IB地理 IA Fieldwork 采样方法 数据分析技巧
IB 地理内部评估（IA）要求你亲自完成一次田野调查，把课堂上的地理概念落到真实的地点、真实的仪器与真实的数据上。本文按 IB 地理 IA 的评分维度，拆解从研究问题设定、采样设计、数据收集到统计分析与评价的完整流程，并给出可复用的句式与常见失分点。
The IB Geography Internal Assessment asks you to design and carry out your own fieldwork, then turn raw observations into a defensible geographical argument. This guide covers question design, sampling strategy, data collection, statistical analysis and evaluation, with the practical detail that examiners reward.
一、IB地理 IA 的评分逻辑与字数边界
IB 地理 IA 的核心不是"我去了哪里"，而是"我如何用地理学的工具回答一个有地理意义的问题"。报告上限约 2500 字，SL 与 HL 的 IA 在最终成绩中占 20% 到 25%，因此每一个字都必须服务于论证。评分维度通常覆盖六个方面：研究问题与地理情境、调查方法、数据质量与处理、书面分析、结论、评价。前三项决定你的数据是否可信，后三项决定你的分析是否有地理深度。最常见的失分并不是数据不够多，而是研究问题太宽泛，导致数据根本无法回应问题。
The IB Geography IA is graded on the relationship between question, method and analysis, not on how exotic your field site is. The report is capped at roughly 2,500 words and contributes between 20 and 25 per cent of the final grade depending on level. The assessment criteria group into six strands: fieldwork question and geographical context, method of investigation, quality and treatment of information collected, written analysis, conclusion and evaluation. Candidates rarely lose marks for collecting too little data; they lose marks because an over-broad question never allowed the data to answer it.
二、研究问题与研究假设的设定
一个好的研究问题必须满足三个条件：可测量、有空间分布、能在有限时间内采集到足够样本。例如"河流下游的河床沉积物粒径是否随距离减小"比"某河流的环境质量如何"更适合 IA，因为前者可以用游标卡尺与皮尺直接测量，并能用散点图与斯皮尔曼等级相关检验。研究问题确定后，把它转化为一到两个可检验的假设或地理问题，并明确自变量与因变量。若研究涉及多个变量，务必分清主变量与控制变量，否则分析部分会退化成描述性罗列。
A workable fieldwork question must be measurable, spatially variable and answerable within the time and equipment you actually have. "Does bedload particle size decrease downstream along the River X" is far better than "How good is the environment of River X", because the first version can be measured with calipers and a tape and tested with a scatter graph and Spearman's rank correlation. Once the question is fixed, restate it as one or two testable hypotheses with explicit independent and dependent variables. Where several variables are involved, separate the main variable from the controlled ones, otherwise the analysis degenerates into a list of descriptions.
三、数据收集方法：初级数据与次级数据
IA 要求以初级数据为主体，次级数据用于提供背景与验证。初级数据包括实地测量、样方调查、问卷、访谈、照片与视频记录、环境质量指数评分。次级数据包括气象局记录、人口普查、政府统计年鉴、卫星影像与 GIS 图层、历史地图。一个高分的做法是让次级数据承担基准角色：例如用十年的气象数据说明采样当天降雨是否异常，从而判断数据是否具有代表性。所有设备必须记录型号与精度，例如 0.01 毫米精度的游标卡尺、精确到 1 米的皮尺，因为精度直接决定你能否讨论误差范围。
Primary data must dominate an IA; secondary data support and validate it. Primary sources include direct measurement, quadrat surveys, questionnaires, interviews, photographic records and scored environmental quality indices. Secondary sources include meteorological records, census and government statistics, satellite imagery, GIS layers and historical maps. A strong technique is to let secondary data act as a baseline: ten years of rainfall records tell you whether your sampling day was unusually wet and therefore unrepresentative. Record the model and precision of every instrument, such as a vernier caliper reading to 0.01 millimetres or a tape accurate to one metre, because precision sets the error you are allowed to discuss.
