Research Article ii FF o o r r e e s s t t doi: 10.3832/ifor2021-010 Biogeosciences and Forestry vol. 11, pp. 162-170 Monitoring of changes in woodlots outside forests by multi-temporal Landsat imagery M Mahmudur Rahman, Woodlots outside forests play a significant role in diversifying household in- M Shafiqul Islam, come, reducing poverty, supplying timber and fuel-wood, and regulating the M Abu Taleb Pramanik local environment in many countries with low forest cover. This study aimed to develop a method for delineating the spatial footprint of woodlots outside forests and assessing their changes over time. The test site was located in the Jhalokati District of south-western Bangladesh, one of the world’s most dense- ly populated regions. Landsat images from 2010 were classified using a super- vised method. Woodlots were extracted, converted to vector layers, and man- ually edited. The overall accuracy of the 2010 land cover map was 87%-89%. A change vector layer was generated by further updating of the vector layer by overlaying a 1989 Landsat image. The total coverage of woodlots in the district increased between 1989 and 2010, from 19,638 ha (27%) to 27,836 ha (39%). The study identified two primary reasons for changes in woodlot coverage: (i) woodlot expansion associated with the population growth and establishment of new households; and (ii) conversion of cropland to orchards because of eco- nomic reasons. The results will improve understanding of the spatial distribu- tion of woodlot coverage in the study area and their dynamics over time. Keywords: Woodlot, Trees Outside Forests (TOF), Landsat, Change, Mapping Background roadsides, along railways, and in human ers are located one after another and form Forests supply and sustain various forest settlements. These trees cannot be in- a continuous pattern, which is a common products for household uses, support habi- cluded in the category “forests” or “other pattern of rural landscape found in the tats for wildlife, conserve biological diver- wooded land” in forest mapping (Belle- study area. In Bangladesh, the definition of sity, and provide a range of ecosystem ser- fontaine et al. 2002, Rawat et al. 2003). TOF is usually synonymous to village for- vices to local communities. Forests are dy- Woodlots are typically clusters of trees ests which belong to the private ownership namic and change over time with the influ- usually growing outside of forests on crop- and are managed by the respective own- ence of human and natural drivers. Earlier land and in human settlements; it does not ers; in contrast, forests generally belong to efforts at assessing land cover changes include trees growing in isolation or trees the Government, and are managed by the have estimated the extent of forest cover planted in lines or in narrow strips. In state-owned Forest Department. change in different regions (Leimgruber et Bangladesh, the definition of TOF includes TOF, in woodlots, in lines or narrow strips al. 2005, Desclé et al. 2013, Redowan et al. all trees growing outside of forests, usually or in isolation, can be a major source of 2014, Song et al. 2015). Compared to stud- clusters of trees distributed on the rural timber and non-timber forest products in ies of forest cover change, there have been landscape. The extent of woodlots is not countries where forest coverage is low a limited number of efforts to map and limited to a specific threshold, for example (FAO 2001). Woodlots produce forest prod- quantify changes in Trees Outside Forests 0.5 ha, which is the definition the FAO of- ucts (fuel wood and timber), mitigating the (TOF – Plieninger 2012), though it is impor- ten uses for defining forest and assessing demand of local and regional markets and tant to assess the spatial footprints of forest resources. playing a role in diversifying household changes in these trees over time (FAO The extent of woodlots under individual economies (cash income from selling tim- 2001). ownership is often variable and sometimes ber, fruits etc.). Woodlots also contribute TOF usually represents the aggregation smaller than the threshold can be suitably to biodiversity conservation and provide of scattered trees grown on cropland, monitored using Landsat data; it becomes habitat for wildlife. Trees help to maintain grazing land, marginal land, canal banks, larger when the woodlots from many own- water quality and quantity (slowing rainfall runoff, reducing sediment and nutrient loading to waterways, and protecting Forestry Division, Bangladesh Space Research and Remote Sensing Organization, Agargaon, stream banks from erosion), soils (minimiz- Sher-E-Bangla Nagar, 1207, Dhaka (Bangladesh) ing wind and water erosion), and air quality (filtering air pollutants, creating shade, and @ M Mahmudur Rahman ([email protected]) cooling – Mason et al. 2014). Woody plants in TOF will play a role in climate change Received: Feb 21, 2016 - Accepted: Nov 23, 2017 adaptation, mitigation (reducing the im- pact of floods, storm-surges, droughts Citation: Rahman MM, Islam MS, Pramanik MAT (2018). Monitoring of changes in woodlots etc.), and carbon sequestration (Guo et al. outside forests by multi-temporal Landsat imagery. iForest 11: 162-170. – doi: 2014, Mason et al. 2014). 