Flood Mapping to Inform Disaster Response in Central America

Flooding regularly impacts the Central American region. These flooding disasters are often related to extreme climatic events, including heavy rainfall and hurricanes. In November 2020, Hurricanes Eta and Iota both hit Central America weeks apart, causing severe landslides and flooding. It is estimated that these hurricanes affected at least 7.3 million people across the region.  

Due to the frequency of hurricanes in Central America, understanding how to prevent, prepare, and monitor flooding is crucial for reducing their impacts on local communities in the future. The NASA DEVELOP Central America Disasters project aimed to assist partners in Central America in preparing for floods in the event of hurricanes or heavy rainfall.  

Focusing on Belize, Guatemala, El Salvador, Honduras, Nicaragua, Panama, and Costa Rica, the Central America Disasters team used NASA SERVIR’s Hydrologic Remote Sensing Analysis for Floods (HYDRAFloods) tool to better understand the impacts of flooding on Central America, with the goal of supporting effective response and monitoring efforts across the region.  

As climate change threatens to increase the frequency and intensity of these storms, project partners in Central America want to better understand the impacts of inundation to prepare for future flooding events. This project aimed to use remote sensing and training tutorials to assist partners’ current disaster monitoring efforts. The primary objectives of this project include improving partners’ ability to monitor surface water, understand flood risks, and create informed response efforts.  

To support partners’ current disaster monitoring and response through remote sensing analysis and training materials, the DEVELOP team produced four end products: 1) historical surface water maps, 2) a case study of Hurricanes Eta and Iota, 3) a precipitation analysis over the region, and 4) a code tutorial on the HYDRAFloods tool to enable real-time flood monitoring and response.

The team created historical surface water maps and precipitation maps to identify possible areas of vulnerability for floods. Additionally, the team analyzed a case study of Hurricanes Eta and Iota using HYDRAFloods to demonstrate the potential of the tool for disaster analysis. Finally, we designed a coding tutorial for HYDRAFloods to enhance the partners’ monitoring capacity. 

We first mapped land cover using the European Space Agency's WorldCover dataset to help us contextualize the areas where floods occur. The map below shows the major land cover types for the study region. Central America is characterized by large forested and vegetated regions. Large, visible sections of cropland are seen on the southwestern coast of Guatemala, El Salvador, and Nicaragua, with smaller sections seen in Costa Rica and Panama.

For this analysis, we focused primarily on cropland, built-up, and forest land cover classes to overlay on flooded areas. Areas of interest in Guatemala, shown by the inset maps, were largely forested and vegetated, with small built-up areas. The Sula valley is characterized by cropland and built-up areas, surrounded by forest cover. The southwestern coast of Honduras is similarly classified, with wetland areas at the coast. 

The inset maps display many differences are between the 2015 surface water maps and the 2021 surface water maps. The areas of interest in El Salvador, Guatemala, Panama, Honduras, and Nicaragua show noticeable increases in the presence of surface water from 2015 to 2021. Belize and Costa Rica show similar levels of surface water presence between the study period. However, these inset maps only show a select number of sites throughout the region. The changes in surface water throughout the region are visualized below:

These maps show the results of the change detection analysis from 2015 to 2021. The purple regions represent areas that changed into surface water, while the orange areas represent areas that were surface water in 2015 but not surface water in 2021. Gray areas indicate regions of no change. All of the inset maps have some presence of purple, indicating that these areas experienced changes to surface water across the study period. This is especially noticeable in Panama, Nicaragua, and Costa Rica. Belize, Guatemala, El Salvador, and Honduras also show an increase in surface water.  

We also performed calculations on how the surface water changed over time using these two maps. The number of pixels classified as surface water actually slightly decreased for all countries except for El Salvador and Guatemala. This could potentially be due to an increase in urban development, but this is an area for future research.

The first slideshow examines flooding in Guatemala and Honduras using SAR imagery.

This next slideshow examines five different satellites and sensors used to produce flood maps. The three inset maps represent different regions of interest with priority sites provided by CEPREDENAC. Each inset map has a during and after flood extent for the back-to-back hurricanes. "During Eta and Iota" represents aggregated imagery from October 31, 2020, to November 18, 2020. The "After Eta and Iota" period is from November 18, 2020, to December 2, 2020. Red represents flood extent, while permanent water from the JRC Global Surface Water dataset is in blue.

In the Alta Verapaz region, Landsat 7 does not detect much flooding in either period. Meanwhile, in the Sula Valley, we see extensive flooding in the "during" period, whereas the "after" period shows almost no flooding. This lack of flooding in the "after" period is likely a result of extensive cloud cover preventing the sensor from detecting flood. Lastly, the Reserva de Vida Silvestre San Bernardo has surrounding floods during and after the two hurricanes.

The results for Landsat 8 were minimal in these regions. The time and area with clear flood extent are in the "during" Eta and Iota period for the region near the Reserva de Vida Silvestre San Bernado. The other regions being characterized by little to no flood is likely a result of excessive cloud cover and limitation in the temporal resolution of the imagery.

In contrast, Sentinel-2 picks up significant flooding most notably in the Sula Valley and in the Reserva de Vida Silvestre San Bernardo. With the flood extent being quite large in the "during" Sula Valley map, we can see this flooding occurs in nearby cropland. We also extracted flood using Suomi NPP VIIRS imagery. Although the images are coarse in spatial resolution, we can still see similar areas indicate flooding around the same time periods. Lastly, we measured flood using Terra MODIS imagery. Similar to the prior optical imagery results we see similar flood patterns and extent, although the satellites and sensors differ.

The images in this slider examine flood extent in the Sula Valley across three time periods after Hurricanes Eta and Iota. One week after the events is represented in orange, two weeks after is shown in green, and the three week after period is in purple. These images compare flood mapping using SAR and Optical imagery.

The precipitation results demonstrate that Hurricane Eta released more precipitation in Central America than Hurricane Iota. Most of the precipitation was concentrated on the east coast of Central America, with Honduras and Guatemala receiving the majority of the precipitation, as evident by the precipitation graph. Additionally, El Cajon Dam in Honduras received 74% of the total precipitation of Hurricane Eta and Iota. Countries like Costa Rica and Panama received less rain than Honduras and Guatemala but still suffered heavy impacts on the community. 

1. Use HYDRAFloods to create flood depth estimation maps. 

2. Analyze why surface water presence decreased slightly from 2015 to 2021 across countries as a whole. 

3. Explore data fusion methods like logistic regression to combine resulting SAR and optical flood maps.

4. Find rain thresholds for areas deemed vulnerable by the surface water maps.

5. Create a GUI utilizing HYDRAFloods for Central America.