The spatial and temporal reconstruction of a medieval moat ecosystem

ABSTRACT Moats and other historical water features had great importance for past societies. The functioning of these ecosystems can now only be retrieved through palaeoecological studies.

Here we aimed to reconstruct the history of a stronghold’s moat during its period of operation. Our spatio-temporal approach allowed mapping of the habitat changes within a medieval moat for

the first time. Using data from four cores of organic deposits taken within the moat system, we describe ecological states of the moat based on subfossil Chironomidae and Ceratopogonidae

assemblages. We found that over half (57%) of the identified dipteran taxa were indicative of one of the following ecological states: limnetic conditions with or without periodic water

inflow, or marshy conditions. Samples representing conditions unfavourable for aquatic insects were grouped in a separate cluster. Analyses revealed that the spatio-temporal distribution of

midge assemblages depended mostly on depth differences and freshwater supply from an artificial channel. Paludification and terrestrialization did not happen simultaneously across the moat

system, being greatly influenced by human activity. The results presented here demonstrate the importance of a multi-aspect approach in environmental archaeology, focusing not only on the

human environment, but also on the complex ecology of the past ecosystems. SIMILAR CONTENT BEING VIEWED BY OTHERS THRESHOLDS FOR THE PRESENCE OF GLACIAL MEGAFAUNA IN CENTRAL EUROPE DURING

THE LAST 60,000 YEARS Article Open access 21 November 2022 COLLAPSE OF THE MAMMOTH-STEPPE IN CENTRAL YUKON AS REVEALED BY ANCIENT ENVIRONMENTAL DNA Article Open access 08 December 2021

FOOTPRINT BEDS RECORD HOLOCENE DECLINE IN LARGE MAMMAL DIVERSITY ON THE IRISH SEA COAST OF BRITAIN Article 26 September 2022 INTRODUCTION The diversity of habitats and of the mutual

connections between organisms and their environment can now be studied in a wide variety of ecosystems, including historical features no longer present. Moats were artificial bodies of water

providing defence for inhabited strongholds in medieval times. Ecological studies of moats can therefore provide a window into historical influences of human activity on natural ecosystems,

broadening our knowledge of the functioning of such features, of past human societies, and of our coexistence with (seemingly distant) animal communities. Ecological studies can be carried

out on moats that have either been preserved from historical times (e.g.1,2) or have been recently reconstructed. However, these modern moats do not perform the same function today as in the

Middle Ages or Early Modern Period3, and their aquatic ecosystems may therefore differ. The medieval stronghold’s inhabitants influenced water conditions by changing its trophic state,

creating new habitats by placing construction elements, modifying its hydromorphology and sedimentation of defined types of water deposits, and their enrichment with selected metals

(see4,5). What were the nature, scale, and consequences of such activities? The only way to get a clear window into those past ecosystems is through palaeoecological studies. Geochemical

studies of historical layers (such as core samples) conducted so far have in many cases confirmed their indicator role in environmental archeology6. The clearly higher Pb and Cu

concentrations in the sediments of medieval moats have been used to determine the course of historical watercourses. However, the grain and mineral composition of sediments influences their

susceptibility to metal sorption7. Therefore, geochemical mapping of archaeological sites, aiming to assess the distribution and characteristics of metallic pollutants in the environment,

should take into account the location of the studied cores. The chosen location of cores also influences palaeoecological studies. Typically, palaeoecological studies are based on one core

of sediment, taken from the deepest (mostly central) part of the (palaeo)lake, on the assumption that the subfossil remains of the organisms once living in different parts of the water body

are passively transported over time and deposited into the deepest area of the lake bottom8. This way, representative biotic records can be obtained with only one drilling, reducing cost and

time of analyses. However, in the case of macrofossils, diatoms, or cladocerans, littoral taxa can be underrepresented in sequences derived from one core9,10,11. Analogously, chironomid

head capsules accumulate mostly near the habitat of the larvae, and offshore transport is primarily observed in the less well preserved early instars12, which may produce some biases.

Generally, if the subfossil record is expected to adequately represent the environment of sedimentation, and track particular events, multiple coring within the (palaeo)lake or mire is

needed8,13. In environmental archaeology, it is recommended to take at least two cores for palaeoecological analyses—one from the studied site (on-site profile) and one from an area

separated from direct human activity (off-site) (e.g.14,15). This allows examination of both local and regional vegetation history, and also of environmental information, such as past water

level or temperature changes. Due to differences in deposition between aquatic and terrestrial systems, “wet sites” or “wet features” (such as a moat) allow examination with use of a wider

range of ecological proxy analyses, based on organic deposits within cores sampled directly from archaeological trenches. Our on-site palaeoecological investigations were an integral part of

an archaeological study of a stronghold’s moat system in Rozprza (central Poland; see ‘Study system’ section, below)5,16. The main sediment core from the studied moat has already been

investigated in detail (see5). The results indicated that when the stronghold was inhabited (14th-15th century AD), the moat was infilled with shallow water with possible episodes of rinsing

through an artificial channel from the nearby river. Changes in water trophic state were tracked using chironomid-based models, and the Chironomidae-inferred summer temperatures of the Late

Vistulian came from the nearby palaeomeander17. The Holocene chironomid sequence in the palaeomeander core was too low in abundance for quantitative reconstructions. However, even profiles

with a depauperate biotic record can be used to obtain some ecological information, especially if treated simultaneously with other analyses. Therefore, we took additional cores from

potentially different habitats (cf.18) within the moat, in order to obtain a more complete picture of environmental changes and human impact on the ecosystem in both time and space. Though

chironomids have proven to be useful in environmental archaeology (e.g.19,20,21,22,23), they have not previously been used for spatial reconstructions of artificial features. Moats, barays

and other anthropogenic ponds have more limited potential to accumulate most of the subfossil remains in one point than lakes, mostly because of the specific basin morphology. Therefore, a

carefully constructed spatial approach is required, taking into consideration habitat mosaics within the moat ecosystem. Midge larvae (Diptera: Chironomidae and Ceratopogonidae), especially

