Space Beats Time: a spatio-temporal method for analyzing abrupt environmental change

We present a methodological framework called Space Beat(s) Time (SBT) to detect and analyze abrupt environmental changes. SBT is based on the understanding that predictions based on neighboring observations (geographic context) outperform predictions based on past observations in the immediate aftermath of disruptive events, such as natural disasters or human induced rapid change (e.g. fast urbanization). SBT differs from other change methods in several aspects: (1) it focuses on disruptive events or human interventions, (2) it explicitly leverages the statistical properties of interrupted spatio-temporal series by comparing the predictive power of spatial and temporal autocorrelation; and, (3) it provides a framework for interpreting natural and socio-ecological phenomena. We use measures of vegetation greenness (NDVI) and night light use emissivity (NLU) as proxy variables for pre- and post- event change trajectories. We run spatial autoregressive (SAR) and ARIMA temporal models to predict values for pre- and post- event dates. We estimate each model's performances by calculating their overall prediction mean absolute error (MAE), and by inspecting maps of residuals. These residual images provide potential spatial clusters of abrupt change/impact that can be used to target relief, identify hotspots of change for health monitoring and to monitor longer-term change.

We illustrate SBT's application using case studies including Hurricane Maria (2017) in Puerto Rico as well as how urbanization and flooding events relate to malaria infection prevalence in Dar es Salaam (Tanzania) and Dakar (Senegal). We evaluate the framework potential in the context of current and previous case studies.

Results indicate that SBT is a promising novel methodological framework with the potential to inform the answers to key questions for many disciplines, such as the measurement of second-order recovery and resilience in socio-ecological systems.