Fernerkundung in Äthiopien zum Monitoring großer Landpachten Matthias Hack 29.11.2018 Nationales Forum für Fernerkundung und Copernicus 2018 1
Umbau des Landwirtschaftssektor in Äthiopien 80% der Bevölkerung leben im ländlichen Raum, hauptsächlich Kleinbauern mit sehr wenigen Produktionsmitteln und minimalen Einkünften Ungenutzte Produktionsmittel (Land, Arbeitskraft) sollen in Wert gesetzt werden 2
Großflächige Landpachten in Äthiopien (GLP) Kommerzialisierung der Landwirtschaft Auf ungenutzten Flächen (circa 1 Million Hektar) Cash crops (Exportgewinne) + Lebensmittel für den eigen Markt (Ernährungssicherheit erhöhen) Arbeitsplätze Technologietransfer 3
Nur ein Bruchteil ist landwirtschaftlich produktiv geworden Erfolg weit hinter den Erwartungen stattdessen Landkonflikte Scheitern der GLP
Monitoring zum Steuern der GLP Sentinel-1 and 2 (Source: ESA) Identification and demarcation Transferring land Implementation Monitoring of the investments Check and demand compliance Before investment During investment 5
Einfacher, zuverlässiger Ansatz Visual interpretation: Identification of structures indicative of agricultural activities (green arrows), e.g. clearance, fields, canal, rods, etc. On-screen digitization of the identified structure in a GIS (green polygon) Final product: map containing information about agricultural activities (green polygon) within LSAI. Uncultivated LSAI clearly visible (red arrow) 6
Incrementally from basic to sophisticated Data access/ collection Sentinel 1 and 2 data Additional data, e.g. Landsat Reference data Single images Time stacks Automated image classification Processing phase Visual interpretation of satellite data Change detection Crop identification Other analysis Basic monitoring tool Addition of functionality Increasing sophistication and higher prerequisites Sophisticated monitoring tool Delivery to end-users Maps showing cultivated land Reports and statistics Mapping of Land Use Change Maps showing crop types Maps plant condition Maps showing deforestation rates Etc. Setup of the tool Operational use Phased functional evolution of the monitoring tool over time
Ein Laptop und Internetverbindung reichen aus 8
Fernerkundung ist nichts neues aber hat sich enorm weit entwickelt. Die neue Qualität gilt es zu nutzen! l
Vielen Dank für Ihre Aufmerksamkeit Matthias Hack Advisor Rural Development and Agriculture Sector Project Land Governance Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH P.O. Box 5180 65726 Eschborn Germany F + 49 6196 79-1680 E matthias.hack@giz.de I www.giz.de
Backupfolien
Incrementally from basic to sophisticated Data access/ collection Sentinel 1 and 2 data Additional data, e.g. Landsat Reference data Single images Time stacks Automated image classification Processing phase Visual interpretation of satellite data Change detection Crop identification Other analysis Basic monitoring tool Addition of functionality Increasing sophistication and higher prerequisites Sophisticated monitoring tool Delivery to end-users Maps showing cultivated land Reports and statistics Mapping of Land Use Change Maps showing crop types Maps plant condition Maps showing deforestation rates Etc. Setup of the tool Operational use Phased functional evolution of the monitoring tool over time
Sentinel 1 C-band SAR (5.6 GHz) Dual polarization (VV+VH or HH+HV) Interferometric Wide mode for land 10 m pixel spacing S1A launched 3 April 2014 S1B launched 25 April 2016 Same orbit, 6 day revisit
Multi-spectral sensor Sentinel 2 12 bands Visual, near and shortwave infrared 10, 20 and 60 m pixel spacing S2A launched 23 June 2015 S2B launched 7 March 2017 Same orbit, 5 day revisit l
Comparison of time composites from S2 and S1 data NDVI time composite from S2 data VH time composite from S1 data (own figure)
From plot to crop identification In situ data Multi-temporal cloudfree optical images / composites Multi-temporal SAR images Existing VHR (e.g. GoogleEarth) Land use / land cover map Auxiliary data Reference data randomly divided into training (50%) and testing (50%) Supervised image classification Advanced machine learning algorithms (e.g. Random Forest, Support Vector Machines) Additionally post classification smoothing Accuracy assessment