CBRN training simulation software: detection, decontamination, and response

CBRN — chemical, biological, radiological, and nuclear — training is among the most technically demanding domains in defense simulation. Unlike military training simulation architecture for kinetic scenarios, CBRN simulation must model invisible threats that propagate through the environment according to atmospheric physics, interact with specific sensor types through well-defined detection chemistry, and require procedurally exact response behaviors that are both life-critical and cognitively demanding under protective equipment. Building software that can deliver this training fidelity while remaining usable by unit instructors without specialist CBRN modeling expertise is the central design challenge.

This article covers the software architecture and implementation details for CBRN training simulation across the full training envelope: atmospheric dispersal modeling, detector simulation, decontamination procedure training, MOPP level management, federation integration, and after-action review.

Why CBRN training requires dedicated simulation software

General-purpose military training simulators do not adequately support CBRN training because the threat physics are fundamentally different from kinetic engagements. A tank engagement has line-of-sight geometry and ballistic trajectories — both are geometrically tractable and computationally well-understood in military simulation. A chemical agent release has atmospheric dispersion physics, dose-response casualty modeling, detector alarm logic with probabilistic false positives, and decontamination procedure sequences. None of these are modeled in standard simulation entity libraries.

The fidelity requirements for CBRN training simulation are distinct in three ways. First, the threat is invisible: trainees cannot directly observe the hazard, so the simulation must provide realistic indirect cues — detector alarms, casualty symptoms, MOPP order triggers — that require trainees to reason about the threat rather than observe it directly. Second, response procedures are life-critical and procedurally exact: in real operations, an incorrect decontamination sequence can cause secondary contamination that is as dangerous as the original exposure. The simulation must enforce procedural correctness rigorously, not approximately. Third, protective equipment degrades performance: in MOPP 4 gear, cognitive load increases, manual dexterity decreases, and communication is impaired. Training simulation that does not model these MOPP-induced performance penalties misses a critical part of the operational picture.

Hazard realism in CBRN training simulation requires validated atmospheric dispersion models that produce agent concentration fields consistent with real-world field trial data. Standardized response training requires a scenario library built around doctrinal response procedures — FM 3-11, ATP 3-11.32, STANAG 2352 — with procedural error detection that flags deviations from doctrine rather than approximating them. These requirements together make CBRN training simulation a specialized engineering domain that general military simulation frameworks address only partially.

Chemical agent dispersal models

The Gaussian plume model is the baseline for chemical agent dispersal in CBRN training simulation. It models downwind agent concentration as a bivariate Gaussian distribution in crosswind and vertical dimensions, with dispersion parameters (sigma-y and sigma-z) derived from Pasquill-Gifford atmospheric stability classes (A through F, representing convective to very stable conditions). For a continuous release at height H above ground, the on-axis downwind concentration at distance x is:

C(x, y, z) = (Q / (2π · u · σy · σz))
             · exp(-y² / (2σy²))
             · [exp(-(z-H)² / (2σz²)) + exp(-(z+H)² / (2σz²))]

Where:
  Q   = release rate (mg/s)
  u   = mean wind speed (m/s)
  σy  = crosswind dispersion coefficient (m), function of x and stability class
  σz  = vertical dispersion coefficient (m), function of x and stability class
  H   = effective release height (m)
  y   = crosswind distance from plume centreline (m)
  z   = height above ground (m)

Pasquill-Gifford sigma parameters are tabulated by stability class and downwind distance. For training applications, these are typically implemented as polynomial fits to the Pasquill-Gifford curves, allowing fast computation at each simulation time step without lookup tables.

The Gaussian plume model has well-known validity limitations that must be communicated to training audiences. It assumes a flat terrain with homogeneous wind field, a steady-state source, and wind speeds above approximately 1 m/s. For training scenarios in complex terrain, urban environments, or calm-wind conditions, the Gaussian puff model provides better fidelity by tracking individual puff trajectories through a non-uniform wind field.

CAMEO/ALOHA (Computer-Aided Management of Emergency Operations / Areal Locations of Hazardous Atmospheres), developed by NOAA and EPA, implements Gaussian plume and heavy-gas models validated against industrial chemical release data. Its scenario library includes military-relevant chemical agents (GB, VX, HD are included in ALOHA's chemical database as their civilian equivalents) and its output — a hazard zone polygon at specified concentration thresholds — can be imported into exercise mapping tools as a simulation inject. Many CBRN training programs integrate ALOHA as a validated reference model for instructor use, with the training simulation running a computationally lighter Gaussian model for real-time entity concentration sampling.

