), Near Infrared (NIR), Red and Blue band along with derived NDVI as band combinations.

Software: All analysis and data manipulations of satellite image were performed using QGIS 2.2 along with freely available open source plugins. Pre-processing including atmospheric correction and ortho-rectification was not necessary as pre-processed satellite data were procured from NRSC. However, a slight shift in geo-referencing was observed during analysis between two scenes. This image of 2005 was co-registered with the 2015 image using geo-referencing tool in QGIS 2.2.Positional accuracy of co-registering were ensured by keeping the root mean square error less than 0.5 pixels. In this case the RMSE was 7.6 meters with is much less than 11.75 m which is 0.5 times the pixel size.

Ancillary data: Additionally, various types of ancillary data including previous reports, working plan from the forest department and administrative maps already available with us were used as source of information at various stage of analysis of data. Example: to decide the landuse/landcover class and to create a signature library during classification.

2.2 Methods

2.2.1. GIS data base creation : The base map was prepared using hardcopy maps from survey of India toposheets of the study area. The methodology adopted for preparation of base map and different thematic layers is depicted in the following flowchart (Fig. 2). Then individual top sheets were geo-referenced, mosaicked and sub-setting to the study area. The spatial details like the district and taluk boundaries, forest boundaries, drainage/rivers network and road network were digitized with QGIS 2.2. These captured spatial layers were captured as different thematic layers. The thematic layers thus captured include administrative boundaries like district, and taluk. Forest department’s administrative boundaries like forest division, range boundary and section boundary (for only the territorial forest of the district). Similarly, protected area classification and their boundaries, transportation network and drainage network layers were generated.

2.2.2 Satellite data processing and analysis: For the classification into different LULC classes semi-automatic classifier an open source plug-in for QGIS was employed (Congedo, 2014). This plug-in work similar to that of supervised classification in other software’s. The actual process of classification was started by acquisition of image in two time frames, one in 2005 and the other in 2015. After initial pre-processing the images were visually interpreted to understand the landscape and chose of land use land cover classes. The landscape was first segmented into different regions based on prior knowledge of land-use type. The forest areas were masked using the inputs for toposheets when classifying the central, agriculture-coffee agroforestry dominated areas. Further, the agriculture-coffee agroforestry dominated central area were segmented into wet zone and dry zone to account for the evergreen and deciduous tree types in them while classifying, to account for the evergreen and deciduous forest types. Each of these segments were analysed separately. Government owned forest areas were masked based on Survey of India toposheets data to reduce the confusion of forest pixels with coffee category. Training sets of individual classes were captured separately for each of these segmented masks. Thus we derived a signature library for all land use classes separately for each segment. This was done to prevent the mixing of land use type with similar signatures. Approach to assess LU/LC changes from 2005 to 2015 is summarised in the following flow chart (Fig. 3).

Initially, we started the analysis with LULC such as evergreen forest, moist deciduous forest, dry deciduous forest, teak plantation, mixed plantation, coffee agroforest, paddy and built-up area. However, during analysis some of the classes were clubbed and additional classes were included to improve the classification. Finally, the LULC classes identified in this study includes natural forest, coffee agroforest, paddy, irrigated agriculture, built-up area, water bodies, grasslands, rubber plantations, and forest plantations. Table 2 describes the classification scheme used in mapping.

Class

Description

Coffee agroforest

Coffee farms with high tree cover are found in the central part of the landscape. These areas look dark red to bright red on the FCC. Difficult to differentiate from forests type. Rice paddy areas which have been converted to other land use like areca along with coffee fall in this category.

Paddy

Areas seen as grey pixels on the FCC with regular shape.

Irrigated   agriculture

Agricultural areas mainly in the eastern part of the district
have regular shape.

Built-up area

Include urban areas, large residential area, areas like closely
scattered house in village, and coffee processing and drying
yards.

Water bodies

The main rivers, large tanks and ponds.

Grassland

Found usually on mountain tops and distinguished as gray
open areas on the FCC.

Natural forest

Areas look dark red on the FCC and shadows of large tress are
visible. Confusing reflectance with high shade coffee class.

Rubber plantation:

Seen as bright red areas with even texture on a FCC.

Forest plantation:

Usually Tectona grandis (teak) plantation areas have high tree
density and crown cover and such areas are clearly
demarcated with regular shapes on the False Color
composite (FCC) of the satellite images.

 

With these inputs, we proceeded with the classification of the images using minimum distance algorithm. Along with the available four bands of the satellite image we also included Normalized difference vegetation index (NDVI) as an additional band as they improved the classification considerably because of the presence of high vegetation cover in the study landscape. The classification output was converted to vector format and the segments were merged into one output. To proceed with classification cleaning, the classification was converted into a raster and further cleaned using software algorithms within QGIS and using manual cleaning. The output classification was simplified using majority filters and boundary cleaning algorithms. The same procedure was followed for both years of assessment.

 

3. Results and discussions

3.1 GIS data base : The thematic layers such as district and taluk boundaries, forest boundaries, drainage/rivers network and road network are usually present as hard copy data in every departments. These data have limited use as they are not available in GIS compatible formats and also not easily available for the users/managers of natural resources in the region. To fill this void, we captured the data as GIS useable formats which are presented below (Fig. 4-11) and are useful in future for various activities related to landscape management and governance.