四、采样策略：随机、系统、分层与便捷采样
四种采样策略各有适用范围。随机采样用随机数生成坐标，避免主观选择，但在河流或街道这类线性环境中往往难以定位。系统采样按固定间隔取样，例如每 50 米设一个测点，适合沿河、沿坡或沿街的连续变化研究，操作最可靠。分层采样先按土地利用类型、坡度或街区划分地层，再在每层内取样，适合比较不同区域。便捷采样速度最快，但样本代表性最弱，只能在时间受限时使用，且必须在评价部分明确承认偏差。样本量方面，30 个观测值通常是统计检验的最低门槛；若研究涉及多个站点，每站至少需要 10 个读数。
Four sampling strategies cover most school fieldwork. Random sampling uses randomly generated coordinates and removes personal bias, but is hard to apply in linear environments such as a river channel or a single street. Systematic sampling takes readings at a fixed interval, for example every 50 metres, and is the most reliable option for studying continuous change along a river, slope or road. Stratified sampling first divides the site into strata by land use, gradient or neighbourhood and then samples within each stratum, which suits comparisons between areas. Opportunistic sampling is quickest but weakest, acceptable only under real time pressure and only if the bias is acknowledged in the evaluation. As a rule, thirty observations is the minimum for a statistical test, and where several sites are compared, each should yield at least ten readings.
五、数据呈现与图表选择
图表不是装饰，而是分析工具，选错图表会直接丢掉分析分。连续变化用折线图或散点图，类别比较用条形图，比例分布用饼图要谨慎，等级分布用分级设色地图或点状密度图。空间数据建议至少包含一张手工或 GIS 绘制的地图，标注采样点、比例尺、指北针与图例。若数据分层明显，可用三轴图或雷达图展示多变量关系。每个图都必须有编号、标题、单位与数据来源，并在正文中被引用与解释，切勿只贴图不分析。横断面剖面图是河流研究中的高频工具，它能同时展示河道形态与流速变化，比单纯列数据更能体现地貌理解。
Graphs are analytical instruments, not decoration, and the wrong choice directly costs analysis marks. Use line graphs and scatter graphs for continuous change, bar charts for categorical comparison, and graded shading or proportional symbols for spatial distribution; be cautious with pie charts because they hide distribution. Include at least one map showing sampling points with a scale, north arrow and legend, produced by hand or with GIS. Where data have clear layers, three-axis graphs or radar charts can display multivariate relationships. Every figure needs a number, caption, units and source, and must be referred to and interpreted in the text; a figure that is never discussed earns nothing. Cross-section profiles are especially useful in river studies because they show channel shape and velocity change more clearly than a table of numbers.
六、统计检验与地理分析工具
IB 地理常见的定量工具包括：斯皮尔曼等级相关系数，用于检验两个连续变量之间的单调关系；卡方检验，用于比较实际观测频数与期望频数；曼-惠特尼 U 检验，用于判断两组非正态分布数据是否来自同一总体；最近邻指数，用于判断聚落或商铺的空间分布是聚集、随机还是均匀；区位商，用于衡量某类功能在某区的专业化程度；辛普森多样性指数，用于比较生态或土地利用多样性。选择检验方法时必须说明理由，并写清零假设、显著性水平（通常取 0.05）、自由度与结论方向。不要只报一个数值，要把它翻译成地理语言，例如"该检验在 95% 置信水平下支持下游粒径减小的假设，说明河流搬运与磨蚀作用在下游占主导"。
The quantitative toolkit for IB Geography includes Spearman's rank correlation coefficient for monotonic relationships between two continuous variables, the chi-squared test for comparing observed and expected frequencies, the Mann-Whitney U test for comparing two non-normal groups, the nearest neighbour index for judging whether settlements or shops are clustered, random or regular, the location quotient for measuring functional specialisation, and Simpson's diversity index for comparing ecological or land-use diversity. Justify the test you choose, and state the null hypothesis, the significance level, usually 0.05, the degrees of freedom and the direction of the result. Never report a number alone: translate it into geographical language, such as "the test supports the downstream fining hypothesis at the 95 per cent level, indicating that fluvial transport and abrasion dominate in the lower course".