Chironomidae, represent an ideal proxy for spatial reconstruction of artificial features. They are ecologically diverse, sensitive to environmental changes, and indicative of particular

ecosystem conditions and processes24,25. Therefore, their distribution in shallow reservoirs with complex morphometry may be diversified and earlier studies based on one sediment profile5

need to be complemented in order to cover the spatial aspect. Our study addresses the general problem of lack of replicability in the environmental sciences. The issue is particularly

visible when deriving data from remote places or from the past (e.g. in geology, oceanography, palaeontology, archaeology). In palaeoecology, replicability is applied when studying and

comparing the same phenomenon/event (e.g. Younger Dryas) in different sites. However, few studies replicate from the same locality, collecting cores several metres apart. On the few

occasions that this small-scale replication has been carried out, important variation has been identified (e.g.26). Considering these issues, our main goal was to reconstruct changes in

habitat distribution across the stronghold’s moat system over time. We aimed to examine: (1) the extent to which the Chironomidae and Ceratopogonidae subfossils from the different cores

reflected differences in past community structure and environmental conditions in different parts of the moat (spatial aspect), and (2) the extent to which midge communities changed

simultaneously and/or evenly throughout the moat basin (temporal/time aspect). We hypothesised that even in such a small and shallow water body as a moat, the differences in midge community

structure between cores and time periods would be significant and, thus, the habitat mosaic and other aspects of stronghold functioning could be more fully reconstructed using multiple

cores. STUDY SITE Rozprza is located in central Poland, about 60 km south of Łódź in the Piotrków Plain. The study site is situated in the middle reach of the Luciąża River valley, a

tributary of the Pilica River in the Vistula River basin (Fig. 1). Nowadays, the late medieval stronghold remnants with their moat system are situated in an area covered by fields and

meadows between the Rozprza and Łochyńsko villages (51° 18′ 07″ N; 19° 40′ 04″ E). The poorly preserved traces of moats, ramparts and baileys are however still visible in the field (Fig. 1C)

and on Digital Elevation Models (DEMs). The study site is located on the valley floor with regulated channels of the Luciąża, Rajska and Bogdanówka rivers, as well as a dense network of

drainage canals (Fig. 1B). The studied fortress was established on the Plenivistulian fluvial terrace remnant, in the area of the widely spread valley floor. Such a location of the

motte-and-bailey was reasonable for defensive reasons, as the sandy terrace remnant was protected by the surrounding swampy areas within the valley floor16,27. However, the hillock of the

terrace remnant occupied by the motte in the late Middle Ages was very low (up to 1 m). The late medieval motte-and-bailey timber castle at Rozprza was built about 1330 AD and replaced an

earlier timber and earth ringfort of unclear chronology (between 11th and 13th century AD). Motte-and-bailey castles were common in western Europe already in the 11th century28 but

introduced to Poland much later, in the 13th century29. In the 14th and 15th centuries AD it was one of the most popular types of rural noble residences. The main moat of the Rozprza

motte-and-bailey was established ca. 1330 AD and was later filled with organic (gyttja and peat) and partially inorganic deposits containing rich remains of wood (Fig. 2, Supplementary Fig.

S1B–D). The fill of the main moat was the subject of a detailed palaeoenvironmental study by Kittel et al.5. The accumulation of overbank silty sandy organic mud took place within the moat

ditch system as late as in the 18th or 19th c. AD5,16. The main moat had a width of 17–21 m and a trapezoidal cross-section with a depth of 0.5 m, up to ca. 1.3 m. Wooden constructions were

recorded near the inner slope of the moat ditch—in the form of a palisade created by two rows of vertical, sharpened wooden poles, and horizontal beams lying behind them. Those constructions

were covered with thick layers of slope deposits (sand with organic mud). Many large chunks of wood (branches and boughs) were recorded in peat and organic mud of the upper unit of the moat

fill close to the inner slope of the main moat16. The second moat had a width of about 11 m and a depth of ca. 0.5 m. In the trapezoid cross-section it has a flat bottom. The fill of the

moat consisted of sand with organic admixtures at the very bottom, limnic deposits, peat with silts, sands and organic mud, and also slope wash deposits on the moat slopes. Generally, these

sediments are characterised by a massive structure, dark or grey-brown uniform colour, resulting from a strong concentration of dispersed organic matter and iron compounds (Supplementary

Fig. S1A). The chronology of the feature formation has been estimated to 1485–1634 AD based on 14C data (Table 1). It demonstrates that the secondary moat was built probably in the early

16th century AD. However, the analysis of most archaeological small finds obtained during field works, mostly pottery fragments, estimated their age to the period between 14th and a mid-15th

century AD. Evidence of human activity on the stronghold in the 16th century AD is also very limited. Therefore, an establishment of the second moat in the 15th century AD cannot be

excluded (see16). RESULTS AND INTERPRETATION CHRONOLOGY OF MOAT DEPOSIT ACCUMULATION In total six radiocarbon dates were used for the construction of the age-depth model for the RP W3(2)

core from the deepest studied part of the main moat (Table 1, Supplementary Table S1)5,16. Based on dendrochronological data from a wooden fragment from the moat’s bottom, establishment of

the main moat was defined to ca. 1330 AD. In the early phase, the moat was filled with gyttja. In the first half of the 16th century AD, a sedentation of peat began. In the early 18th c. AD

the peat was covered with overbank alluvia (Fig. 2). A comparable pattern of evolution of the main moat was reconstructed based on the age-depth model for the RP W3(4) core (Table 1,

Supplementary Table S2). The lacustrine deposition was replaced by peat sedentation in the mid-15th c. AD. The accumulation of overbank alluvia may have been initiated in the first decades

of the 18th c. AD. The fill of the western part of the moat was covered in 1944 AD by an embankment from the destroyed stronghold mound5. The chronology of the main moat fills in the RP W1

core confirms an establishment of this moat in the 1st half of the 14th c. AD. Moreover, the eastern part of the wet defensive system was filled with peat from its beginning up to 16th c. AD

