by Marthalina Indhawati (Indonesia)
Nowadays, idle land (land abandonment) is a serious problem in Indonesian land management. About 7.3 million hectare is estimated as idle land spread over the whole country. It causes many disadvantages and increases land conflicts in Indonesia. Hence, the national government gave instructions to the National Land Agency of the Republic of Indonesia to control idle land in Indonesia. However, idle land controlling has some major obstacles during indicated idle land inventory; one of them is that Hak Guna Usaha land parcels (HGU - right to cultivate) are located far away from the capital city, resulting in difficulties to enforce controlling mechanisms. Remote sensing has been used successfully for many years in detection of changes and together with time series analysis it will lead to a better understanding of land use changes. In this study we provide a possible method for performing indicated idle land inventory based on remote sensing methods in which, we aimed: 1) to find out whether there is any land clearance/deforestation on HGU land parcels or not and 2) to find what is the HGU land cover, four years after the publication of idle land controlling regulation and compare, whether it is the same as the purpose of the entitlement. Firstly, we used Landsat satellite images for the years 2008 until 2014 and applied the Break For Additive Season and Trend Monitor (BFAST Monitor) with a monitoring period from the 1st of January 2013 to the 1st of July 2014 to detect spatial and temporal changes for five HGU land parcels in the Riau Province – Indonesia. We combined the breakpoints and set a threshold of 0, -0.025, -0.05, -0.075, -0.1 for the magnitude value to evaluate which threshold would obtain the best accuracy in detecting land clearance for the five study areas. We validated our BFAST monitor results, by using SPOT 6 images of the years 2013 and 2014. The purpose of this validation was to validate, the accuracy of the BFAST monitor in detecting land clearance. We chose the cross point between user’s and producer’s accuracy as an unbiased estimator in order to get the best threshold for detecting land clearance for each study area. In general, a magnitude threshold of -0.075 gave the best results in land clearance detection for this particular study area. This threshold value corresponded to the highest overall accuracy of 88.24%. We concluded that the BFAST Monitor could be used for detecting land clearance in the five HGU land parcels by combining breakpoints and magnitude values. Secondly, we performed a maximum likelihood supervised classification for a Landsat 8 image using a SPOT 6 image acquired on 17 June 2014 as the ground truth for observing the existing land cover. We validated our classification results by using a SPOT 6 image for the year 2014 in combination with the Kampar land cover map of the year 2014. We performed random sampling and combined the results into one confusion matrix for the five study areas. The results of this accuracy assessment showed an overall accuracy of 86.13% and a Kappa coefficient of 0.78, which corresponds to a substantial agreement. We concluded that detecting bare soil is more accurate than vegetation i.e. oil palm, forest, and grassland. In general, it can be concluded that Landsat time series are suitable to be used for idle land early detection. However, high cloud cover was a major aspect causing a low data availability, which limited the usability of our methodology. By combining all the results in this study, it gives many opportunities for the National Land Agency of the Republic of Indonesia to have an idle land early detection method to decide whether a land parcel will be targeted in the idle land controlling or not. A continuously development in the idle land controlling stages is necessary for the National Land Agency of the Republic of Indonesia to gain a proper land management in Indonesia.
Keywords: idle land; Landsat; time series; BFAST Monitor; land cover classification; Indonesia.