2.1 Study area
The study was conducted in KC and adjacent areas (Udagaji, Mgugwe and Ukami villages), referred to as the Kihansi area. It is located in the Kilombero and Mufindi Districts of south central Tanzania (07815°–08845°S, 35800°–37800°E). Kihansi catchment covers about 0.9 km2 and rises from 300m to around 2,500m above sea level (asl) in the east and north-south respectively. The area contains a closed canopy of wood and rain forests with mixed tree species which including Mninga (Pterocarpus angolensis), Mpangapanga (Millettia stuhlmannii), Msekeseke (Bobgunnia madagascariensis), Pamosa (Afzelia quanzensis), Msufi (Bombax rhodognaphalon), and Mtondo/Mtondoro (Julbernardia globiflora), Mhongo, and other diverse endemic and endangered flora and fauna (Kibbassa, 2014). Crop production and livestock keeping are the major economic activities in this area and covering approximately 85% and 10 % respectively (Kibbassa, 2014). The study area varies in weather and elevation leading to the cultivation of different crops, fruits and vegetables. Maize (Zea mays), cassava (Manihot esculenta), sweet potatoes (Ipomoea batatas), bananas (Musa sp), oranges (Citrus sinensis), and guava (Psidium guajava) are grown in both areas (Kibbassa, 2014). Wheat (Triticum aestivum), finger millet (Eleusine coracana) and round potatoes are crops grown in highlands (Kibbassa, 2014) whereas paddy (Oryza sativa), sorghum (Sorghum bicolor), coconut (Cocos nucifera), cashew nuts (Anacardium occidentale), palm-oil and sugarcane (Saccharum officinarum) are crops grown in lowlands (Kibbassa, 2014; Nwilene et al., 2013; Yudha et al., 2019).
2.2. Image acquisition, processing and analysis
The area under the study were identified, demarcated and different categories of LULC were classified using Google earth Pro as suggested by Ngongolo, Estes, Hudson, & Gwakisa (2019). The area was monitored for the last 25 years from 1995-2020. Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) imagery were used to analyze land cover change of the years1995, 2004 and 2020, respectively. The aim of choosing these years was to compare the LULC of the KC and its adjacent areas before dam construction (1995), after dam construction (2004) and after restoration and reintroduction of the toads (2020). The Kihansi catchment area extends within a single Landsat path and row. High resolution satellite imagery (30m spatial resolution) was used to download images from the US Geological Survey (USGS) Earth Explorer (http://earthexplorer.usgs.gov).
Table 1
Satellite image descriptions
(Source: http://earthexplorer.usgs.gov).
Description |
Year | Path | Raw | Space craft ID | Sensor ID | Date acquired |
1995 | 168 | 66 | Land Sat 5 | TM | 23 June 1995 |
2004 | 168 | 66 | Land Sat 5 | TM | 17 July 2004 |
2020 | 168 | 66 | Land Sat 8 | OLI TIRS | 14 August 2020 |
The images were selected based on season (dry season) as previously suggested by (Mmbaga et al., 2017) because during the dry season, it was possible to identify forested land and the remaining bare lands were considered agricultural lands. All images (1995, 2004 and 2020) were obtained within the same time of the year (ranging between June 23 and August 14) (Table 1). Shape files for study site boundaries were obtained from The National Bureau of Statistics (NBS) available at (https://openmicrodata.worldpress.com). Satellite images of different years were imported into ArcGIS (version 10.3) for processing and analysis. The geographic coordinate system was defined as the World Geographical System (WGS) 1984 and projected to the Universal Transverse Mercator (UTM) Zone 37S prior to analysis. Image processing and analysis included image clipping and composite bands formation. In this study, natural colour bands were used. Three Landsat images were classified by using the maximum likelihood function, which is the most common decision rule among the supervised classification (Mmbaga et al., 2017). It is also considered to give very accurate results because each pixel is assigned to the class to which it has a highest probability of belonging (Mmbaga et al., 2017) Visual interpretation and digital image classification were then combined using GIS functions.
Three land use classes defined in the study area were: (1) agricultural land (seasonal cultivated lands, bush lands, bare lands open areas, grasslands, and livestock grazing); (2) forested land (3) Settlement (any type of buildings). Training sites were determined and signature files were created to be used in the classifications by using ArcGIS (version 10.3). After that, the classified images were compared with ground truthing (field visit) that was achieved with the assistance from Kihansi management team members to validate the collected information and modified accordingly. Additionally, to improve the accuracy on the past images of the years 1995, 2004, and 2020 interviews with local people were conducted. To improve and analyze the data at a better quality scale, the LULC classification and analysis were done across the three villages: Udagaji, Mgugwe and Ukami.
2.3. Data collected on Land use land covers changes through questionnaire survey from local communities adjacent the Kihansi area.
The interviews involved 156 respondents from the three study villages (Udagaji, Mgugwe and Ukami). Extensive field observations and interviews were conducted by administering a structured questionnaire to respondents aged 45 years and above, who had lived in the respective location for at least 25 years from the age of self-awareness (15years). The information captured through the questionnaire was as follows: awareness of respondents on land use types of KC and adjacent areas in relation to KST conservation within the range of 1995 to 2021. In addition, the socio-economic activities of local communities and the trends of different land use types over the past 25 years were captured. In addition, reasons that could promote changes in land use and cover were addressed in the questionnaire survey. Furthermore, the dynamic changes in terms of the human population, livestock, agricultural land, water bodies and forested land for the past 25 years were investigated during the questionnaire survey. The variation of opinions obtained from respondents across the three study villages was analyzed using the Kruskal Wallis statistical test.