(Table 1, Supplementary Table S3). The upper part of the moat fill in the RP W1 area was probably removed during melioration works in 20th c. AD. The absolute chronology of the fill of the

secondary moat, studied in the RP F2 core, demonstrates that this additional defensive ditch has been established most probably in the first half of the 16th c. or possibly in the late 15th

c. AD (Table 1, Supplementary Table S4). From 18th c. AD, an effect of flooding is visible, recorded by sandy admixtures in organic deposits of the moat fill (cf.16). DIPTERANS AND

ENVIRONMENT: IDENTIFICATION OF THE MOAT PHASES The samples from depths of 45-39 cm in RP W3(4) core and the samples from depths of 67 and 49-21 cm in RP F2 core were empty, i.e. without any

taxon present. The self-organising map (SOM) allowed clusters of non-empty core samples with similar community composition to be produced. The taxa significantly associated with them were

then identified with Indicator Species Analysis. Two main clusters were distinguished in the output layer of the SOM: X and Y, comprising the respective pairs of subclusters: X1 and X2, and

Y1 and Y2 (Fig. 3). The subclusters were ordered according to the gradient observed in the number of indicator species (Fig. 4). Subcluster X1 represents unfavourable conditions for

Chironomidae development, mostly overbank deposits (see Fig. 2). It contains surface samples from RP W3 cores (from 51 cm depth in W3(2) and from 55 cm in the W3(4) profile), samples from 91

to 87 cm of W3(2) core depth, 85–83 cm and 73 cm of W3(4) core depth, most samples from RP W1 core, and the whole sequence of the non-empty samples from the second moat (RP F2 core).

Subcluster X2 included samples from 83 to 55 cm of RP W3(2) core, 75 cm and 63–57 cm of RP W3(4) core, as well as one sample (29 cm depth) from RP W1 core. They were associated with high

organic matter (OM) content (mean 59.9%) and slightly acidic (mean pH = 6.6), probably telmatic (marshy) conditions (Fig. 2). Cluster Y represents limnetic conditions dominated by fully

aquatic midges and low values (i.e. < 20) of TOC/N ratio for deposits. The latter suggests that organic matter mainly came from algal phytoplankton49,81. Subcluster Y2 reflects habitat

with higher detrital (K, Mg, Ca) and sulphide (Cu, Fe) element concentrations, grouping bottom samples from the RP W3(2) core up to 95 cm and samples of 39–37 cm depth from the RP W1 core

(Fig. 3). Samples grouped in the subcluster Y1 (81–77 cm and 71–65 cm core RP W3(4) depth, sample from 21 cm core RP W1 depth) are associated with lower sulphide element content (Fig. 3). A

total of 55 (57%) dipteran taxa were significantly associated with a certain subcluster, i.e. they were indicators of its respective environmental conditions (Fig. 4). Among these, 24

exhibited IndVals significant at _p_ ≤ 0.001, 20 at _p_ ≤ 0.01, and 11 at _p_ ≤ 0.05. An upward trend was observed in the number of such taxa for subclusters in the order X1, X2, Y1, Y2. No

palaeoindicator was significantly associated with subcluster X1, eight taxa were significantly associated with X2, 19 with Y1 and 28 with Y2. Therefore, this order of subclusters corresponds

to increasingly favourable conditions for development of a rich biota. The most indicative (at _p_ < 0.001) morphotypes for subcluster X2 were _Limnophyes-Paralimnophyes_ and

_Parametriocnemus-Paraphaenocladius_, which are associated with the semi terrestrial habitats with slightly acidic water30. Ceratopogonid species grouped as _Dasyhelea_-type seem to have

similar preferences31, while Chironomini taxa linked to X2 prefer shallow, muddy water bodies and can occur in seasonal surface water. Chironomids associated with Y1 were mostly typical of

warm, productive, littoral habitats, and many of them are phytophilous (e.g. _Glyptotendipes pallens-_type, _Lauterborniella_). However, also associated with this subcluster were _Tanytarsus

lugens_-type and _Paratanytarsus austriacus_-type, often recorded in cold, oligotrophic conditions. Morphotypes significant to subcluster Y2 included both taxa associated with warm,

eutrophic stagnant water (e.g. _Micropsectra pallidula_-type, _Cladotanytarsus mancus_-type, _Cryptochironomus_) and those preferring meso- and oligotrophic conditions (e.g. _Psectrocladius

barbatipes_-type and _Bezzia_-type). Many of them, such as _Zavreliella_ and _Polypedilum sordens_-type are associated with macrophytes. Moreover, several chironomids associated with running

water (such as _Nanocladius rectinervis_-type, _Corynoneura coronata_-type and _Psectrotanypus varius_) were recorded with a significant IndVal in this subcluster. This differentiation is

confirmed by the results of the chemical composition filling from the upper part of the RP W3(2) core (Fig. 2), because the rich organic sediments (OM even above 90%) are covered by acidic

deposits with organic matter content below 7% and very variable concentration of lithophilic elements (for example K range 0.17–2 mg/g). The changes in time of sorption capacity were

probably caused by changes in the porosity of the sediments that accumulated in the moat. This feature is the result of the difference between natural and dry bulk density and it is

particularly modified by the content of very fine fraction in the sediments32. In turn, the increased abundance in nutrients results from a high proportion of the clay fraction, which in the

sediments from RP W3(2) often exceeds 3%, with a maximum of 6.45% (Fig. 2). These features of the biogenic accumulation environment influenced conditions for the development of vegetation

and chironomids. Canonical correspondence analysis (CCA) was carried out to detect midge (Chironomidae and Ceratopogonidae) geochemical signal correlations. Axis 1 (Ax1) explained 11.3% and

Axis 2 (Ax2) 4.2% of species data variance among individual core samples. For the species-environment relationship variance, 42.1% was explained by Ax1 and 16.0% by Ax2. The analysis (Fig. 

