Sigma · Ekumene v4
Ekumene · 寰宇 —— 逐期动画Ekumene — Period-by-period viewer
这是一次完整运行的历史模拟。世界从公元前四千年的两千一百万人开始，粮食、人口、耕地、农业技术、政体、战争和瘟疫分别按自己的规则逐期演算，直到公元 1500 年。下面的动画每十期截取一帧，可切换八个图层，也可拖动时间轴查看它们的变化。 One completed run of a historical simulation. The world starts in 4000 BC with about twenty-one million people; food, population, cropland, farming technique, polities, war, and epidemics each advance by their own rules, period after period, through to AD 1500. The animation below samples every tenth period; switch between eight layers and drag the timeline to watch each one change.
这一版做了什么What changed in v4
此前版本中，只要格子里有可耕地，第一期就可以种地，因此美洲、撒哈拉以南非洲和两河流域在公元前四千年同时出现农民。这一版改为按照作物驯化年代开始农业。起点依据 ArchaeoGLOBE：这项数据集汇集全球考古学家对各地古代土地利用的评估。距今六千年时，农业已属“常见”以上的 24 个地区，共 9,570 格，从第一期就有农业；中美洲、萨赫勒、苏丹东部、印度南部、安第斯高原和北美东部这六处独立驯化中心，则要到各自的考古年代才在当地出现农业。其他地方的人只能采集渔猎，等农业从相邻格子传来，不能自行发明农业。 In earlier versions any cell with cultivable land could farm from the first period, so the Americas, sub-Saharan Africa, and Mesopotamia all had farmers in 4000 BC. This version starts agriculture by the dates of crop domestication. The starting map comes from ArchaeoGLOBE, a dataset that collects archaeologists' assessments of ancient land use around the world: the 24 regions where farming was already at least "common" six thousand years ago — 9,570 cells — farm from the first period, while six independent centers of domestication (Mesoamerica, the Sahel, the eastern Sudan, southern India, the Andean highlands, and eastern North America) appear in place at their archaeological dates. Everyone else lives by foraging until farming arrives from a neighboring cell; they cannot invent it themselves.
农业只从接壤的格子传入，不经过模型中的远距离商路和海路。定耕农业的扩散，依靠农民一代代迁往邻近土地，或依靠附近居民学会种地，并不是商人把作物带到千里之外。按这种方式传播，农业在欧洲的速度约为每年 0.9 公里，接近欧洲新石器时代扩散的实测速度（每年约一公里）。 Farming now spreads only between adjacent cells, never along the model's long-distance trade and sea routes, because settled agriculture spread as farmers moved onto neighboring land generation after generation, or as their neighbors took it up — not as merchants carried a crop a thousand kilometers. Spreading this way it advances across Europe at about 0.9 km a year, close to the measured pace of the European Neolithic expansion, roughly one kilometer a year.
瘟疫的起点也作了调整。天花和麻疹只能在人口足够多的地方起源：一个格子的人口超过某个数，病毒才能不断找到新的易感者，而不至于自行消失。按传播规律推算，天花的门槛是 6.7 万人，麻疹是 11 万人。超过门槛的格子每期有千分之一的概率成为起源地；千分之一这个概率没有出处，是三档试算中的中间一档。结核不需要大人群，也能长期潜伏在人体内，因此改为每格每期一千万分之一的起源概率；全世界平均约一千年出现一次，这个数同样没有出处。 Epidemics now start differently too. Smallpox and measles can arise only where population is large enough: a cell must hold more than a certain number of people for the virus to keep finding new susceptible hosts instead of burning out. Worked out from transmission theory, that number is 67,000 for smallpox and 110,000 for measles. Each cell above the threshold has a one-in-a-thousand chance per period of becoming an origin; that probability has no source and is the middle of three trial values. Tuberculosis can lie latent for years without a large population, so it instead gets an origin probability of one in ten million per cell per period — about once a millennium worldwide — a figure that likewise has no source.
战争算法改了一处。此前，一个国家的野战军可以出国进攻，却不参与保卫本国；被进攻的格子只有本格驻军，本国原有格子的驻军只有人口的千分之二，一万人的格子只有二十个兵。进攻方调来的是全国的野战军，上一版全程三十九万次进攻，赢了三十八万九千次。这一版让守方野战军也参与防守，按守方都城到被攻格子的距离折算兵力，算法与进攻方投送兵力相同；守方野战军作战也会减员，损失比例与进攻方对调使用，不另设参数。守方同一期在多处受到进攻时并不分兵，每一处都按全额野战军乘距离折扣计算，这是这一版仍然存在的缺口。 One change to war. Until now a state's field army could march out to attack but never came home to defend; a cell under attack had only its own garrison, and a home cell's garrison was two people per thousand — twenty soldiers in a cell of ten thousand — against an attacker bringing its whole field army. In the previous run 389,000 of 392,000 attacks succeeded. The defender's field army now joins the defense, discounted by the distance from the defender's capital to the attacked cell by the same formula that discounts an attacker's force, and it takes losses by the attacker's loss rates with the two sides swapped, so no new parameter is added. A defender attacked in several places in one period does not divide its army: each defense counts the whole field army times its distance discount, a gap this version still leaves open.
