Back to the fields: COVID-19 impact on agricultural activity detected with satellite data

This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.


Download Preprint


Ahmed Hammad, Giacomo Falchetta , I. B. M. Wirawan


In response to the 2020 COVID-19 pandemic, policymakers worldwide adopted unprecedented measures to limit disease spread, with major repercussions on labour markets and economic growth. Here we provide empirical evidence of their impact on agricultural activity due to sectoral labour reallocation. Analysing daily satellite data in a non-parametric machine learning statistical framework over cropland in Badung, a highly populated regency of Bali, Indonesia, we generate a counterfactual synthetic Enhanced Vegetation Index (EVI) based on gradient boosted decision trees trained on a set of environmental variables assuming no lockdown occurrence. Based on the counterfactual, we estimate a significant increase in the EVI over cropland after the beginning of the lockdown period. The finding is robust to a placebo test. We then exploit the heterogeneity of the region analysed, where the South is dominated by tourism and the tertiary sector and the North is already mostly agricultural, and we find a stronger effect in the former. This results suggests that the observed increase in remotely sensed agricultural productivity indexes is likely driven by a labour force reallocation towards the primary sector to compensate for the income lost from previous employment. Overall, our results show that statistical analysis of satellite data can be an effective methodology to observe the impact of a labour force crowding into the agricultural sector in response to an exogenous shock in other labour sectors.



Environmental Studies


remote sensing, machine learning, COVID-19, agricultural productivity, causal inference, labour market


Published: 2021-02-03 07:26


CC BY Attribution 4.0 International

Additional Metadata

Conflict of interest statement:

Data Availability (Reason not available):
The R and Google Earth Engine JavaScript code to replicate the analysis and the figures are publicly hosted at The repository includes references to retrieve the input data.

Add a Comment

You must log in to post a comment.


There are no comments or no comments have been made public for this article.