5) demonstrated that pH, Ca, Pb, Fe and organic matter (_p_ < 0.01), as well as Cu and K (_p_ < 0.05), were significant in shaping midge assemblages in the moat, with 6.4% of the

variance explained by pH, 4.6% by Ca, 4.2% by Pb, and 3.0% by Fe. Organic matter (OM) and Cu both explained 2.3%, while K explained 1.5% of the total variance. Pb was positively correlated

and pH negatively correlated with Ax1. The rest of the variables were positively correlated with Ax2. The samples grouped in the X1 subcluster generally represented conditions unfavourable

for aquatic biota. Many of them were characterised by relatively high Pb, probably reflecting increased denudation processes after stronghold abandonment and increased flooding activity in

the 18th–19th centuries AD (cf.5). This series describes monofraction of the mineral admixture (share of the sand mainly ranges between 70–90% and Mz for 70% number of samples is 1.6–2.2

_phi;_ Fig. 2). According to Kittel et al.21, in the absence of a clear boundary between individual layers, identification of flood activity should include changes of colour sediments,

caused by admixture of decomposed and diffused organic matter (Supplementary Fig. S1). In our case, organic matter values varied little among X1 subcluster samples (mean for this section

29.3%) and corresponded with light-grey horizon (overbank organic mud and overbank sandy organic mud _vide_:5; Fig. 2) a dozen cm thick. The overbank deposits accumulated in rapidly changing

oxygenation of water. Such conditions stimulated microbial decomposition of organic matter and allowed accumulation of sulphide elements, such as Cu, Mn and Fe (Fig. 2). The taxa typical of

the telmatic phase of the moat (X2) were associated with low pH. Moreover, _Limnophyes-Paralimnophyes, Parametriocnemus-Paraphaenocladius_ and _Polypedilum sordens_-type prefer habitats

with high organic matter and iron compounds content. However, among samples classified to the X2 cluster, those from RP W3(2) core were more related to acidic conditions than those from RP

W3(4). Generally more alkaline conditions are preferred by the chironomids indicative of the limnetic stage of the moat, in particular to subcluster Y2 (Fig. 5). Those taxa (e.g.

_Zavreliella_ and _Cricotopus bicinctus_-type) prefer habitats with high Ca and K values. Several phytophilous taxa (such as _Paratanytarsus penicillatus_-type, _Corynoneura coronata_-type

and _C. arctica_-type) seemed to be more associated with sulphide elements (Cu, Zn), than with pH level. Alkaline conditions are also important to several taxa indicative of the Y1

subcluster (like _Tanytarsus pallidicornis_-type 2), but unlike Y2, these taxa are associated with low element levels (Fig. 5). The occurrence of allochthonic mineral matter with variable

grain-size parameters (Supplementary Fig. S2) in the Y1 samples may be responsible for the increased lithophilic elements (mainly Mg and K) content in deposits (Fig. 2). The high

concentrations of Ca (often above 60 mg/g) (Fig. 2) are documented in the deposition environment of the hypergenic zone in Central Europe (cf.33). The intensive chemical denudation and

leaching of mineral substrate of variable origins were confirmed in the catchment of Luciąża River valley, the surface geology and mineralogy of which was documented in detail by

Wachecka-Kotkowska34. Moreover, the documented sediment structures and textures (Fig. 2, Supplementary Figs. S1 and S2) prove the supply of mineral matter in two different accumulation

environments. The first one is related to upper flow regime (mainly samples from RP F2 and RP W3(4)), the second one in the lower flow regime conditions, which is related to the sorting by

absorbing detrital material by selective deposition of fine grains from the suspension35. In most of the studied profiles the concentration of sulphide elements was low (Cu: 3.49–57.2 μg/g;

Zn: 3.12–210 μg/g; Fe: 1.2–99.3 mg/g) (Fig. 2), but irrespective of lithology, these results are typical for a river valley environment in Central Europe36. The stratigraphy differentiation

in deposit chemistry indicated that enrichment of Cu and Fe took place during the changes of sedimentation type from organic rich to mineral input or increased humification. Precipitation of

colloidal forms of these elements was dependent on changes in the local groundwater level37. In the Luciąża River valley the water budget was represented by water flowing underground into

the alluvia coming from the post-glacial areas surrounding the Rozprza stronghold (e.g. Radomsko and Dobryszyce Hills), the water of the Luciąża river system, and precipitation water that

did not participate in the evapotranspiration processes. TESTING DIFFERENCES AMONG CORES IN RELATIVE TAXON ABUNDANCE To examine whether an individual core can be considered representative of

the whole site, we tested for differences in relative dipteran taxon abundances among cores, controlling for sample volume and sample age effects. Diagnostics revealed no significant

deviations from model assumptions for the fitted poisson family generalised linear mixed model (GLMM). Fixed effect model selection based on AICc revealed the full model to be the

best-fitting model, including an interaction between core and species (LL = − 3727.7, d.f. = 294, AICc = 8079.1, weight = 1). This indicated a significant difference among cores RP W1, RP

W3(2) and RP W3(4) in chironomid relative species abundance distributions. Therefore, there may be some error associated with extrapolating results from a single core across a whole site.

The species with the biggest differences among cores in relative abundance were _Chironomus plumosus_-type and _Dicrotendipes notatus_-type, which both are common, often with high share in

the samples. However, they had much higher relative abundance in core RP W3(2) than RP W1 (Supplementary Table S5). DISCUSSION AND CONCLUSIONS The results based on chironomid and

ceratopogonid assemblages generally confirm three main stages of the moat history: limnetic, telmatic and terrestrial. In addition to the previous study5, however, we reveal variation in

moat habitat overall and in temporal habitat changes across the moat system. The afore-mentioned three stages of the moat are visible only in its deepest, south-western part (both RP W3

cores) (Fig. 2). Here the ecological processes, such as paludification, lasted longer and the habitat changes were less dynamic, resulting in more stable conditions for biota. However, the

limnetic phase was of a different nature in the RP W3(2) and RP W3(4) cores, which were located close (12.5 m) to each other. While fresh water from the artificial canal (see Figs. 1C, 6,5)

firstly supplied the southern part of the moat, the inflow may have been higher in RP W3(2) than in RP W3(4), probably because of its greater depth (Fig. 6). The significant presence of

rheophilic taxa in the former core (indicative for cluster Y2) supports this. Moreover, the sediment chemistry record (Fig. 2), in particular values of Fe/Mn ratio, suggest higher

oxygenation in this part of the moat, which may indicate the course of the water current. A crucial factor in this case could also be the structure of the moat bottom closely related to the