分工规则也改了一处。模型按野生食物能养活多少人来分配人口：养得起的人去采集，超出的人去种地。上一版里，天花把旧大陆人口压到这条线以下，所有人都被分配去采集，旧大陆耕地从公元前 3245 年的 75.9 万平方公里撂荒到前 995 年的 770 平方公里，此后一千五百年没有恢复，而任何一个格子都没有失去种地技术。历史上遭遇大瘟疫的农业社会并没有退回采集。这一版让上一期在种地的人，只要种地的人均所得不低于采集，就继续种地。 One change to how people divide between foraging and farming. The model sends to foraging as many people as wild food can feed and puts the rest on the land. In the previous run smallpox pushed the Old World below that line, everyone was sent to forage, and Old World cropland fell from 759,000 km² in 3245 BC to 770 km² by 995 BC and never came back in the next fifteen hundred years — although not one cell had lost the knowledge of farming. Farming societies struck by great epidemics did not go back to foraging. In this version, people who farmed last period keep farming as long as farming pays each of them at least as much as foraging would.
这一版的人口曲线The population curve
| 年份Year | 模型Model | HYDE | 比值Ratio |
|---|---|---|---|
| 3000 BC | 75.1M | 43M | 1.73× |
| 2000 BC | 40.1M | 71M | 0.57× |
| 1000 BC | 53.2M | 107M | 0.50× |
| AD 1 | 63.7M | 225M | 0.28× |
| AD 1000 | 89.7M | 314M | 0.29× |
| AD 1500 | 54.5M | 488M | 0.11× |
HYDE 是荷兰环境评估署的全球历史人口与土地利用数据库，这里只用于事后对照，没有参与任何取值。 HYDE is the Netherlands Environmental Assessment Agency's history database of the global environment; it is used here only for after-the-fact comparison and played no part in setting any parameter.
人口的起落How population rose and fell
模拟开始后的五百年里，有农业的地区人口迅速增长。公元前 3480 年，全世界人口达到 2.93 亿，其中 2.69 亿在旧大陆，而 HYDE 对公元前三千年全世界人口的估计只有 4,300 万。下一期，即公元前 3475 年，天花在尼罗河谷（北纬 29.0°、东经 30.9°）起源，随后传遍旧大陆。旧大陆人口在公元前 2495 年降到 1,850 万，全世界人口在公元前 2545 年降到最低的 3,380 万。与上一版不同，这一次天花始终没有从模拟中消失，此后三千多年一直在八千到两万六千个格子里流行，到公元 1500 年仍有 9,283 格。 In the first five hundred years, population in the farming regions grew very fast. By 3480 BC the world held 293 million people, 269 million of them in the Old World, while HYDE puts the whole world at 43 million in 3000 BC. The next period, 3475 BC, smallpox arose in the Nile valley (29.0° N, 30.9° E) and spread across the Old World. Old World population fell to 18.5 million by 2495 BC, and world population reached its low of 33.8 million in 2545 BC. Unlike the previous run, smallpox never disappeared from the simulation this time: for more than three thousand years it circulated in between eight thousand and twenty-six thousand cells, and 9,283 cells still had it in AD 1500.
美洲的农业要到各地的驯化年代才出现，天花也没有传过去。美洲人口从公元前 3495 年的 720 万增长到公元元年的 5,120 万、公元 1005 年的 8,010 万，此后降到公元 1500 年的 4,720 万；这次下降的原因尚未查明。公元 1000 年前后，全世界 8,970 万人里近九成在美洲。 In the Americas farming appears only at each region's date of domestication, and smallpox never crossed over. American population grew from 7.2 million in 3495 BC to 51.2 million in AD 1 and 80.1 million in AD 1005, then fell to 47.2 million by AD 1500; the cause of that fall has not yet been traced. Around AD 1000, nearly nine in ten of the world's 89.7 million people lived in the Americas.
旧大陆的农业为什么还是没有回来Why Old World farming still did not come back
这一版让已经在种地的人继续种地，但旧大陆农业仍然萎缩。旧大陆靠采集为生的人，公元前 3245 年只占 4%，公元前 2495 年占 41%，公元元年占 90%，公元 1500 年占 93%；耕地从公元前 3245 年的 81.9 万平方公里降到公元 1500 年的约一千平方公里。种地技术没有丢，末期旧大陆仍有 36,599 个有人居住的格子掌握定耕农业，其中九成以上按模型算法种地更划算：种地每人一年得到约 410 万大卡，采集约 59 万。可是这些格子里只有 2,418 个还有人种地。 This version lets people who already farm keep farming, yet Old World farming still retreated. The share of Old World people living by foraging was 4% in 3245 BC, 41% in 2495 BC, 90% in AD 1, and 93% in AD 1500; cropland fell from 819,000 km² in 3245 BC to about a thousand square kilometers by AD 1500. The knowledge was not lost: at the end, 36,599 inhabited Old World cells still know settled farming, and in more than nine in ten of them farming pays better by the model's own reckoning — about 4.1 million kilocalories a year per person, against about 0.59 million from foraging. Yet only 2,418 of those cells have anyone farming.
