Fishing for better tracking: Tracing anglers in the Gulf Coast 

The Gulf Coast is recognized worldwide for its exceptional fishing opportunities, offering anglers a wide variety of species such as trout, red snapper and grouper. Recreational fishing represents a significant — yet historically undermeasured — aspect of overall fisheries pressure in the region, raising concerns about its impact on fish populations and the need for effective conservation and management.   

Dr. Alan Kuhnle, assistant professor in the computer science and engineering department at Texas A&M University, is using smartphone mobility data collected from anglers to develop machine-learning approaches to estimate recreational fishing effort and classify fishing behaviors.     

“This work addresses a significant gap in fisheries monitoring,” Kuhnle said. “While commercial fishing is tracked through mandatory automatic identification systems (AIS) and vessel monitoring systems, private recreational fishing has remained difficult to quantify at scale. With better data on where and when recreational fishing occurs, managers can make more targeted decisions that protect fish populations while maintaining opportunities for anglers.”   

For sustainable fisheries management, understanding total fishing effort from commercial and recreational sources is also essential for setting science-based catch limits and preventing overfishing. From an economic perspective, recreational fishing contributes billions of dollars annually to Gulf economies. Better data on fishing effort and participation support informed policy decisions that affect coastal communities dependent on fishing tourism and related industries.   

Current methods for estimating recreational fishing rely on surveys and dockside intercept programs, costing millions while missing substantial activity. In contrast, Kuhnle’s approach leverages the ubiquity of smartphones combined with modern machine learning to provide continuous, cost-effective monitoring. The trajectory data from these devices can be used to identify likely fishing trips and estimate effort without requiring anglers to report their activities.     

Their model works by first ingesting and standardizing raw GPS trajectories from angler smartphones and AIS data from commercial and recreational vessels. Distinguishing the type of fishing trip is essential, as ecological impacts differ between trip types.     

“Private recreational anglers, charter boats and dive operators target different species and have different catch rates,” Kuhnle said. “Understanding which type of trip is occurring helps managers understand pressure on specific fish populations. Fishing regulations differ based on vessel type, gear, target species and location. Attributing fishing effort to the correct category is necessary for evaluating compliance and the effectiveness of different regulations.”    

According to Kuhnle, accurately attributing fishing pressure to specific sectors enables managers to develop targeted policies rather than one-size-fits-all regulations. If a particular industry is growing rapidly or has an outsized impact on specific species, managers can respond with sector-specific measures.    

After determining the trip type, the machine-learning technology continuously streams data into discrete trips and applies quality control filters to manage noise and missing observations. From there, they can compute more than 70 behavioral features that capture movement patterns, speed profiles, spatial clustering, trajectory statistics and temporal dynamics of the trip.     

In the final modeling stage, they train the machine-learning classifiers to differentiate vessel types and identify when fishing occurs, both at the trip level and at finer timepoint resolution. The outputs from these models are combined to produce robust predictions of private recreational fishing efforts, which are calibrated against traditional survey data for consistency.   

The team has made significant progress on the vessel classification component of the pipeline. Using the first quarter of 2019 AIS data from the Gulf, which comprises over 182 million position reports, they validated their machine-learning approach on approximately 80,000 trips across 7,159 unique vessels. For multi-class vessel type classification, they achieved 74.3% accuracy in categorizing vessels into cargo, tanker, fishing and pleasure craft types. For binary detection tasks, their models achieved 94.3% accuracy in detecting pleasure craft versus other vessel types, and 95.3% accuracy in detecting fishing vessels versus other vessel types.  

These results demonstrate that trajectory-based features alone can reliably distinguish vessel types without requiring vessel registration lookups. These results are a critical building block for their goal of identifying private recreational fishing trips from smartphone data.  

The two-phase approach of the larger project — combining smartphone mobility data with satellite imagery — provides multiple complementary methods for estimating fishing efforts and validating results across different data sources.   

 Looking forward, they hope to provide fisheries managers with high-resolution, near-real-time estimates of recreational fishing effort broken down by area and time.   

 “We aim to enable evidence-based policies that balance conservation goals with sustainable angler access. If our methods prove successful in the Gulf, they could be applied to other marine regions and management contexts worldwide, wherever smartphones and vessel tracking systems can provide trajectory data on recreational boating activity,” Kuhnle said.    

 This project is funded by National Oceanic and Atmospheric Administration Fisheries, via the Gulf States Marine Fisheries Commission and LGL Ecological Research Associates, with a subaward to Texas A&M.  

This work builds directly on the methodology developed by Drs. Mona Ahmadiani and Richard Woodward in Texas A&M’s Department of Agricultural Economics. Ahmadiani is the principal investigator for the subaward. Dr. Nathan Putman at LGL Ecological Research Associates is the principal investigator while Taylor Beyea is a co-Investigator. 

The Texas A&M work is being carried out in conjunction with Drs. Ahmadiani and Woodward, along with contributions from graduate research assistant Jessica Magana.