groundwater level, which determines the habitat diversification of the plant cover. These processes could lead to the aggregation of soils grains/sediment into concretions and lumps, which

when combined with Fe and decomposed organic matter, can lead to development of dense hardly permeable zones38. Such a situation in the studied area would have a direct impact on the

disturbance of vertical water movement and the possibility of plant rooting, determining the specific geochemical cycle between moats-plants-sediments. While habitats in both RP W3 cores

during the limnetic phase, with high pH and dense vegetation, could support well-functioning biotic communities, conditions in the shallower part of RP W3(4) were slightly less favourable

for midges. There, the limnetic stage is reflected mainly in the Y1 subcluster, interrupted by single samples with lower concentration of midge larvae. The telmatic phase in this core could

have started earlier than in RP W3(2), as indicated by peat deposits and the X2 cluster. This effect may be caused by the location of RP W3(4) close to the moat edge, resulting in faster

sedentation and shallowing of the moat bottom. The north-eastern part of the main moat had worse conditions for chironomid development, as is shown by cluster X1 containing most samples from

the RP W1 core (Fig. 2). Generally, the RP W1 core is characterised by relatively uniform lithology, consisting mostly of peat (with periodical supply of the mineral fraction, including

sands). It is the result of much shallower conditions in this part of the main moat resulting in dominating sedentation of peat. During the first phase of the moat history (up to ca. 1440

AD), episodes of more complex midge larvae assemblages were recorded, as reflected by clusters Y2, X2 and Y1 (Fig. 2), and higher chironomid richness and abundance (Supplementary Fig. S5).

They may indicate some limnetic episodes also in the NE part of the moat (Fig. 6), though not as clear and stable as in its deeper southern part. Despite slight differences in the

concentration of most elements, they are confirmed by selected geochemical indicators (e.g. increase of Fe/Mn ratio to 82). These episodes are also accompanied by a clear decrease in the

values of denudation indicators (i.e. Na/K from 0.12 to 0.04 and Ca/Mg from 0.03 to 0.01) (Fig. 2). The record of the RP W1 core ends approx. 1560 AD, probably because of the anthropogenic

removal of the top parts of the moat filling during drainage works in the 20th c. AD. Generally, the history of the main moat lasted for six centuries, since its establishment in ca. 1329 AD

till the earthwork from 1944 AD, when the moat remnants were levelled (Fig. 6). In fact, the moat was only functioning as a defensive water body for the first 120–170 years of its

existence. Even then it was fully aquatic only in its deepest, south-western part, with only several recorded episodes of higher water level in the whole feature (Fig. 6B). Its status

changed in the 15th century, when it dried up (or was drained), becoming a kind of wetland, and ca. 300 years later it was fully filled with organic deposits and covered by overbank mineral

matter. The second moat (RP F2 core) was functioning briefly, as it was created not earlier than in the 2nd half of 15th century, and most probably ca. 1500 AD (see Fig. 2, Supplementary

Table S4). In fact, it is not certain whether it was built as a functional moat, or possibly as a dry ditch (e.g. for melioration or defensive purposes). The chironomid scarcity and high

admixture of sands (Figs. 2, 5) rather support the latter possibility. Moreover, the secondary moat was much narrower (ca. 11 m) than the main one, so active slope processes provided a

constant supply of mineral matter (mostly sands). The results of habitat reconstruction confirm the modelling (GLMM) outcome that one core does not show the entire history of the moat. This

is because a moat is not a typical water body—not only very small and shallow, but also with a specific shape, which hinders transport of fossils to the deepest part of the moat. The steep,

almost vertical banks, artificial channels, depth differences and many other features had a crucial impact on the sediment spatial composition and, hence, also on moat functioning.

Anthropogenic wood and other artefacts and ecofacts in the bottom can serve as an additional habitat, e.g. for periphyton development. Besides Rozprza, only a few moats in Europe had

hitherto been studied using palaeoecological analyses (e.g.39,40,41), and they mostly focused on human economy and functioning, rarely touching the issue of the moat ecology itself.

Moreover, some of these studied defensive objects were dry (e.g. in Prague42 and Gdańsk43). The external moat of the Czermno stronghold44 seems to be comparable with the Rozprza site, though

they cover different time spans. Both features evidenced relatively fast peat sedentation and paludification in comparison with natural water bodies (cf.17). The moat system in the Tum

(Łęczyca) stronghold has been well studied, including palynological and plant macrofossil analyses from its different parts45,46. However, no further spatial reconstruction of the

environmental conditions within the feature was provided. Moats and similar human-made features are hardly comparable with natural ecosystems. While the multiple coring approach is sometimes

undertaken i.e. to track past water-level changes47,48, in such archaeological sites as Rozprza motte, various factors need to be considered, in particular human impact. In waterlogged

sites, such as the Serteya Neolithic pile-dwelling, the human–environment relation can be tracked alongside the quantitatively reconstructed climatic background22,49. Palaeoecological

methods are of great importance while tracking the history of the cities, like Gdańsk50 and London51 with the use of profiles of wet sediments. In the majority of archaeological sites, if

palaeoecological studies are undertaken, they focus on the surroundings of the excavations, mainly because of the lack of wet organic sediments to take core from (e.g.52,53). Another issue

is the cost and time, which need to be taken into consideration with any additional core. In our case, the cores of sediments were taken as monoliths directly from the walls of

archaeological trenches, and the profiles were relatively short, which was a great convenience. The additional cores, however, were examined only with respect to lithological and geochemical

composition, accompanied by chironomid (and ceratopogonid) analysis. With the use of macrofossil analysis, habitat diversity could be even more accurately mapped, which is worth considering

in future research. To sum up, multiple cores are required to get a complete picture of the spatio-temporal changes within the ecosystem. The environmental reconstructions from the deepest

part of the moat (RP W3(2)) presented in Kittel et al.5 are substantiated here, and the results are largely consistent with the core taken from the same trench (RP W3(4)). However, the

sequences from the second moat (RP F2) and the NE part of the main moat (RP W1) significantly complement the reconstruction and help give a better understanding of the functioning of moat

ecosystems and motte-and-bailey strongholds. This study represents the first reconstruction of moat habitats during its functioning that considers spatial variation. It is likely that many

similar water bodies could be investigated this way, broadening our knowledge about past societies and ecology of such human-made ecosystems. MATERIAL AND METHODS FIELDWORK The research in

