AF447 Case: How Ísis Could Have Optimized the Search
Hypothetical case study using Ísis software to simulate the search planning for Air France flight 447
Rio de Janeiro → Paris | 01 Jun 2009
Introduction
On June 1, 2009, Air France flight 447 (AF447), operated by an Airbus A330-200, departed Rio de Janeiro bound for Paris, carrying 228 people. During the Atlantic crossing, the aircraft stopped sending scheduled reports, ceased communications, and disappeared in an oceanic region far from any radar coverage.
The first physical evidence – bodies and debris on the surface – were only located days later, and identifying the impact point and large debris on the seabed required a prolonged search effort.
This document describes, hypothetically, how the search mission planning would have been if Ísis software had been available at the time. The goal is not to reconstruct every actual decision made in 2009, but rather to show the step-by-step reasoning a SAR coordinator could employ using the system's uncertainty, drift, and probability modules.
Methodology Note
In a real open-sea disappearance scenario, there are different modeling approaches. Since this study aims to evaluate Ísis performance in a real scenario as if we were on that night, the simulations are configured only with information available to the coordinator at the time of disappearance. The initial location is defined as a point datum from the last ACARS, and the radial error is not adjusted retrospectively, but created directly from the aircraft's performance. These parameters are kept constant throughout the entire simulation, and only at the end are the actual coordinates of found bodies plotted on the maps for model validation purposes.
Ísis Premises and Capabilities
For this case study, it is assumed that Ísis is fully operational at the Rescue Coordination Center responsible for the AF447 oceanic region. The following modules are particularly used:
Maritime Drift Module (Monte Carlo)
Simulates particle dispersion in open sea from current, wind, and leeway fields, generating probabilistic position maps over time. These are automatic maps, based on IAMSAR principles, with optimizations that capture meteorological data directly and at shorter time steps.
Position Uncertainty Module (Total Error)
Allows representing the initial object position as: point with radial error (point datum), Course & Distance vector with uncertainty cone and directional error calculations, or line datum along a route.
Effort Planning Module
Converts POC maps into optimal search areas based on Available Effort, delivering the spacing and region that will yield the highest possible POS for the scenario, regardless of POC distribution.
It is also assumed that the system has access to historical current and wind reanalyses for the accident date, and leeway tables compatible with typical objects from an aircraft accident at sea (bodies with life vests, seats, fuselage pieces, etc.). The planning described here is conducted as if we were in the initial phase of the case, using only information that could be available to the coordinator at that moment.
Initial Case Data
The starting point for this study is the data set recorded by the coordinator at the time, consolidated in a spreadsheet. Among this data, the coordinate adopted as AF447's Last Known Position (LKP) stands out.
LKP (Last Known Position)
02° 58.08' N / 030° 35.04' W
(2.968 / -30.584)
06/01/2009 – 02:10 UTC
Geometric and Particle Projection Model
In terms of uncertainty geometry, Ísis allows representing the initial position of an object in open sea in three main ways:
1. Point Datum with Radial Error
Particles generated around a single datum, according to a circular error characterized by a radius containing 50% of the probability (R₅₀). Suitable when there is a clear reference position (LKP).
2. Course & Distance Model
Starts from an initial coordinate, a heading, and an estimated distance traveled. The system builds an uncertainty cone in the displacement direction.
3. Multipoint Trajectory Model (Line Datum)
Represents uncertainty along a route: between an initial and final point, associated with a time interval.
For AF447, the combination of automatic failure messages in a short interval, sudden loss of communication, and subsequent knowledge that the impact occurred only a few miles ahead of the LKP indicates a catastrophic event in cruise. Therefore, for this study, the problem is treated as a typical case of point datum with reinforced radial error.
Error Definition Around the Datum
From the point model, the first step is to define how Ísis represents uncertainty around the LKP. Instead of treating the last position as exact, the system considers it the center of a probability distribution, within which the impact may have occurred in different directions and distances.
In this study, the error is described by a 50% Probability of Containment radius (R₅₀): a circle around the LKP that contains half the probability that the target was located at the moment of impact. The other half of POC is distributed decreasingly as distance from this circle increases.
Accident reports indicate the impact point would be within a maximum 40 NM radius from the last transmission received. This radius was adopted as the region with 100% probability. Considering a normal distribution, 50% of POC concentrates in approximately 15 NM. This determination considers AF447's characteristics and creates regions without undersizing the area or expanding it excessively.
Drift Simulation
First Projection Parameters
In this simulation, Ísis no longer models the aircraft trajectory, but rather the behavior of a person in water (PIW) after impact. Parameters are defined from the moment the body is considered to be on the surface, subject to the combined action of current, wind, and leeway.
