Computing the Rate of Disappearance of Cropland Using Satellite Images

Proceedings of the 3rd International Conference on Computational Sustainability

Sunandan Chakraborty, Scot Dalton, Yaw Nyarko, Lakshminarayanan Subramanian

Croplands worldwide are in decline. Degradation of arable land is a cause for concern, especially in developing countries where agriculture, including subsistence farming, makes up a significant percentage of economic output. In developing regions, urban population is increasing, leading to expansion of cities and development of new cities or townships. Often these expansions are done on arable lands. Apart from urban expansion, industrial developments are often done on agricultural land [2012]. All these result into acquisition and loss of arable lands. On many occasions, these acquisitions are unplanned and unauthorized. Such loss of arable land can have huge impact, particularly for agrarian economies. Not only it can affect the lives and livelihoods of the population who are directly dependent on agriculture, it can directly impact food security due to reduced production. Apart from human-led development, changing climate is also leading towards a change in the land pattern. There are reports of Sahara desert expanding southwards in an alarming rate [2012], loss of land in low lying coastal areas due to rising sea level etc. In both the two scenarios described above, a solution to the problem can be a monitoring tool, which can identify the change in land pattern over the years. In this paper, we present a tool that can monitor this change through satellite images. Google Earth (GE) offers a huge corpus of satellite images across the globe. GE has the current image of a location, as well as an archive of older images of the same places. Given a location or a geographical area, our tool can access the latest available satellite image in addition to earlier images available and classify the images, as cropland, developed, forest or barren. Following this classification process, the tool computes the total amount of change of pattern in the region and also the type of change (e.g. crop land changed to developed land etc.). Figure 1 shows some GE images depicting the loss of vast open land over the years in 2 African cities. This figure also explains how GE images can be used to detect such changes. Similar approaches can be seen in various fields, like famine and agriculture [Quinn et al, 2010][Nivens et al, 2002], environmental changes to detect outbreak of diseases [Ford et al, 2009] etc. In our case, we used raw photographs taken from the satellites from a freely available source with extensive coverage. This makes the approach much more scalable.