Rozprza began with a non-destructive survey carried out in 2013–2015. Methods included analytical field walking, aerial photography, geophysical and geochemical prospecting, as well as

thorough geological mapping. A dense network of cores taken with a hand auger resulted in elaboration of detailed cross-sections of the ringfort vicinity. Thanks to this investigation it was

possible to localise some archaeological and palaeogeographical features27,54. This led to the next extensive, interdisciplinary investigation. This fieldwork was conducted in 2015–2016,

with the use of archaeological trenches, geological outcrops and a wide range of palaeoecological studies. They aimed to reconstruct the environmental conditions and settlement history of

the mediaeval stronghold at Rozprza5,16,55. The procedure of exposing trenches included removing successive 10-cm layers of sediments, distinguishing stratigraphic units within them, and

wet-sifting with a 4 × 4 mm sieve in order to collect archaeological artefacts and ecofacts. All trench walls and collected features were thoroughly documented as orthophotos. The sediments

for palaeoecological analyses were collected from the trench walls as monoliths using metal boxes with dimensions of 50 × 10 × 10 cm (Supplementary Fig. S1). Thanks to this method, the

undisturbed structure of the sediments was preserved. The RP W1 profile was collected from the deepest section of the trench 1/2015. This trench, with dimensions 2.5 × 12 m, was exposed in

the eastern part of the main moat (Fig. 1C). Wooden vertical posts, associated with numerous fragments of wood, were revealed in the bottom of the moat. The moat was shallow here, reaching

up to 50 cm depth. Two cores of sediments were collected from the trench 3/2015 (1.5 × 25 m), situated in the south-western part of the main moat. The RP W3(2) profile was taken from the

deepest part of the main moat and RP W3(4) from its shallower part. Trench 3/2015 exposed the very well preserved moat fill, adjoining the outer rampart and the motte mound, allowing for

their full cross-section. The RP F2 profile was taken from the thoroughly deepened and purified wall of the drainage ditch, which currently crosses the secondary moat. The deposits were

collected from the deepest part of the smaller southern moat (Fig. 1C). DIGITAL RECONSTRUCTION OF THE MAIN MOAT RELIEF The 3D reconstruction model of the bottom of the main moat was prepared

within the GIS environment (Qgis, SAGA GIS and PlanlaufTerrain softwares) using point cloud of Airborne Laser Scanning (ALS) already accessible via the Geoportal.gov.pl web service. This

was supplemented with results of detailed coring (80 drillings 1 to 2 m apart) of the moat as well as results of excavation of archaeological trenches 1/2015 and 3/2015. Contemporary bare

earth points covering the moat in the ALS derived point cloud were replaced by points with the height values of the surface of mineral bedrock indicating the original bottom of the moat.

Subsequently, all the points were interpolated to obtain a Digital Elevation Model of the stronghold area with the main moat virtually reconstructed and emptied. This allowed for modelling

water circulation and subsequent changes of moat states. GEOCHEMICAL AND SEDIMENTOLOGICAL ANALYSIS Detailed geochemical tests covered material from the four cores presented here (133 samples

from 4 cores) (Fig. 2). The basic physical and chemical parameters were the following: organic matter content (LOI—loss on ignition), calcium carbonate (CaCO3) content (volumetric

measurement of CO2 from conversion of CaCO3 by 10% HCl), biophilic elements such as: total organic carbon (TOC), total nitrogen (TN) and total sulphur (TS) content (combustion method at the

Rapid CS cube and VarioMax analyser—Elementar), and reaction (pH in distilled water). All parameters were measured in 2-cm resolution according to the procedure by Tolksdorf et al.56. Ash

material without organic matter (remaining after LOI) was dissolved with concentrated 65% HNO3, 10% HCl and H2O2 in a Berghof Speedwave microwave mineralizer. Elements with

palaeoenvironmental significance (Na, K, Ca, Mg, Fe, Mn, Cu, Zn and Pb) identified in the resulting solution were marked by the atomic absorption spectroscopy (AAS) method with use of Solar

Unicam and following procedure after Borówka37. When interpreting the individual elements of the chemical composition and grain size distribution for the mineral fraction for all four cores,

the variable character of the sediment accumulation conditions was taken into account57,58. Palaeoenvironmental conditions responsible for the sedimentation of the studied deposits were

interpreted by determining the quantitative ratios of the elements (such as: Na/K, Fe/Mn and Cu/Zn) with the assumption that the individual lithogeochemical components came from different

sources (cf.49). The grain size composition of mineral ash (treated as terrigenous silica) remaining after solution was prepared as in Clift et al.59, using a Mastersizer 3000 laser

particle-size analyser (Malvern). The grain-size data were stored and processed using GRADISTAT software v. 8.060. CHIRONOMIDAE AND CERATOPOGONIDAE ANALYSIS The samples for midge analysis

were taken as contiguous 2-cm slices of the sediment from each profile, apart from the RP W3(2) core, from which they were collected with 4 cm resolution. The number of samples analysed in

each profile was similar (ranging between 23 in RP W1 and 27 in RP W3(2)), while sample volume varied between 5 and 70 cm3. Chironomidae preparation methods followed Brooks et al.25. The

sediments were passed through a 63 μm mesh sieve. If head capsule (hc) concentration in the sediments was low, kerosene flotation was used following the procedure of Rolland and Larocque61.

Processed sediment was scanned under a stereo-binocular microscope. Where applicable, a minimum of 50 (preferably 100) chironomid head capsules from each sample were picked and mounted in

Euparal®. Identification of chironomids followed Schmid62, Brooks et al.25, and Andersen et al.63, while ceratopogonids were divided into two morphotypes distinguished by Walker64.