Reports show that after the last ACARS position message, there was only a few minutes interval until impact. This makes the hypothesis of an organized ditching with evacuation to rafts unlikely, which is why the 'Person in Water – vertical posture' configuration was chosen.
Simulation Parameters
Search Timeline
Below is the day-by-day progression of Ísis simulations, showing POC heat maps and actual search patterns executed. Found bodies are marked on the maps for model validation.
Day 1 Search
06/01/2009
First pattern assigned to SAR 7100 (18:54z, 2h flight). Second pattern to SAR 2474 (22:30z, 3:10h flight, 1000 ft, 2NM spacing).
Patterns within or near initial probability area.
Day 2 Search
06/02/2009
All patterns at 1000 ft altitude, 150 kts speed, and 2 NM spacing.
All remained OUTSIDE Ísis probability area.
Day 3 Search
06/03/2009
Same configuration: 1000 ft, 150 kts, 2 NM spacing.
All remained OUTSIDE Ísis probability area.
Day 4 Search
06/04/2009
Three patterns outside probability area, one tangent to it.
Only one pattern touched Ísis probability area.
Day 5 Search
06/05/2009
Spacing varying between 2 NM and 3 NM.
All sorties remained OUTSIDE Ísis probability area.
Day 6 Search
06/06/2009
First day with positive results. First bodies located at 12:05z and 13:17z.
Bodies are very close to Ísis simulation. Only one pattern covered the probability area.
Day 7 Search
06/07/2009
14 bodies found between 04:20z and 20:40z.
Bodies found ~29 NM from heat map center.
Day 8 Search
06/08/2009
11 bodies found between 16:45z and 23:50z.
Bodies found between 35 NM and 40 NM from heat map center.
Day 9 Search
06/09/2009
14 bodies found between 07:30z and 17:05z.
Bodies found ~40 NM from heat map center.
Day 10 Search
06/10/2009
4 bodies found between 10:05z and 19:00z.
Bodies found ~44 NM from heat map center.
Day 11 Search
06/11/2009
4 bodies found between 12:30z and 17:30z.
Bodies found ~45 NM from heat map center.
Day 16 Search
06/16/2009
1 body found at 14:30z.
Body located ~80 NM from heat map center.
Day 17 Search
06/17/2009
1 body found at 17:30z.
Body located ~87 NM from heat map center.
Final Visualization
Final Considerations
The AF447 study with Ísis clearly shows two complementary aspects: on one hand, the physical adherence of the drift model to the actual evolution of bodies; on the other, the enormous potential efficiency gain if such a tool had been available from the first day of search.
Until the 5th day of operations, practically all actual aerial effort was concentrated in areas far from the heat map generated by Ísis – in some cases, hundreds of nautical miles from the core of the probability cloud. Additionally, in all analyzed patterns, briefings indicated a life raft as the search object, when, given the accident characteristics, person in water (PIW) was much more likely.
Incorrect Search Object
All briefings indicated a life raft as the search object, when person in water (PIW) was much more likely.
Inadequate Flight Configuration
The 1,000 ft configuration with 2-3 NM spacing results in intrinsically low POD for PIW. Beyond searching mostly in the wrong area, the search was conducted with inadequate configuration for the target type.
Drift Model Validation
When bodies start being located (Day 6), actual positions are near the heat map edge and aligned with the predicted displacement trend: Ísis's probability cloud elongates in the same direction where bodies are being found.
Potential Efficiency Gain
In real system employment, the coordinator would have had from the start a probability map concentrating effort in the hottest areas, instead of dispersing sorties in distant sectors.
Each search pattern executed over these higher POC zones would feed Ísis's Bayesian module: detections and non-detections would be used to recalibrate the probability field, reinforcing or attenuating specific regions and progressively narrowing the area of greatest interest.
This means transforming flight hours into information, and information into rapid search convergence.
Case Statistics
228
People on board
51
Bodies recovered
22.4%
Recovery rate
5 days after accident
First body (Day 6)
~115h (mostly outside POC area)
Flight hours (Days 1-5)
It's important to highlight that even when actual body locations appear outside the heat map core, it cannot be concluded that the highest concentration of victims wasn't in the area indicated by Ísis. Only 51 bodies were recovered from 228 people on board – a statistically insufficient sample to rule out the hypothesis that the debris field's center of mass was in the simulated higher POC region. What this study demonstrates is that Ísis offers, from the start, a physically consistent and operationally actionable reference framework, capable of drastically reducing the amount of flight hours spent in low-probability areas and concentrating effort where the chance of success is higher.
Transform Your SAR Operations
Discover how Ísis software can optimize your Search and Rescue missions with precise drift simulations and Bayesian planning.