A higher‐level classification of the Pannonian and western Pontic steppe grasslands (Central and Eastern Europe)

Abstract

QuestionsWhat are the main floristic patterns in the Pannonian and western Pontic steppe grasslands? What are the diagnostic species of the major subdivisions of the class Festuco‐Brometea (temperate Euro‐Siberian dry and semi‐dry grasslands)?LocationCarpathian Basin (E Austria, SE Czech Republic, Slovakia, Hungary, Romania, Slovenia, N Croatia and N Serbia), Ukraine, S Poland and the Bryansk region of W Russia.MethodsWe applied a geographically stratified resampling to a large set of relevés containing at least one indicator species of steppe grasslands. The resulting data set of 17 993 relevés was classified using the TWINSPAN algorithm. We identified groups of clusters that corresponded to the class Festuco‐Brometea. After excluding relevés not belonging to our target class, we applied a consensus of three fidelity measures, also taking into account external knowledge, to establish the diagnostic species of the orders of the class. The original TWINSPAN divisions were revised on the basis of these diagnostic species.ResultsThe TWINSPAN classification revealed soil moisture as the most important environmental factor. Eight out of 16 TWINSPAN groups corresponded to Festuco‐Brometea. A total of 80, 32 and 58 species were accepted as diagnostic for the orders Brometalia erecti, Festucetalia valesiacae and Stipo‐Festucetalia pallentis, respectively. In the further subdivision of the orders, soil conditions, geographic distribution and altitude could be identified as factors driving the major floristic patterns.ConclusionsWe propose the following classification of the Festuco‐Brometea in our study area: (1) Brometalia erecti (semi‐dry grasslands) with Scabioso ochroleucae‐Poion angustifoliae (steppe meadows of the forest zone of E Europe) and Cirsio‐Brachypodion pinnati (meadow steppes on deep soils in the forest‐steppe zone of E Central and E Europe); (2) Festucetalia valesiacae (grass steppes) with Festucion valesiacae (grass steppes on less developed soils in the forest‐steppe zone of E Central and E Europe) and Stipion lessingianae (grass steppes in the steppe zone); (3) Stipo‐Festucetalia pallentis (rocky steppes) with Asplenio septentrionalis‐Festucion pallentis (rocky steppes on siliceous and intermediate soils), Bromo‐Festucion pallentis (thermophilous rocky steppes on calcareous soils), Diantho‐Seslerion (dealpine Sesleria caerulea grasslands of the Western Carpathians) and Seslerion rigidae (dealpine Sesleria rigida grasslands of the Romanian Carpathians).

GRIMP: A machine‐learning method for improving groups of discriminating species in expert systems for vegetation classification

“…Intuitively, one might think that higher numbers of species per group might result in more accurate classification. If so, it would be straightforward to use diagnostic species of individual vegetation types, defined using statistical indices such as the phi coefficient (Chytrý et al 2002, Tichý & Chytrý 2006 or constancy ratio (e.g., Willner et al, 2017) directly for discriminating among vegetation types in the expert system. We used such an approach as one variant of our tests ("From data") and found that it did not perform very well regarding the number of correctly classified plots.…”

“…Each vegetation type has one (and just one) discriminating species group, but single species can occur in more than one discriminating species group. The discriminating species group of a particular vegetation type is usually different from the diagnostic species group (e.g., Chytrý et al 2002;Willner et al, 2017) of that type, although a significant proportion of species is shared between these groups. Besides species, other taxa such as subspecies or species aggregates can be included in these groups, but for simplicity, all of them are called "species" throughout this paper.…”

“…Taking groups of statistically determined diagnostic species from classified vegetation‐plot datasets (e.g., Chytrý et al. ; Willner et al., ) is an option. However, such species groups may not yield the best discrimination of the existing vegetation types.…”

“…While this issue has recently been partly solved for bottom-up approaches (Bruelheide, 2016), it remains to be addressed for top-down approaches. Taking groups of statistically determined diagnostic species from classified | 7 vegetation-plot datasets (e.g., Chytrý et al 2002;Willner et al, 2017) is an option. However, such species groups may not yield the best discrimination of the existing vegetation types.…”

New syntaxa of steppe vegetation of the Kuialnyk Estuary (Odesa Region, Ukraine)

“…Нові синтаксони встановлювали шляхом порівняння фітоценонів один з одним та з описаними раніше синтаксонами (Korotchenko, Didukh, 1997;Krasova, Smetana, 1999;Moysiyenko et al, 2005;Korotchenko et al, 2009а, b;Vynokurov, 2014а, 2016Kolomiychuk, Vynokurov, 2016;Willner et al, 2017;etc.). Для порівняння використано 174 описи, результати відображені в синоптичній таблиці (табл.…”