Terrain effects on chemical agent dispersal are a significant source of simulation error when training in realistic environments. Valleys channel agent downwind at higher concentrations than flat-terrain models predict; building wakes trap agent pockets on the leeward side of structures; dense vegetation reduces wind speed and extends agent persistence. Training simulations operating in complex terrain should apply terrain-correction factors to Gaussian model outputs, or use a Lagrangian particle dispersion model that explicitly tracks particle trajectories through a terrain-following wind field derived from a mesoscale weather model or range meteorological system.

JCAD and detector simulation

Detector simulation translates the continuous agent concentration field from the dispersal model into the discrete alarm outputs that trainees actually observe during training. The simulation must model each detector type's specific sensing technology, sensitivity, alarm threshold, false-positive behavior, and output format. Using a generic binary alarm model for all detectors removes the training value that comes from understanding each system's specific limitations.

The M8A1 Chemical Agent Alarm is the legacy standard, using ion mobility spectrometry (IMS) with a single alarm threshold for G-series nerve agents and blister agents. Its simulation model requires four parameters: the detection threshold (approximately 0.02 mg/m³ for GB), the alarm latency (8–12 seconds from threshold crossing to audible alarm), the saturation level (above which the detector may not function reliably), and the false-positive probability as a function of interferent concentration (diesel exhaust, aircraft fuel vapors are common interferents in field conditions). M8A1 simulation does not require modeling concentration bands — the output is binary.

JCAD (Joint Chemical Agent Detector) simulation is more complex. JCAD provides three alarm levels (low, medium, high) corresponding to concentration bands above detection threshold, and transmits its alarm state over a standardized radio interface. Simulating JCAD requires:

JCAD Alarm State Logic:

Input: C = agent concentration at detector location (mg/m³)

if C < 0.005 mg/m³ (GB equivalent threshold):
    state = NO_ALARM

elif 0.005 ≤ C < 0.02:
    state = LOW_ALARM

elif 0.02 ≤ C < 0.1:
    state = MEDIUM_ALARM

elif C ≥ 0.1:
    state = HIGH_ALARM

Latency: 8–15 s from concentration threshold crossing
False-positive P(alarm | no agent) = f(interferent_type, interferent_conc)
Hold duration: 60 s minimum before returning to NO_ALARM
MOPP recommendation: transmitted via radio on HIGH_ALARM

False-positive modeling is essential for training realism. JCAD in operational environments produces false positives from jet exhaust, vehicle diesel emissions, and certain cleaning compounds. A training simulation that never generates false alarms teaches units to treat every alarm as a confirmed chemical event — a dangerous cognitive habit in the field where false alarms are common and unwarranted MOPP 4 posture degrades operational performance. The simulation should inject false positives at a historically realistic rate (approximately 1–3 per 8-hour exercise day in a mechanized unit's environment) and train units to verify alarms through M256A2 kit detection or multi-detector confirmation rather than responding to single-detector alarms with full-force MOPP escalation.

Point detector placement strategy is itself a trainable skill. The simulation should allow exercise controllers to reposition virtual detectors during planning phases and observe the resulting coverage pattern against a sample chemical release plume. Detectors placed upwind of the unit formation provide no useful detection, a fundamental placement error that training simulation can reveal clearly.

Decontamination procedure training

Decontamination procedure training is the component of CBRN simulation that most directly reduces training errors with life-critical consequences. The individual decontamination sequence under FM 3-11 doctrine is precisely specified, and deviations from the sequence — particularly incorrect removal order of MOPP gear — can transfer agent from the outer surface of protective gear onto previously clean skin or inner clothing. Training simulation must model this consequence explicitly rather than treating decontamination as an abstract procedure completion.

