Linear-shaped convective systems (LCSs), accompanied by band-shaped areas of heavy rainfall with a length of 50-300 km and a width of 20-50 km, are one of the extreme weather phenomena that can generate record-breaking heavy rainfall in Japan. Early warning information of LCSs considering the meteorological aspects of observation and numerical simulation is essential in ensuring lead time for flooding controls and evacuation. The current study examines the diagnostic approach of multifractality of 3D atmospheric field for LCSs using a cloud-resolving model (CRM) with 500-m horizontal resolution. Multifractal is defined as power laws of 3D spatial scale for water vapor flux (QVF) and turbulence kinetic energy (TKE). The CRM simulation of LCS at Kyoto on 15 July 2012 represented the process that convective clouds organized into a band-shaped cluster. QVF and TKE showed intense multifractality corresponding to convection initiation and the development of a back-building-like convective cluster. To define the period of the organization of LCS, we estimate the temporal variation of convective mass flux (CMF) which is the total volume of upward moisture flux and fractal dimension of updraft (FDU). When the isolated convective clouds initiated FDU showed a gradual increasing tendency. CMF and FDU increase as merging convective clouds and forming LCS and we defined this period as the organization of LCS. Our analyses revealed multifractal signals on QVF and TKE that appeared about 1-h before LCS was organized. The similar characteristics were also observed in the LCS event at Hiroshima on 20 August 2014. We conducted the ensemble forecast experiment of Kameoka LCS event to evaluate the robustness of precedent multifractal behavior. The composite analysis for 10 members with larger accumulated rainfall indicated increasing FDU and intense multifractal features on QVF and TKE before the organization of LCSs.