Rapid detection of new and expanding human settlements in the Limpopo province of South Africa using a spatio-temporal change detection method

Abstract

Recent development has identified the benefits of using hyper-temporal satellite time series data for land cover change detection and classification in South Africa. In particular, the monitoring of human settlement expansion in the Limpopo province is of relevance as it is the one of the most pervasive forms of land-cover change in this province which covers an area of roughly 125 000km 2 . In this paper, a spatio-temporal autocorrelation change detection (STACD) method is developed to improve the performance of a pixel based temporal Autocorrelation change detection (TACD) method previously proposed. The objective is to apply the algorithm to large areas to detect the conversion of natural vegetation to settlement which is then validated by an operator using additional data (such as high resolution imagery). Importantly, as the objective of the method is to indicate areas of potential change to operators for further analysis, a low false alarm rate is required while achieving an acceptable probability of detection. Results indicate that detection accuracies of 70% of new settlement instances are achievable at a false alarm rate of less than 1% with the STACD method, an improvement of up to 17% compared to the original TACD formulation.

A novel framework for parameter selection of the Autocorrelation Change detection method using 250m MODIS time-series data in the Gauteng province of South Africa

“…The temporal ACF change detection (TACD) method proposed in Kleynhans et al(2012) with a slight generalization provided in Kleynhans et al (2015) was developed specifically using an 8-daily 500 m MODIS time-series based on the MCD43A4 product and can be defined as:…”

“…Data for the second cube were similar to that of the first cube with the key difference being (1) the spatial resolution was 250 m as opposed to 500 m, (2) the temporal resolution was daily as opposed to 8-daily and (3) only two bands were available as opposed to seven (due to fact that only the first two MODIS bands have a spatial resolution of 250 m). It was shown in Kleynhans et al (2015), where a similar land cover change was considered, that multiple band combinations as well as vegetation indices including NDVI and EVI were not able to provide significant improvement over that of using only band 1. Band 1 is the red band in the visible spectrum range 620-670 nm and is known to be very sensitive to changes in vegetation.…”

“…As for the selection of the δ * parameter, there is a requirement to produce a false alarm rate of ≤ 1% as discussed in section 1, this is due to the fact that the change algorithm is run over large areas and the validation of a large number of false alarms could be costly and time consuming. To determine the threshold value that would yield a false alarm rate in this range, a similar approach was used as in Kleynhans et al (2015) where the Bayesian decision error was evaluated based on the distribution of the inferred change index (δ) using a training no-change dataset as follows:…”

“…In many cases, new human settlements as well as existing settlements expand informally and these expansions occur in areas that were previously covered by natural vegetation. Satellite time-series data has proven to be an effective data source for change detection (Verbesselt et al ,2010;Lunetta et al, 2006;de Beurs and Henerby, 2005) and in particular, time-series analyses of hyper-temporal satellite data has been successfully applied to land cover change detection in South Africa Kleynhans et al, 2015Grobler et al, 2012, Grobler et al, 2013. In Kleynhans et al, 2012 a temporal Autocorrelation change detection (TACD) method was demonstrated to detect the development of new informal settlements in South Africa.…”

From Land to Sea, a Review of Hypertemporal Remote Sensing Advances to Support Ocean Surface Science

“…Since 1995, a diverse range of applications have been developed which involved extracting meaningful information from hypertemporal data. These include landcover mapping [10,13], landuse mapping [14], change detection [15][16][17][18], ecosystem structure and species modelling [19][20][21][22][23], phenology mapping [24][25][26], gradient analysis [27][28][29], and data quality assessment [30]. To date, a limited range of ocean-focused hypertemporal studies have been conducted (Figure 3), hampered by validation challenges and limited knowledge on the partitioning of temporal patterns over the ocean's surface.…”