七、代表性与局限性讨论
评价部分要回答四个问题：样本是否有代表性、测量是否准确、方法在多大程度上可复制、误差可能来自哪里。常见的误差来源包括观测者偏差（例如对颜色分级的主观判断）、设备误差、天气与季节差异、样本量不足、时间限制导致的采样点分布不均。评价不是自我否定，而是展示方法学意识。高分写法是先指出局限，再说明该局限如何影响结论的强度，最后给出改进方案，例如"本次因退潮时间限制仅采集 24 个样本，低于统计检验建议的 30 个，因此斯皮尔曼系数只能作为趋势参考；后续研究应分两个潮周期采样以扩大样本量"。同时要区分准确度与精密度、可靠性与有效性这两组常被混淆的概念。
The evaluation must answer four questions: was the sample representative, were the measurements accurate, how reproducible was the method, and where did the error come from. Typical sources of error include observer bias such as subjective colour scoring, instrument error, weather and seasonal variation, insufficient sample size, and uneven distribution of sampling points caused by time limits. Evaluation is not self-criticism; it demonstrates methodological awareness. The high-scoring pattern is to identify the limitation, explain how it weakens your conclusion, and then propose a concrete improvement, for example "only 24 samples were collected because of the falling tide, below the recommended thirty, so the Spearman coefficient should be read as a trend only; a follow-up study should sample across two tidal cycles". Also keep the pairs accuracy versus precision and reliability versus validity clearly separated.
八、写作结构与提分要点
报告建议按"研究问题与地理背景、方法、数据呈现、分析、结论、评价"的顺序组织，每一部分用小标题标注，方便阅卷人对应评分维度。地理背景必须引用真实的位置信息、地质或气候条件、相关政策或理论模型，例如城市研究可引用同心圆模型或核心-边缘模型并说明其适用程度。分析部分是把数据与地理过程连接起来的段落，要使用专业术语：侵蚀、搬运、沉积、径流、渗透、梯度、可达性、集聚效应、绿地供给、土地利用冲突。字数控制在 2500 字以内，图表与引用不计入，但不要靠堆砌图注凑数。拿到题目后先画一张"研究问题到数据到检验方法"的逻辑链草图，再开始动笔，可以显著减少后期返工。
Structure the report as fieldwork question and geographical context, method, presentation of data, analysis, conclusion and evaluation, using subheadings so the examiner can map your text onto the criteria. The geographical context should cite real location data, geology or climate, and relevant policies or models: an urban study might apply the concentric zone or core-periphery model and justify how well it fits. The analysis paragraph is where data meet process, so use precise terminology such as erosion, transportation, deposition, runoff, infiltration, gradient, accessibility, agglomeration, green space provision and land-use conflict. Keep within 2,500 words, noting that figures and references are excluded, but do not pad with captions. Before writing, sketch a logic chain from question to data to test; it saves a great deal of rewriting later.
九、高频失分点清单
一、研究问题太宽泛，无法用一个变量检验。二、方法描述缺少设备与步骤，别人无法复制。三、采样点集中在一处，缺乏空间代表性。四、只贴图表不做解释，或图表没有单位与来源。五、统计检验只报数值不报含义，缺少零假设与显著性水平。六、评价部分只写时间不够，没有说明影响与改进方案。七、字数超限，或把地理背景写成旅游介绍。避开这七点，IB 地理 IA 的分数通常会稳定在较高区间。若你在采样设计或统计检验的选择上仍有疑问，可以把研究问题和数据表发来，我们一起把逻辑链补完整。
The most frequent losses come from a question too broad to test, a method section lacking equipment and step detail so it cannot be replicated, sampling points clustered in one place, figures presented without interpretation or units, statistical results quoted without a null hypothesis or significance level, an evaluation that blames a lack of time without explaining the consequence or a fix, and prose that drifts into a travel description instead of geography. Avoiding these seven patterns usually lifts an IA into the upper band.
CASIO fx-991CN X 科学计算器 考试必备¥168去京东购买京东精选需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导
或截图后用「扫一扫 → 相册」选择本图