Seeking Flood Justice in the Rio Grande Valley
Researchers launch a new program that uses AI and collaborative mapping to help border residents in need after flooding disasters.
“Back in 2020 when we had the hurricane, my house fell from the foundation, so we had to pick it up and strap it down. It was a huge mess,” said Lizbeth Ramos, a community organizer who lives near Donna, Texas, in a colonia—a type of unincorporated community that tends to be among the hardest hit by large storms. “Whenever it floods where I live, I see a pattern: my light goes off, the yard will flood, the middle of the street will flood to the point where it’s thigh high, and the entrance to the community also floods. If you have a little car, you won’t make it through.”
As The Border Chronicle reported, from March 26 to 28, at least 21 inches of rain fell on the Rio Grande Valley of South Texas and northern Tamaulipas. In some colonias, such as Las Milpas near the town of Pharr, water spilled into residents’ living rooms. Yet while colonias bear the brunt of these events’ damage, they are also often overlooked for government aid. This happens because of inadequate flood maps, barriers to navigating applications for flood relief, and racial and economic marginalization.
This month, a team of researchers launched a new tool called FLUJOS (Flood Justice Utilizing Satellite Observation), developed in partnership with colonia residents, that uses satellite imagery and machine learning to provide fine-grained flood maps that more accurately represent on-the-ground realities. They hope these will help communities advocate for improved disaster preparation and relief.
According to the federal government, colonias are “rural communities within the U.S.-Mexico border region that lack water, sewer, or decent housing, or a combination of all three.” South Texas is home to the vast majority of colonias—over 2,000, according to the official count. FLUJOS cocreator Lucas Belury added that the communities are predominantly Mexican and Mexican American, and they often include manufactured homes and irregular housing. “Lack of infrastructure typifies a colonia,” he said. At the same time, he added, “the lack of physical infrastructure has a counterweight in the social infrastructure of solidarity and community and the opportunity for low-income folks to own homes.”
Belury said there are three main reasons that colonias often experience exacerbated effects of flooding relative to other areas. First, they are unincorporated: instead of receiving services from municipalities like McAllen or Brownsville, they are built on county land. South Texas’s small, largely poor counties have correspondingly small tax bases, translating into limited resources to provide services or ensure code enforcement.
Second, for most of the 20th century, the Rio Grande Valley was dominated by agriculture. Many colonias were built on former farmland. “The water infrastructure that exists is to bring water to these areas from the Rio Grande, not drainage to bring it away,” explained Belury.
Finally, the flooding itself is worsened by a lack of access to postdisaster support from government agencies like FEMA. “We see time and time again that colonias are really limited in the postflood aid that they are provided,” said Belury. Though FEMA doesn’t provide race data, he added, studies have shown that predominantly Black and Hispanic census tracts have higher rates of denial for postdisaster support programs. After 2008’s Hurricane Dolly, the organization La Unión del Pueblo Entero (LUPE), represented by Texas RioGrande Legal Aid (TRLA), sued FEMA over denying compensation to residents whose homes the agency considered to be in “poor condition” before the storm.
Belury, a Mexican American geographer originally from Austin, came to the Rio Grande Valley in 2019 to research how colonia residents recovered from floods, including using predatory loans like car title loans to obtain funds to rehabilitate their homes. That work caught the eye of Ana Laurel, a disaster attorney at TRLA. After Hurricane Harvey, many of her clients had described problems with flooding and drainage. But there wasn’t good data to back up their claims. There is no digital FEMA floodplain map for much of the region, only scans of maps dating to the 1980s, and FEMA maps don’t incorporate pluvial (rainstorm) flooding, the primary culprit in the valley.
“We were looking at this gap between what is going on and what our clients have access to,” Laurel said, “and we thought that a map that has satellite data but that also considers what is actually happening in the colonias, from the perspective of the colonias, might be effective to show what is actually happening.”
In 2022, after Belury started a doctorate in geography at the University of Arizona, he and Professor Beth Tellman of the Social Pixel Lab partnered with five organizations in the valley—LUPE, Proyecto Azteca, TRLA, Border Workers United, and ARISE Adelante—to bring these maps to life.
During five workshops, colonia residents used markers and sticky notes to indicate on large maps where they remembered seeing flooding after different storms, how high the water was, and how long the water lasted. Then, Social Pixel Lab members used the maps to teach a machine-learning algorithm how to recognize floods in images from the satellites Planetscope and Sentinel-1. Machine learning is a type of artificial intelligence in which computers learn to recognize patterns that they can then apply to new data.
“Imagine three screens: the participatory map on the left, before-and-after satellite imagery on the right, and in the middle an image of a flooded neighborhood,” explained Belury. “And I’m marking by hand where I see flooding onto the middle map, dozens of times, in concert with other labelers. We feed that into a machine-learning algorithm to teach it what a flood looks like in the RGV.”
The result is FLUJOS, which went live on April 1, 2025. The bilingual website offers users a drop-down list of 12 past flood events that affected the valley. (This month’s floods are not included, said Belury, because obtaining and preparing the data takes weeks, but they are a priority to be added.) After selecting an event, users can type in an address, a colonia, or another location. The map will zoom to the area, showing locations that were flooded in light blue.
FLUJOS is designed to be applicable to a variety of uses, from individuals’ decision-making and self-advocacy to community efforts to hold officials accountable and secure funds for services and infrastructure.
“Let’s say I want to buy another house or lot. It would be good to know that the place is not at risk, because that will affect how we want to lay the foundation or whether we even want to buy there,” said Ramos. “For community organizing, it can be a way to report to Hidalgo County or FEMA that, hey, even though you don’t believe it’s a flooding area, these are the facts. Look at the satellite.”
Another use, said Belury, could be identifying priority areas for services like clearing property titles, a requirement for qualifying for FEMA assistance. But, he said, any use that community members find valuable is a positive outcome.
“My dream is that this serves the needs of community-based organizations and that they find it useful and valuable,” said Belury. “I would be over the moon if someone were to use FLUJOS’s tool to identify flood-vulnerable places and to bring needed resources and infrastructure to those places.”
This is such good news ! I had not known how much the flood responses bypassed colonias and other poor areas. I did live in a colonia outside of Brownsville in the 90's during one small hurricane and remember the flooding. Thanks for such a sensible and helpful use of AI !