Individual decontamination is modeled as a finite state machine with states for each step of the IEDK (Individual Equipment Decontamination Kit) procedure. The canonical state graph for nerve agent individual decon is:

INDIVIDUAL DECON STATE MACHINE

[CONTAMINATED]
    │ (within 60s of exposure)
    ▼
[SKIN_DECON] — M291 kit: wipe exposed skin (face, neck, hands)
    │ (minimum 60 s, buddy check)
    ▼
[EQUIP_DECON] — M295 IEDK: decontaminate weapon, mask, equipment surfaces
    │ (minimum 90 s)
    ▼
[MOPP_REMOVAL_OUTER_GLOVES] — buddy removes outer gloves (touching exterior only)
    │ (buddy-assisted, 30 s minimum)
    ▼
[MOPP_REMOVAL_SUIT] — buddy removes JSLIST/MOPP suit (rolling outward)
    │ (60 s minimum)
    ▼
[MOPP_REMOVAL_BOOTS] — buddy removes overboots
    │ (30 s minimum)
    ▼
[MOPP_REMOVAL_INNER_GLOVES]
    │
    ▼
[MOPP_REMOVAL_MASK] — last item removed (highest risk)
    │
    ▼
[SKIN_WASH] — soap and water, or RSDL at earliest opportunity
    │
    ▼
[DECON_COMPLETE]

Procedural errors detected:
  - State transition out of sequence → SECONDARY_CONTAMINATION_RISK
  - Missing buddy-assist → SOLO_REMOVAL_ERROR
  - Step duration below minimum → INCOMPLETE_DECON_WARNING
  - Mask removed before suit → CRITICAL_ERROR (modeled as exposure event)

Collective decontamination station simulation models throughput as a function of crew proficiency, equipment availability, and agent type. A standard PDDE (Powered Decontamination and Detection Equipment) station using the M12A1 processes vehicles at 45–90 minutes each depending on vehicle type and contamination level. The simulation should track the decon line queue, compute throughput based on crew skill modifiers, and calculate residual contamination on each vehicle post-decon. Residual contamination is modeled as a probabilistic function of agent persistence (tabun and sarin are non-persistent; mustard agent (HD) is persistent at temperate temperatures), decontaminant type (DS2, bleach slurry, supertropical bleach), dwell time on surface, and surface material (rubber retains agent longer than painted metal).

Decision branching in decon simulation covers the medical decision points that commanders face: when to accept residual contamination risk and continue operations versus halting for complete decontamination. The simulation should present commanders with explicit decision nodes at which they observe the current contamination status of their formation, the decon throughput rate, and the tactical situation, and must choose a decon posture. The AAR should then analyze whether their chosen posture was consistent with the contamination data available at the time of decision.

Protective posture and MOPP level exercises

MOPP (Mission Oriented Protective Posture) level management is a collective training task that operates at all echelons simultaneously. Individual soldiers must don or remove specific components of protective gear within specified time limits; commanders must order level changes based on threat assessment and balance protection against performance degradation; and the entire formation must coordinate transitions to avoid windows of partial protection at the collective level.

The four MOPP levels specify which protective equipment is worn:

MOPP Level	Mask	Suit	Gloves	Boots	Transition Time
MOPP 0	Carried	Carried	Carried	Carried	—
MOPP 1	Carried	Worn	Carried	Carried	8 min from MOPP 0
MOPP 2	Carried	Worn	Carried	Worn	+2 min from MOPP 1
MOPP 3	Worn	Worn	Carried	Worn	+3 min from MOPP 2
MOPP 4	Worn	Worn	Worn	Worn	+2 min from MOPP 3

Cross-unit coordination is the critical MOPP training task at battalion level and above. When a battalion receives a MOPP 4 order, not all companies transition simultaneously — units in contact may be unable to mask safely while under fire, logistical elements may be in vehicles with collective protection, and medical elements have specific MOPP procedures for patient care. The simulation must model each unit independently, allowing the training audience to observe the patchwork MOPP posture across the formation and practice coordinating simultaneous transitions without degrading tactical continuity.

Time pressure is introduced by linking simulated chemical release events to the MOPP transition clock. When the simulation releases a chemical agent while a formation is transitioning from MOPP 2 to MOPP 4, individuals who have not yet completed their transition receive contamination exposure. The simulation should track and report the number of individuals exposed during a transition — this is a directly actionable training data point that motivates faster, more disciplined MOPP transition drill.

Performance degradation modeling under MOPP applies heat stress, reduced manual dexterity, impaired communication (voice attenuation through the mask, reduced radio clarity), and reduced visual field to all entities at MOPP 3 and MOPP 4. These modifiers affect the simulation's movement rates, engagement time parameters, and communication reliability. Training scenarios that do not apply these modifiers systematically underestimate the operational cost of chemical protection and train units to accept MOPP 4 too casually.

