Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Assessing Condensation Exposure Risk in Maritime Container Transport: A Data-Driven Screening Methodology Using Pre-Departure Conditions

Polozani, Arban LU (2026) MTTM10 20261
Packaging Logistics
Department of Design Sciences
Abstract
Condensation within maritime transport containers can damage moisture-sensitive cargo, yet the risk is typically assessed only after conditions during transportation are known, leaving limited opportunity for preventive action. The present study develops a data-driven approach to quantifying condensation-exposure risk from temperature and relative humidity sensor data and evaluates whether that risk can be anticipated from information available prior to departure.
With the use of sensor data from 322 closed-phase container transports, incorporated with ERA5 reanalysis climate data, condensation exposure was quantified as Econ, the proportion of closed-phase observations fulfilling established relative humidity and dewpoint depression... (More)
Condensation within maritime transport containers can damage moisture-sensitive cargo, yet the risk is typically assessed only after conditions during transportation are known, leaving limited opportunity for preventive action. The present study develops a data-driven approach to quantifying condensation-exposure risk from temperature and relative humidity sensor data and evaluates whether that risk can be anticipated from information available prior to departure.
With the use of sensor data from 322 closed-phase container transports, incorporated with ERA5 reanalysis climate data, condensation exposure was quantified as Econ, the proportion of closed-phase observations fulfilling established relative humidity and dewpoint depression thresholds. Across the data cohort, exposure was strongly right skewed, with 30.4% of transports reaching or exceeding the 25% exposure criterion used to separate exposed from non-exposed transports. Four pre-departure factors, historical route risk, thermal direction, departure-month deviation and initial loading relative humidity, displayed significant associations with exposure and were combined into a logistic regression model aimed at predicting the likelihood of elevated exposure.
The model achieved moderate discrimination (AUC 0.74) and was well calibrated across the predicted probability range. Transports ranked in the upper 40% by predicted probability accounted for 71% of all exposed transports, substantially more than expected under random selection. The regression model reliably distinguishes highly elevated-risk transports from the rest but does not as reliably separate degrees of risk among transports already given a low probability. This limitation is attributable in part to outcome variability unexplained by the applied predictors.
The findings from this study indicate that condensation-exposure risk can be quantified directly from sensor data and meaningfully anticipated, though not precisely predicted, using only information already available pre-departure. (Less)
Abstract (Swedish)
Kondens inuti containrar inom sjöfrakt kan skada fuktkänsligt gods, men risken bedöms vanligtvis först efter att förhållandena under transporten är kända, vilket lämnar litet utrymme till förebyggande åtgärder. Detta examensarbete utvecklar ett datadrivet tillvägagångssätt för att kvantifiera risken för kondensexponering utifrån sensordata av temperatur och relativ luftfuktighet, samt undersöker om denna risk kan förutses med information tillgängligt före avgång.
Med hjälp av sensordata från 322 slutna containertransporter, kompletterat med klimatdata från ERA5-reanalys, kvantifierades kondensexponering som Econ, andelen observationer i den slutna fasen som uppfyller etablerade kriterier för relativ luftfuktighet och daggpunktsdepression.... (More)
Kondens inuti containrar inom sjöfrakt kan skada fuktkänsligt gods, men risken bedöms vanligtvis först efter att förhållandena under transporten är kända, vilket lämnar litet utrymme till förebyggande åtgärder. Detta examensarbete utvecklar ett datadrivet tillvägagångssätt för att kvantifiera risken för kondensexponering utifrån sensordata av temperatur och relativ luftfuktighet, samt undersöker om denna risk kan förutses med information tillgängligt före avgång.
Med hjälp av sensordata från 322 slutna containertransporter, kompletterat med klimatdata från ERA5-reanalys, kvantifierades kondensexponering som Econ, andelen observationer i den slutna fasen som uppfyller etablerade kriterier för relativ luftfuktighet och daggpunktsdepression. Exponering över hela datasetet var starkt högerskev, där 30.4% av transporterna nådde eller översteg tröskelvärdet på 25% som användes för att särskilja exponerade från icke-exponerade transporter. Fyra faktorer kända före avgång, historisk ruttrisk, termisk riktning, avvikelse i avgångsperiod och initial relativ luftfuktighet vid lastning, visade signifikanta samband med exponering och integrerades i en logistisk regressionsmodell avsedd att prediktera sannolikheten att en transport är exponerad.
Modellen uppnådde måttlig diskrimineringsförmåga (AUC 0.74) och var väl kalibrerad över hela predikterade sannolikhetsintervallet. 40% av transporterna som rankades högst efter predikterad sannolikhet stod för 71% av samtliga exponerade transporter, betydligt fler än vad som skulle förväntas vid slumpmässigt urval. Modellen särskiljer på ett tillförlitligt sätt tydligt högriskexponerade transporter från övriga, men särskiljer inte grader av risk lika väl bland transporter som redan tilldelats en låg sannolikhet. Denna begränsning kan delvis förklaras av utfallsvariation som de applicerade faktorerna inte fångar.
Resultaten visar att risken för kondensexponering kan kvantifieras direkt från sensordata och till viss del förutses, om än inte exakt förutsägas, med information tillgänglig före avgång. (Less)
Please use this url to cite or link to this publication:
author
Polozani, Arban LU
supervisor
organization
alternative title
Utvärdering av Kondensrisk vid Sjöburen Containertransport: En Datadriven Screeningmetodik Baserad på Förhållanden Före Avgång
course
MTTM10 20261
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Condensation risk, Maritime container transport, Screening model, Logistic regression, Pre-departure prediction
language
English
id
9250276
date added to LUP
2026-09-08 08:05:48
date last changed
2026-09-08 08:05:48
@misc{9250276,
  abstract     = {{Condensation within maritime transport containers can damage moisture-sensitive cargo, yet the risk is typically assessed only after conditions during transportation are known, leaving limited opportunity for preventive action. The present study develops a data-driven approach to quantifying condensation-exposure risk from temperature and relative humidity sensor data and evaluates whether that risk can be anticipated from information available prior to departure.
With the use of sensor data from 322 closed-phase container transports, incorporated with ERA5 reanalysis climate data, condensation exposure was quantified as Econ, the proportion of closed-phase observations fulfilling established relative humidity and dewpoint depression thresholds. Across the data cohort, exposure was strongly right skewed, with 30.4% of transports reaching or exceeding the 25% exposure criterion used to separate exposed from non-exposed transports. Four pre-departure factors, historical route risk, thermal direction, departure-month deviation and initial loading relative humidity, displayed significant associations with exposure and were combined into a logistic regression model aimed at predicting the likelihood of elevated exposure. 
The model achieved moderate discrimination (AUC 0.74) and was well calibrated across the predicted probability range. Transports ranked in the upper 40% by predicted probability accounted for 71% of all exposed transports, substantially more than expected under random selection. The regression model reliably distinguishes highly elevated-risk transports from the rest but does not as reliably separate degrees of risk among transports already given a low probability. This limitation is attributable in part to outcome variability unexplained by the applied predictors.
The findings from this study indicate that condensation-exposure risk can be quantified directly from sensor data and meaningfully anticipated, though not precisely predicted, using only information already available pre-departure.}},
  author       = {{Polozani, Arban}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Assessing Condensation Exposure Risk in Maritime Container Transport: A Data-Driven Screening Methodology Using Pre-Departure Conditions}},
  year         = {{2026}},
}