Fig. 4 Administrative division with three taluks of the district

Fig. 5 Map showing the road network in the district

Fig. 6 Map showing the drainage network in the district

Fig. 7 Map showing the reserved and protected forests in western and eastern region of the district

Fig. 8 Contour map of the district

Fig. 9 Map showing forest working plan compartments managed by the state forest department

Fig. 10 Map showing different forest ranges and range head quarters under state forest department

Fig. 11 Map showing different forest sections and section head quarters under state forest department

3.1 Land-use land cover mapping and change detection

The classified LULC maps of Kodagu district for the years 2005 and 2015 are given in Fig. 12 and 13. The achieved overall classification accuracies were around 91% and overall kappa statistics was 0.900. The classification results for 2005 and 2015 are summarized in Table 3. Percentage of classes based on these results show the land cover/land use practices observed in the district during 2005 and 2015. The results show that major decline with respect to area coverage in the district was observed in paddy lands (19.03 sq. kms), irrigated agriculture (12.65 sq. kms), natural forest (4.79 sq kms) and rubber plantations (3.08 sq kms). Whereas, area under coffee agroforest (26.82 sq kms), grasslands (7.14 sq. kms) and built-up and settlements (3.92 sq. kms) have increased. There was no change in the area under forest plantations from 2005 to 2015.

 

Land-use classes

2005

2015

Change in area
(sq. Kms)

Area
(Sq. kms)

%

Area
(Sq. Kms)

%

Natural forest

1306.53

31.83

1301.77

31.72

-4.79

 

Coffee Agroforest

 

2083.77

 

50.77

 

2110.59

 

51.42

 

+26.82

 

Rice paddy

 

367.33

 

8.95

 

348.2963

 

8.49

 

-19.03

 

Built-up

 

74.25

 

1.81

 

78.16493

 

1.90

 

+3.92

 

Agriculture (irrigated)

 

35.76

 

0.87

 

23.08378

 

0.56

 

-12.65

 

Water bodies

 

10.57

 

0.26

 

12.24

 

0.30

 

+1.67

 

Grassland

 

80.48

 

1.96

 

87.59693

 

2.13

 

+7.14

 

Rubber plantation

 

32.78

 

0.80

 

29.69568

 

0.72

 

-3.08

 

Forest plantation

 

112.86

 

2.75

 

112.8545

 

2.75

 

0.00

Total

4104.294

100

4104.294

100

 

 

Major land-use changes have occurred in coffee agroforest, paddy lands and irrigated agriculture areas and grass lands. The area under coffee agroforests has increased to the extent of around 2600 ha while area under paddy has declined to the extent of 1900 ha from 2005 to 2015. Similarly, area under irrigated agriculture has declined by 1265 ha during period of 10 years. Other major changes were under natural forest areas which have decline to the extent of 479 ha while the area under settlement has increased by 392 ha from 2005 to 2015. The area under grasslands has increased by 714 ha while area under rubber plantations has decreased. These patterns of land cover/land use change in Kodagu district reinforces that economic forces are commonly a major stimulus on anthropogenic change of land and it is the main reason why the area under coffee has increased. The decrease in paddy and agriculture could be attributed mainly to the cost of production and non-availability of labour. Mostly area under low lying paddy is either being converted into palm plantations and for construction of houses in paddy areas near towns and that is the reason where we observed increase in the area under built-up/settlements. It has also been observed in the present analysis that the Settlements or Built up areas are mostly surrounded by the paddy/Agricultural area, Decrease in natural forest area could be probably due the encroachment and other anthropogenic disturbances like forest fires. The area under rubber plantations have declined mainly because of eviction of the encroachments by the state forest department in natural forest areas of Sampaje and Makutta regions of the district. There could have been probably forest fire in the grassland areas before 2005 and later as a result of protection the area under this category has increased over 10 years. Area under water bodies has increased mainly due to the increase in number of farm ponds for irrigating coffee plantations during summer to enhance coffee flowering (in absence of blossom showers during March-April).

 

4. Linkages with Sentinel Landscape project of CIFOR

Since College of Forestry, Ponnampet is a partner in CGIAR-CIFOR-FTA supported Western Ghats Sentinel Landscape (WGSL) project where Kodagu is one among the four sites in Western Ghats of India. The WGSL project is focused at long term monitoring at landscape level using both biophysical and socio-economic data. One of the major objective of the WGSL is to monitor the landscape dynamics (both historical and future predictions) in Western Ghats and outcome of the current project could be linked to these initiatives. The possible linkage would be to use the data collected under WGSL (10 x 10 km grid near Madikeri) for ground truthing and more intensive analysis could be done coupled with satellite data procured under current project. The other major linkage is to integrate the WGSL with outcome of the current study is to assess the carbon dynamics (both above and below ground) associated with land-use changes.

 

Conclusion:

Analysis of the spatial and temporal pattern of land-use and assessment of the key driving factors behind the associated changes is imperative for proper land-use planning and sustainable utilization. The present study characterized the land-use change of Kodagu district for a period between 2005-2015 where there has been increase in some land-uses and decreased in others. Based on these results we can identify the land-uses where there is desirable change and in some land-uses where the change is undesirable in terms of the sustainability of the landscape. The outcomes of this project are very useful to KMFT which could emphasize to the policy makers the need for policy intervention to halt/reduce the degradation of landscape which is evident from the current analysis.

 

Action Forward:

Based on our long standing association with College of Forestry (UAHS) and Other Stakeholders since inception of our Model Forest Programme we look forward to our continued association and cooperation in the years to come.

© 2018 Kodagu Model Forest Trust. All rights reserved | Design by Yeshua Melech Softwares