Ecological preferences of identified taxa are based mainly on Brooks et al.25, Vallenduuk and Moller Pillot65, Moller Pillot30,66, and Luoto31. The midge sequences are presented on

stratigraphic diagrams (Supplementary Figures S3–S6) created with C2 software67. RADIOCARBON AND DENDROCHRONOLOGICAL DATING The chronology of the Rozprza moat system was estimated using

radiocarbon (14C) and dendrochronological methods. Both dendrochronological and conventional radiocarbon dating of organic material using the LSC technique were performed in the Laboratory

of Absolute Dating in Kraków (Poland). A few wood fragments sampled during moat system exploration16 were dendrochronologically dated using standard procedures68. A total of 15 samples from

moats of the Rozprza motte-and-bailey were collected for radiocarbon (14C) dating (Table 1). Thirteen of these were sampled from three cores of the main moat and two were from the southern

secondary moat (cf.16). For the full cross-section of the deepest part of the main moat, seven dates were obtained for the RP W3(2) core and four for RP W3(4)5. Two more 14C datings were

made for the RP W1 core from the eastern shallow part of the main moat, and a further two for the RP F2 core in the southern additional moat. Twelve samples of bulk organic deposits (organic

mud, peat or gyttja) were dated using the liquid scintillation technique (LSC) and three samples of selected terrestrial plant macrofossils dated using the accelerator mass spectrometry

technique (AMS). All samples were chemically pre-treated using the AAA (acid–alkali–acid) method. The procedure for conventional radiocarbon dating of organic material using the liquid

scintillation counting method (LSC) included the standard synthesis of benzene from organic samples69. 14C measurements were carried out with a 3-photomultiplier spectrometer, the HIDEX

300SL and Quantulus 1220. Organic samples dated using the AMS technique were combusted, purified, and graphitised with Fe catalyst70. The mixture of graphite and Fe powder was pressed into a

target holder and measured with the AMS system at the Centre for Applied Isotope Studies at the University of Georgia, USA or in the Accelerator Mass Spectrometry Laboratory (D-AMS

laboratory code) in Seattle, USA (see71 for details). Calibrated radiocarbon ages (BC/AD) were made using the IntCal20 radiocarbon calibration dataset72 and the OxCal 4.4.2 calibration

software73,74. The age-depth curves for studied cores were elaborated based on the OxCal P_Sequence model75. The age-depth models were obtained separately for four studied cores. More

detailed chronology was obtained for the longest RP W3(2) core, and the new model slightly differs from that published by Kittel et al.5. A dendrochronological date from a fragment of wooden

ecofact found in the very bottom of the main moat was included into RP W3(2) and RP W3(4) age-depth models (cf.5). For an estimation of absolute chronology of selected palaeoenvironmental

events, the probability distributions of the modelled calendar ages for 1-cm intervals of deposits were calculated (Supplementary Tables S1–S4). These were used for estimation of absolute

chronology of selected palaeoenvironmental events. STATISTICAL DATA ANALYSES SELF-ORGANISING MAP AND INDICATOR SPECIES ANALYSIS Patterns in the dipteran assemblages were recognized with

Kohonen’s (unsupervised) artificial neural network (ANN), also referred to as a self-organising map (SOM)76,77. Artificial neural networks (ANNs) are simple structural and functional models

of the brain. ANNs have many advantages, which allow a researcher to apply them to “difficult” data. They do not require any a priori specification of the model underlying a studied

phenomenon because they learn it based on the processed data. They are also robust to noise in data78,79. This is important for the purposes of the present study, because taxa abundances in

field samples do not reflect exactly the original abundances of populations80. Additionally, in palaeoecological research the long time separating the living populations and their sampling,

and resulting decomposition and fragmentation additionally enhance the problem81. ANNs are also robust to non-linear relationships between variables and to non-normal distributions in

data82,83. This is also crucial in this study because the counts of rare species cannot be effectively normalised by any transformation due to their absence in most samples and therefore

strongly skewed variable distributions82,84. Furthermore, dipteran assemblages are shaped by many abiotic and biotic factors that are related in complex ways. The SOM application in

palaeoecology and its advantage in comparison to classical zonation methods (CONISS and Optimal Partitioning) are well described by Płóciennik et al.85. Kohonen’s ANNs are constructed from

data processing units (neurons) arranged in two layers: an input layer used for data input, and an output layer responsible for data structuring and output. The data used for the SOM

analysis comprised log-transformed abundances of 97 taxa recorded in 79 non-empty core samples. They were displayed on the input layer comprising 97 neurons (one input neuron per taxon). The

output neurons were arranged as a two-dimensional rectangular lattice. The number of output neurons should be close to 5√_n_, where _n_ is the number of samples; in this case the result was

44 (see86). Therefore, the final size of the lattice was 7 × 6 (= 42) neurons. Each input neuron repeatedly transmitted signals to each output neuron. These signals were strengthened or

weakened by modifying the weight of the connections between neurons. On this basis, a virtual dipteran core sample (DCS) was created in each output neuron. The distance between virtual DCSs

on the two-dimensional lattice exhibited their mutual dissimilarity, i.e. virtual DCSs in distant output neurons differed considerably while those in neighbouring output neurons were

similar. The latter might not be true when the neighbouring output neurons were in different (sub)clusters as the virtual DCSs, and respective output neurons, were additionally clustered

with hierarchical cluster analysis (with Ward algorithm and Euclidean distance)78,86,87. Finally, each real DCS was assigned to the best matching virtual DCS and the respective output

neuron. Therefore, the mutual distance of the real DCSs on the two-dimensional lattice was a derivative of the mutual dissimilarity and position of virtual DCSs: significantly dissimilar

real DCSs were located in distant neurons, while similar real DCSs were located in the same neuron or in adjoining neurons83. The batch training algorithm was chosen for the purpose of

network training, because it does not require any training rate factor to be specified78. The network training and the clustering of virtual DCSs were performed with the use of the SOM

