@article{fdi:010079135,
  title = {{Q}uantifying the benefits of nonlinear methods for global statistical hindcasts of tropical cyclones intensity},
  author = {{N}eetu, {S}. and {L}engaigne, {M}atthieu and {V}ialard, {J}{\'e}r{\^o}me and {M}angeas, {M}organ and {M}enk{\`e}s, {C}hristophe and {S}uresh, {I}. and {L}eloup, {J}. and {K}naff, {J}. {A}.},
  editor = {},
  language = {{ENG}},
  abstract = {{W}hile tropical cyclone ({TC}) track forecasts have become increasingly accurate over recent decades, intensity forecasts from both numerical models and statistical schemes have been trailing behind. {M}ost operational statistical-dynamical forecasts of {TC} intensity use linear regression to relate the initial {TC} characteristics and most relevant large-scale environmental parameters along the {TC} track to the {TC} intensification rate. {Y}et, many physical processes involved in {TC} intensification are nonlinear, hence potentially hindering the skill of those linear schemes. {H}ere, we develop two nonlinear {TC} intensity hindcast schemes, for the first time globally. {T}hese schemes are based on either support vector machine ({SVM}) or artificial neural network ({ANN}) algorithms. {C}ontrary to linear schemes, which perform slightly better when trained individually over each {TC} basin, nonlinear methods perform best when trained globally. {G}lobally trained nonlinear schemes improve {TC} intensity hindcasts relative to regionally trained linear schemes in all {TC}-prone basins, especially the {SVM} scheme for which this improvement reaches similar to 10% globally. {T}he {SVM} scheme, in particular, partially corrects the tendency of the linear scheme to underperform for moderate intensity (category 2 and less on the {S}affir-{S}impson scale) and decaying {TC}s. {A}lthough the {TC} intensity hindcast skill improvements described above are an upper limit of what could be achieved operationally (when using forecasted {TC} tracks and environmental parameters), it is comparable to that achieved by operational forecasts over the last 20 years. {T}his improvement is sufficiently large to motivate more testing of nonlinear methods for statistical {TC} intensity prediction at operational centers.},
  keywords = {{H}urricanes ; typhoons ; {S}tatistical forecasting ; {N}eural networks ; {R}egression ; {S}upport vector machines},
  booktitle = {},
  journal = {{W}eather and {F}orecasting},
  volume = {35},
  numero = {3},
  pages = {807--820},
  ISSN = {0882-8156},
  year = {2020},
  DOI = {10.1175/waf-d-19-0163.1},
  URL = {https://www.documentation.ird.fr/hor/fdi:010079135},
}