10.3832/ifor2021-010 [online 2018-02-19] The structure and composition of wood- lots and other TOF might be more hetero- Communicated by: Luca Salvati geneous and inconsistent than regular forests; their extent is generally smaller © SISEF http://www.sisef.it/iforest/ 162 iForest 11: 162-170 Rahman MM et al. - iForest 11: 162-170 y r and distribution is sparse. Land tenure is 1901/1905 and 2009 using analysis of multi- were considered in the class “Villages”, t s often different, and consequently their as- temporal maps. which was included in other land category. e r sessment might require a different ap- The different image processing tech- The class consists of small rural settle- o proach than estimation of forest cover (De niques used in woodlot mapping include ments without trees, rural settlements F Foresta et al. 2013). The economic contri- pixel-based methods (Rawat et al. 2004), with trees (size between 0.1 and 0.5 ha), d n bution of these trees to rural livelihoods object-based image analysis (Aksoy et al. and rural settlements with trees (size more a and their ecological roles in rural land- 2010, Pujar et al. 2014), or a hybrid ap- than 0.5 ha). The data was obtained from s e scapes are usually not well-known beyond proach that combines both pixel- and ob- National Forest Inventory and calibrated by c n the local level (De Foresta et al. 2013). ject-based classification techniques (Sheer- remote sensing (BFD 2008). Earlier studies e i Therefore, local level studies are important en et al. 2009). The remotely sensed im- conducted in Bangladesh did not aim to c s to assess and map the spatial footprint of agery used in TOF mapping are aerial pho- generate full coverage maps, statistics and o woodlot resources to analyse their tographs (Liknes et al. 2010, Meneguzzo et change assessments in trees or woodlots e g changes and to understand their spatial ex- al. 2013), high- to medium-resolution satel- outside forests. The contribution of this o i tent, their dynamics, and their role in the lite images, i.e., QuickBird satellite imagery study was to present a technique for as- B local economy and ecology. (Aksoy et al. 2010), IRS Cartosat-1 (Pujar et sessing changes in woodlot coverage over – Local-level studies have been conducted al. 2014), IRS LISS-IV (Pujar et al. 2014), IRS time using multi-temporal Landsat data. t s in different regions to assess and map LISS-III (Rawat et al. 2004), and moderate The current study focused on the map- e r woodlots and other tree resources grow- resolution Landsat Enhanced Thematic ping of woodlots (cluster of trees in regu- o F ing outside of forests. Liknes et al. (2010) Mapper Plus (ETM+ – Premakantha et al. lar or irregular shape), including orchards i estimated tree coverage located outside of 2008). with fruit producing trees located outside forests in North and South Dakota, USA us- The earlier studies mostly generated forests, using Landsat Thematic Mapper ing various layers of data-sets. Lister et al. woodlot and tree cover maps of rural land- (TM) data. We also assessed their changes (2009) conducted an inventory of trees lo- scapes for a single time frame (Liknes et al. over time. Mapping of individual scattered cated outside forest areas in the Great 2010, Pujar et al. 2014); only a few investi- trees or trees planted in lines or narrow Plains states, USA. Perry et al. (2009) used gations have assessed the changes in tree strips (e.g., roadside plantation) was be- Forest Inventory and Analysis (FIA) data of coverage in multi-temporal domains. Plien- yond the scope of current study. The ex- the USDA (United States Department of inger (2012) considered published maps as tent of individual scattered trees or trees Agriculture) Forest Service to estimate the historical information and satellite images planted in linear shapes as tree coverage is extent of tree resources on agricultural as recent information in the change analy- much lower in the study areas compared to lands as well as other landscapes in the sis. The objectives of this study were to: (i) the woodlots. Midwest region of the United States and develop a method for mapping and ana- compared the estimates to those derived lysing change of woodlots (including or- Materials and methods from Moderate Resolution Imaging Spec- chards) located outside the forest from troradiometer (MODIS) Vegetation Contin- multi-temporal satellite (Landsat) data; (ii) Study areas and datasets uous Fields (VCF).