RECENT STRUCTURAL CHANGE IN REMOTE SENSING DATA TIME SERIES LINKED TO FARM MANAGEMENT IN HORN OF AFRICA (1999 – 2009)

摘要

Food security in East Africa region essentially dependsrnon the stability of rain-fed crops farming, which rendersrnits society vulnerable to climatic fluctuations. Thesernones in Africa are most widely and directly related tornrainfall. In this study, the relation between recent spatialrnrainfall variability and vegetation dynamics has beenrninvestigated for East Africa territories. Satellite rasterrnproducts SPOT-4 Vegetation 1 km resolution (Saint,rn1995) and RFE (rainfall estimates) from Famine EarlyrnWarning Systems Network (FEWS NET) are used. Thernsurvey is carried out at administrative level scale usingrn10-day summaries extracted from raster data for eachrnspatial area unit thanks to specific polygonal layers.rnTime series covers two different periods: 1996-2009 forrnrainfall estimates and 1999-2009 for NDVI. The firstrnstep of the analysis has been to build for eachrnadministrative unit a coherent set of data, along the timernseries, suitable to be processed with state-of-artrnstatistical tools. The analysis is based on the assumptionrnthat every structural break in vegetation dynamics couldrnbe caused by two alternative/complementary causes,rnnamely: (i) modifications in crop farming systemsrn(adaptation strategy) related to eventual break-shift inrnrainfall regime and/or (ii) other socio-economic factors.rnBFAST (Verbesselt et al, 2010) R package arernemployed to lead a comprehensive breakpoint analysisrnon 10-day RFE (spatial mean and standard deviation)rnand 10-day NDVI ones (spatial mean, mode andrnstandard deviation). The cross-viewing of the yearsrnwhere significant breaks have occurred, throughoutrnopportune GIS layering, provides an explorativerninterpretation of spatial climate/vegetation dynamics inrnthe whole area. Moreover, the spatial and temporalrnpattern of ecosystem dynamics in response to climaticrnvariability has been investigated using waveletrncoherency by SOWAS R package (Maraun, 2007). Thernwavelet coherency (WCOH) is a normalized time andrnscale resolved measure for the relationship between tworntime series (Maraun and Kurths, 2004). This kind ofrnmulti-scale temporal investigation provides anrnexplanation of break detected in time series, confirmingrnor not their climatic linkage; results of the analysis arernshown.rnFinally, in order to support the dissemination andrnsharing of information, interactive vegetation maps havernbeen implemented with Google Earth mash-up. Thernmaturity of Web-based GIS enables the generation ofrnthematic maps dynamically and efficiently, with arnthin/thick client or hybrid architectures. This could be arngreat support for the understanding environmentalrnphenomena.