原因在分工规则。模型先按野生食物能养活多少人来分配：养得起的人去采集，超出的人才去种地；这一版加上的规则只能留住还在种地的人。一个格子的农民一旦消失，只要人口低于野生食物养得起的人数，就不会有人回去种地，哪怕种地所得是采集的七倍。这些格子里的农民最初怎样消失，尚未逐期追查。下一步应修改规则，使种地更划算时，人能够重新回到地里。 The cause lies in how the model divides people between the two. It first sends to foraging as many people as wild food can feed and puts only the rest on the land; the rule added in this version can only keep people who are still farming. Once a cell has lost its farmers, nobody goes back to the fields as long as its population stays below what wild food can feed — even when farming pays seven times as much. How those cells lost their farmers in the first place has not yet been traced period by period. The next change is to let people return to farming when it pays better.
战争War
守方野战军参与防守之后，守方第一次出现战损：全程守方损失 517 万人，攻方 145 万人。但模型选定的战事里，进攻方仍然几乎都赢，全程 44.1 万次进攻赢了 43.6 万次。原因在目标的选法：进攻方按胜率从高到低挑目标，被攻格子一半以上是无主地，或属于没有野战军的政体；即便是有主的目标，守方野战军按距离折算后也多半只有几人到二十来人。模型里的一次“战争”是一个政体在一期里攻打一个相邻格子，所以次数很多。政体太多、太小，也和叛乱使大国迅速解体有关，这一点还没有改。 With the defender's field army now joining the defense, defenders take losses for the first time: 5.17 million soldiers over the run, against 1.45 million for attackers. But the battles that get fought are still almost always won by the attacker — 436,000 of 441,000 attacks. The reason is how targets are chosen: an attacker picks targets in order of its chance of winning, more than half of the cells attacked are unclaimed or belong to a polity with no field army, and even owned targets usually face a defending army of a few to about twenty soldiers after the distance discount. A "war" in the model is one polity attacking one neighboring cell in one period, which is why the count looks so large. That polities stay so many and so small is also tied to rebellions breaking large states apart quickly, which has not yet been changed.
其他还没对上的地方Other things that do not yet match
麻疹在整个模拟中没有起源。结核在公元前 3435 年于婆罗洲（北纬 4.9°、东经 116.8°）起源，此后两千多年只停留在四百多个格子里，公元 500 年前后才扩散开，到公元 1500 年有 44,253 格。东南亚在公元前 3550 年、日本在公元前 3400 年就有了定耕农业，比考古记录早一千年以上，这一点尚未查明。 Measles never arose in the whole run. Tuberculosis arose on Borneo (4.9° N, 116.8° E) in 3435 BC, stayed confined to some four hundred cells for more than two thousand years, spread only around AD 500, and was present in 44,253 cells by AD 1500. Southeast Asia had settled farming by 3550 BC and Japan by 3400 BC, more than a thousand years earlier than the archaeological record; this has not yet been examined.
八个图层各是什么The eight layers
政体归属：每个国家一种颜色，深蓝底色的是无人统治的格。人口：每格人数按对数着色。耕地：每格熟地面积。采集者：每格依靠采集渔猎生活的人数，标题中给出他们占世界人口的比例。口粮比值：每人每天分得的口粮相对于温饱线的比例，1 是刚够吃，红色表示饥荒。城市：城内人口。记录手段：已有文字或其他记账手段的格。官僚：在编官吏按治下格摊算。 Polities: one hue per state; cells left in the dark blue base color are under no rule. Population: people per cell, shaded logarithmically. Cropland: cultivated area per cell. Foragers: people per cell living by hunting, gathering, and fishing, with their share of world population in the title. Food ratio: daily ration per person against the subsistence line, where 1 is just enough and red is famine. Cities: urban population. Record-keeping: cells that have writing or another means of keeping accounts. Bureaucracy: officeholders spread across the cells they govern.
怎么读它How to read it
这是模型，不是历史。它回答的是“给定这套规则，会出现什么样的世界”，不能回答“历史为什么是这样”。看动画时，应先看形态，不必盯住具体数字：农业从几个中心向外扩散；人口先增长，被天花压低；农业在旧大陆退缩，在美洲扩张；政体从无到有，又随人口起落。与真实历史不相符的地方，说明模型缺少某种机制，偏差本身就是线索。 This is a model, not history. It can answer what kind of world grows out of a given set of rules; it cannot answer why history went the way it did. What is worth watching is shape rather than exact numbers: farming spreading outward from a few centers; population rising and being cut down by smallpox; farming retreating in the Old World and spreading in the Americas; polities appearing from nothing and rising and falling with population. Where it fails to match the real record, that gap is itself useful — it points to a mechanism the model is missing.