TY - UNPB

T1 - Remote sensing of urban cyclone impact and resilience

T2 - Evidence from Idai

AU - Fisker, Peter Kielberg

AU - Malmgren-Hansen, David

AU - Sohnesen, Thomas Pave

PY - 2021/6

Y1 - 2021/6

N2 - Cyclone Idai, the most devastating cyclone ever recorded in Southern Africa, caused havoc in large parts of central Mozambique, especially the port city of Beira, upon its landfall in March 2019. This study reviews and compares measurements of the impact, using various sources of remote sensing data. Furthermore, taking into account pre-cyclone neighbourhood characteristics and post-cyclone developments in building quantity and quality, it is shown that: i)spatial patterns of severity of impact vary substantially by data source and method used, ii) a small but measurable share of the cross-city variation in cyclone damage can be explained by prior neighbourhood-level characteristics, and finally, iii) a convolutional neural network building detector and classifier applied to a panel of high resolution satellite images fails to prove clearpatterns in the post-cyclone rebuilding of Beira up to 17 months after the cyclone. Although the present analysis shows mixed results, and more work validating the various links between satellite data, operational use, and economic analysis is needed, remote sensing may still be the best option in a data-scarce disaster situation.

AB - Cyclone Idai, the most devastating cyclone ever recorded in Southern Africa, caused havoc in large parts of central Mozambique, especially the port city of Beira, upon its landfall in March 2019. This study reviews and compares measurements of the impact, using various sources of remote sensing data. Furthermore, taking into account pre-cyclone neighbourhood characteristics and post-cyclone developments in building quantity and quality, it is shown that: i)spatial patterns of severity of impact vary substantially by data source and method used, ii) a small but measurable share of the cross-city variation in cyclone damage can be explained by prior neighbourhood-level characteristics, and finally, iii) a convolutional neural network building detector and classifier applied to a panel of high resolution satellite images fails to prove clearpatterns in the post-cyclone rebuilding of Beira up to 17 months after the cyclone. Although the present analysis shows mixed results, and more work validating the various links between satellite data, operational use, and economic analysis is needed, remote sensing may still be the best option in a data-scarce disaster situation.

KW - Faculty of Social Sciences

KW - remote sensing

KW - cyclones

KW - resilience

U2 - 10.35188/UNU-WIDER/2021/029-0

DO - 10.35188/UNU-WIDER/2021/029-0

M3 - Working paper

T3 - W I D E R. Working Papers

SP - 1

EP - 15

BT - Remote sensing of urban cyclone impact and resilience

ER -