The Shannon information function (H) has been extensively used in ecology as a statistic of species diversity. Yet, the use of Shannon diversity index has also been criticized, mainly because of its ambiguous ecological interpretation and because of its relatively great sensitivity to the relative abundances of species in the community. In my opinion, the major shortcoming of the traditional perspective (on the possible relation of species diversity with information theory) is that species need for an external receiver (the scientist or ecologist) to exist and transmit information. Because organisms are self-catalized replicating structures that can transmit genotypic information to offspring, it should be evident that any single species has two possible states or alternatives: to be or not to be. In other words, species have no need for an external receiver since they are their own receivers. Therefore, the amount of biological information (at the species scale) in a community with one only species would be $$ { \log }_{2} 2^{1} = 1 $$ species, and not $$ { \log }_{2} 1 = 0 $$bits as in the traditional perspective. Moreover, species diversity appears to be a monotonic increasing function of $$ { \log }_{2} 2^{{\text{S}}} $$ (or S) when all species are equally probable (S being species richness), and not a function of $$ { \log }_{2} {\text{ S}} $$ as in the traditional perspective. To avoid the noted shortcoming, we could use 2H (instead of H) for calculating species diversity and species evenness (= 2H/S). However, owing to the relatively great sensitivity of H to the relative abundances of species in the community, the value of species dominance (= 1 − 2H/S) is unreasonably high when differences between dominant and subordinate species are considerable, thereby lowering the value of species evenness and diversity. This unsatisfactory behaviour is even more evident for Simpson index and related algorithms. I propose the use of other statistics for a better analysis of community structure, their relationship being: species evenness + species dominance = 1; species diversity × species uniformity = 1; and species diversity = species richness × species evenness.

中文翻译：

---

重新审视物种多样性与信息论之间的关系

香农信息函数 (H) 已被广泛用于生态学中作为物种多样性的统计数据。然而，香农多样性指数的使用也受到批评，主要是因为它的生态解释模棱两可，而且它对群落中物种的相对丰度相对敏感。在我看来，传统观点（物种多样性与信息论之间可能存在的关系）的主要缺点是物种需要外部接收者（科学家或生态学家）来存在和传递信息。因为生物体是自我催化的复制结构，可以将基因型信息传递给后代，很明显，任何单个物种都有两种可能的状态或选择：存在或不存在。换句话说，物种不需要外部接收器，因为它们是自己的接收器。因此，一个只有一个物种的群落中的生物信息量（在物种尺度上）将是 $$ { \log }_{2} 2^{1} = 1 $$ 个物种，而不是 $$ { \log }_{2} 1 = 0 $$bits 与传统观点一样。此外，当所有物种的概率相等时，物种多样性似乎是 $$ { \log }_{2} 2^{{\text{S}}} $$（或 S）的单调递增函数（S 是物种丰富度） )，而不是传统观点中 $$ { \log }_{2} {\text{ S}} $$ 的函数。为了避免上述缺点，我们可以使用 2H（而不是 H）来计算物种多样性和物种均匀度（= 2H/S）。然而，由于 H 对群落中物种的相对丰度的相对较大的敏感性，当优势种和从属种之间的差异相当大时，物种优势度(= 1 − 2H/S) 值过高，从而降低了物种均匀度和多样性的价值。这种令人不满意的行为对于辛普森指数和相关算法来说更加明显。我建议使用其他统计数据来更好地分析群落结构，它们的关系是：物种均匀度 + 物种优势 = 1；物种多样性 × 物种一致性 = 1；物种多样性=物种丰富度×物种均匀度。我建议使用其他统计数据来更好地分析群落结构，它们的关系是：物种均匀度 + 物种优势 = 1；物种多样性 × 物种一致性 = 1；物种多样性=物种丰富度×物种均匀度。我建议使用其他统计数据来更好地分析群落结构，它们的关系是：物种均匀度 + 物种优势 = 1；物种多样性 × 物种一致性 = 1；物种多样性=物种丰富度×物种均匀度。