We present a workflow to conduct a full characterization of medium to coarse-grained igneous rocks, using portable, non-invasive, and reproducible approaches. This includes: (i) Image Analysis (IA) to quantify mineral phase proportions, grain size distribution using the Weka trainable machine learning algorithm. (ii) Portable X-ray fluorescence spectrometer (PXRF, Bruker Tracer IV) to quantify the whole-rock's chemical composition. For this purpose, a specific calibration method dedicated to igneous rocks using the open-source CloudCal app was developed. It was then validated for several key elements (Si, Al, K, Ti, Ca, Fe, Mn, Sr, Ga, Ba, Rb, Zn, Nb, Zr, and Y) by analyzing certified standard reference igneous rocks. (iii) Portable Magnetic Susceptibilimeter (pMS, Bartington MS2K system) to constrain the mineralogical contribution of the samples. The operational conditions for these three methods were tested and optimized by analyzing five unprepared surfaces of igneous rocks ranging from a coarse-grained alkaline granite to a fine-grained porphyric diorite and hence, covering variable grain sizes, mineralogical contents, and whole-rock geochemical compositions. For pMS and PXRF tools, one hundred analyses were conducted as a 10 cm × 10 cm square grid on each sample. Bootstrap analysis was implemented to establish the best grid size sampling to reach an optimized reproducibility of the whole-rock signature. For PXRF analysis, averaged compositions were compared to PXRF analysis on press-pellets and laboratory WD-XRF analysis on fused disk and solution ICP-OES (for major) and solution-ICPMS (for trace element concentrations). Ultimately, this workflow was applied in the field on granitoids from three Roman quarrying sites in the Lavezzi archipelago (southern Corsica) and tested against the Bonifacio granitic War Memorial, for which its provenance is established. Our results confirm this information and open the door to geoarchaeological provenance studies with a high spatial resolution.

我们提出了使用便携式，非侵入性和可再现方法对中至粗粒火成岩进行全面表征的工作流程。这包括：（i）图像分析（IA），使用Weka可训练的机器学习算法来量化矿物相的比例，粒度分布。（ii）便携式X射线荧光光谱仪（PXRF，Bruker Tracer IV）用于量化整个岩石的化学成分。为此，使用开源CloudCal应用程序开发了专用于火成岩的特定校准方法。然后，通过分析认证的标准参考火成岩，对几种关键元素（Si，Al，K，Ti，Ca，Fe，Mn，Sr，Ga，Ba，Rb，Zn，Nb，Zr和Y）进行验证。（iii）便携式磁化计（pMS，Bartington MS2K系统）可限制样品的矿物学作用。通过分析五个未准备好的火成岩表面（从粗粒碱性花岗岩到细粒斑岩闪长岩）来测试和优化了这三种方法的操作条件，从而涵盖了可变的粒度，矿物含量和全岩石地球化学成分。对于pMS和PXRF工具，在每个样品上以10 cm×10 cm的正方形网格进行一百次分析。进行了自举分析，以建立最佳的网格大小采样，以实现整个岩石特征码的最佳重现性。对于PXRF分析，将平均组成与压片上的PXRF分析进行比较，将熔融盘和溶液ICP-OES（用于主溶液）和溶液ICPMS（用于痕量元素浓度）与实验室WD-XRF分析进行比较。最终，该工作流程在Lavezzi群岛（科西嘉南部）的三个罗马采石场的花岗岩上应用于野外，并针对Bonifacio花岗岩战争纪念馆进行了测试，该纪念馆的来历得到证实。我们的研究结果证实了这一信息，并为高空间分辨率的地理考古物证研究打开了大门。为此，它的来源已经确定。我们的研究结果证实了这一信息，并为高空间分辨率的地理考古物证研究打开了大门。为此，它的来源已经确定。我们的研究结果证实了这一信息，并为高空间分辨率的地理考古物证研究打开了大门。