Integration with LVC and LSST frameworks

CBRN simulation components must integrate with the broader live virtual constructive integration exercise federation to participate in combined-arms training scenarios. CBRN elements in isolation — a CBRN decon company running a decon exercise — do not represent the realistic challenge of CBRN response during ongoing kinetic operations. The most valuable CBRN training occurs when the formation must respond to a chemical event while simultaneously managing a kinetic threat, sustaining logistics, and maintaining command and control.

The SISO CBRN FOM supplement defines the HLA object classes and interaction classes required to represent CBRN entities within an RPR-FOM federation. Chemical cloud objects carry attributes for agent type (coded per AC 225(D) chemical agent enumeration), source location in geocentric coordinates, release rate, and current atmospheric stability class. The dispersal model updates cloud object attributes at each simulation time step, and subscribing federates can sample concentration at their entity locations using the cloud geometry.

XMSF (Extensible Modeling and Simulation Framework) support allows validated CBRN dispersal models to be exposed as web services discoverable by other federation participants. A scenario manager can invoke a CBRN dispersal service by agent type, release parameters, and meteorological conditions, receiving a contaminated area polygon in return without embedding the dispersal model in the scenario management component. This architectural separation allows the authoritative dispersal model to be updated to a higher-fidelity implementation without modifying the scenario management code.

DIS entity type codes for CBRN hazard entities use the DIS Entity Type enumeration domain 9 (Environmental) with country code 0 (other) and specific entity/category codes defined in SISO ENUM-70 for chemical, biological, radiological, and nuclear hazard representations. Gateway implementations bridging DIS environments to HLA must maintain the mapping between DIS entity type codes and CBRN FOM supplement object class parameters to avoid losing agent type information across the protocol boundary.

The LSST (Live System Software Testbed) architecture provides integration points for live CBRN detector systems. When a real JCAD unit is carried by an instrumented live force participant, LSST's sensor gateway can inject the real detector's alarm state into the constructive simulation as an authenticated HLA interaction, tagged with the entity identity of the carrier. This creates a hybrid scenario where real detector alarms drive simulated consequence modeling — an important validation environment for checking whether unit CBRN response SOPs work as designed before live chemical exercises.

AAR for CBRN scenarios

CBRN after-action review requires an event log with four additional data streams beyond standard exercise AAR: exposure log, detection log, decontamination log, and MOPP state log. Each stream must be time-synchronized with the standard entity position and engagement log so that CBRN events can be correlated with tactical events during the debrief.

The after-action review software should automatically calculate the following CBRN-specific metrics from the event log:

Metric Category	Metric	Doctrinal Standard
Exposure	Cumulative Ct dose per individual (mg·min/m³)	Agent-specific IDLH threshold
Detection	Time from release to first alarm (seconds)	<120 s for upwind detector placement
Detection	Fraction of formation warned before exposure	>90% for effective detector placement
MOPP	Time from MOPP 4 order to 90% compliance (min)	<8 min (FM 3-11 standard)
Decon	Time from contamination to skin decon initiation	<60 s (immediate decon standard)
Decon	Procedural errors per individual decon sequence	0 critical errors (e.g. mask-off before suit)
Collective decon	Vehicle decon throughput rate (vehicles/hour)	1–1.5 vehicles/hr per M12A1 PDDE station

Exposure tracking in the AAR requires recording the concentration-time history for each individual entity during the exercise. Because the concentration field changes at each simulation time step, the raw record is a time series of (entity_id, timestamp, agent_type, concentration_mg_per_m3) tuples sampled at the simulation time step (typically 1 second). The AAR system integrates this time series to compute Ct (concentration-time product in mg·min/m³) for each entity and each agent type, then compares against agent-specific physiological thresholds to assess casualty risk.

Decontamination time recording logs the start and end of each decon procedure step per individual with the responsible buddy partner and any procedural error events. The AAR replays these records in a timeline view that allows the debrief facilitator to step through the decon sequence step by step, highlighting errors and showing the simulated contamination consequence of each error in context.

Procedural error detection operates in two modes: real-time detection (flagging errors during the exercise for instructor notification) and post-hoc analysis (computing error statistics across the formation for the AAR report). Real-time detection allows instructors to observe the decon sequence and intervene when critical errors are detected — removing a mask before the suit is a life-critical error in real operations that should trigger immediate instructor intervention in training. Post-hoc analysis aggregates error types across the formation to identify systematic training deficiencies versus individual mistakes, which drives different training responses.