Limitations of Remote Sensing in Assessing Vegetation Damage Due to the 2019–2021 Desert Locust Upsurge

Abstract

The 2019–2020 Desert Locust (DL) upsurge in East Africa threatened food security for millions in the region. This highlighted the need to track and quantify the damaging impacts of the swarming insects on cropland and rangelands. Satellite Earth observations (EO) data have the potential to contribute to DL damage assessments that can inform control measures, aid distribution and recovery efforts. EO can complement traditional ground based surveys (which are currently further limited due to COVID-19), by rapidly and cost effectively capturing the full spatial scale of the DL upsurge. However, EO-based techniques struggled to accurately quantify damage from this DL upsurge due to the sporadic and localized nature of infestations impacting scale, timing, and anomalous vegetation conditions. This study analyzed time series data from MODIS, the harmonized Landsat Sentinel-2 product, and C-band radar data from Sentinel-1 to distinguish DL damage from normal senescence or other confounding factors from January to June 2020. These data were compared to in situ locust swarm, band, and non locust observations collected by the Food and Agriculture Organization (FAO) and PlantVillage. The methods presented did not produce results that could confidently differentiate senescence from locust activity, and may represent a limitation of publicly available remotely sensed data to detect DL damage. However, the higher spatial resolution data sets showed promise, and there is potential to explore commercially available satellite products such as Planet Labs for damage assessment protocols.

Outbreak of Moroccan Locust in Sardinia (Italy): A Remote Sensing Perspective

“…To address this objective, field data with specific information on ground detected vegetation damage and timely coupled satellite data at higher spatial, temporal, and spectral resolution are required. Previous investigations utilizing MODIS data showed that optical moderate-resolution sensors might be insufficient to detect vegetation damage related to locust infestation [80,81].…”

A framework for modelling desert locust population dynamics and large-scale dispersal

“…Recently, data analytics methods, such as regression and machine learning, have been applied to predict breeding areas of DL (Piou et al, 2013;Gómez et al, 2018;Kimathi et al, 2020;Ellenburg et al, 2021;Klein et al, 2022), potential habitat areas of DL Meynard et al (2017); Guan et al (2021); Saha et al (2021); Youngblood et al (2022), presence of DL (Tratalos et al, 2010;Piou et al, 2019;Samil et al, 2020;Tabar et al, 2021;Gómez et al, 2021a;Shao et al, 2021;Yusuf et al, 2021;Sun et al, 2022;Rhodes and Sagan, 2022;Cornejo-Bueno et al, 2022) and the incidence of locust outbreaks ((Lawton et al, 2022)). They have also been applied to detect DL impact on vegetation senescence (Adams et al, 2021), and cropland damage (Alemu and Neigh, 2022). Major challenges remain in spatiotemporal forecasting of DL, in particular to predict the long-range movements of swarms.…”