Toolbox88 developed by the Laboratory of Information and Computer Science at the Helsinki University of Technology (http://www.cis.hut.fi/projects/somtoolbox/). The SOM Toolbox allows the

associations between dipteran taxa and SOM regions to be visualised in the form of greyness gradients over the two-dimensional lattice83. This visualisation may facilitate the formulation of

ecological conclusions as taxa with the same patterns of greyness usually co-occurred and exhibited similar habitat preferences. However, the SOM Toolbox does not provide a statistical

verification of those associations. For this reason, the untransformed dipteran abundance data were subjected to Indicator Species Analysis (ISA): the associations between each dipteran

taxon and each subcluster of output neurons, and its respective environmental conditions, were expressed in a numeric form with the indicator values (IndVals)89. IndVals complement the

visualisation in the form of greyness gradients. An IndVal (range 0–100%) of taxon _i_ in all real DCSs of subcluster _j_ is a product of three values: (1) Aij—a measure of specificity, i.e.

the mean abundance of taxon _i_ in real DCSs assigned to subcluster _j_ divided by the sum of its average abundances in all subclusters (%), (2) F_ij_—a measure of fidelity, i.e. the

frequency of occurrence of taxon _i_ (%) in real DCSs assigned to subcluster _j_, and (3) the constant 100 in order to produce the percentages: $${\text{IndVal}}_{ij} = {\text{A}}_{ij}

\times {\text{F}}_{ij} \times {1}00$$ $${\text{A}}_{{{\text{ij}}}} = {\text{ taxon}}\;{\text{abundance}}_{ij} /{\text{ taxon}}\;{\text{abundance}}_{i.}$$ $${\text{F}}_{ij} = {\text{

N}}\;{\text{real}}\;{\text{core}}\;{\text{samples}}_{ij} /{\text{N}}\;{\text{real}}\;{\text{core}}\;{\text{samples}}_{.j}$$ The maximum IndVal (100%) was observed when all real DCSs with

taxon _i_ were assigned to subcluster _j_ and when taxon _i_ was present in all real DCSs assigned to subcluster _j_89. Significant maximum IndVals, and therefore significant associations of

individual taxa with a given SOM subcluster (and its respective environmental conditions), were identified with Monte Carlo randomisation statistics. The significance level was calculated

as the proportion of randomised trials with IndVal exceeding or equal to the observed IndVal. The above calculations were performed in PC-ORD90. GENERALISED LINEAR MIXED MODEL We asked

whether chironomid taxon composition was consistent across sediment cores, so that results from a single core could be extrapolated across the whole site. Following Hadfield et al.91, we

used a poisson family generalised linear mixed model (GLMM) with log link to test whether relative species abundances differed between cores RP W1, RP W3(2) and RP W3(4). Core RP F2 was

excluded due to low temporal overlap of RP F2 sample ages with other cores, particularly core RP W1 (Fig. 2). Three samples with no age estimate from radiocarbon dating were excluded.

Chironomid morphotypes that were absent from a core, age category, or sample were included as zero counts. After reducing samples to only those with strongly overlapping ages among the three

cores, sample age was converted to a factor with 6 levels, to allow for nonlinearities in changes in species abundance over time. All resulting age categories were represented by all

studied cores. This resulted in a dataset with 5141 individual species counts. The response variable in the generalised linear mixed model (GLMM) was the untransformed count of individuals

of each chironomid morphotype in each sample. To control for variation in sediment volume among samples, we included ln(sample volume cm3) as an offset (logged because we fit a poisson model

with log link), so that the fixed effect parameter estimates represented the effects of predictors on chironomid counts per unit sediment volume. Fixed effect predictors included core,

chironomid taxon (morphospecies), and their interaction. Sample age category and its pairwise interactions with core and morphotype were included as random effects to control for temporal

variation in abundances. The core: morphospecies interaction fixed effect therefore tested for differences among cores in relative species abundance, controlling for differences in overall

abundance among cores and morphospecies (fixed main effects), sample volume (offset), and any influence of sample age (random effects). All analyses were run in R version 4.1.292. GLMM was

performed using the R package glmmTMB93. Model diagnostics were performed in DHARMa94, and fixed effect model selection based on AICc carried out in MuMIn95. All random effects were included

in every model. Post-hoc comparison of relative species abundance differences among cores for individual chironomid morphotypes, based on the estimated core:morphospecies interaction, were

carried out in the R package phia96. This package is not compatible with mixed effects models and so these analyses are based on a GLM model including fixed effects only, run using base R’s

glm function. CANONICAL CORRESPONDENCE ANALYSIS Because the performed DCA for all four combined cores dataset revealed long biological data gradients (4.499 on Ax 1 and 4.787 on Ax 2 [SD

units]), Canonical Correspondence Analysis (CCA) was selected to compare geochemical and biotic variable patterns. Due to autocorrelation, Na and Mg content were excluded from the further

analysis. The CCA was performed on square-root transformed data with downweighting rare taxa, biplot scaling and inter-sample distance. The significance of environmental variables relating

to the biota was tested with the Monte Carlo permutation with automatic selection and permutation under full model. DATA AVAILABILITY The datasets analysed during the current study are

available from the corresponding author on reasonable request. REFERENCES * Kirilova, E. P., Cremer, H., Heiri, O. & Lotter, A. F. Eutrophication of moderately deep Dutch lakes during

the past century: Flaws in the expectations of water management? _Hydrobiologia_ 637, 157–171 (2010). Article  CAS  Google Scholar  * Scharf, B. & Viehberg, F. A. Living Ostracoda

(Crustacea) from the town moat of Bremen, Germany. _Crustaceana_ 87(8–9), 1124–1135 (2014). Article  Google Scholar  * Rees, S. E. The historical and cultural importance of ponds and small

lakes in Wales, UK. _Aquat. Conserv._ 7(2), 133–139 (1997). Article  Google Scholar  * Brown, A. _et al._ The ecological impact of conquest and colonisation on a medieval frontier landscape:

Combined palynological and geochemical analysis of lake sediments from Radzyń Chełmiński, northern Poland. _Geoarchaeology_ 30, 511–527 (2015). Article  Google Scholar  * Kittel, P. _et

al._ The palaeoecological development of the Late Medieval moat—Multiproxy research at Rozprza Central Poland. _Quat. Int._ 482, 131–156 (2018). Article  Google Scholar  * Hildebrandt-Radke,

I. Geoarchaeological aspects in the studies of prehistoric and early historic settlement complexes. In _Studia interdyscyplinarne nad środowiskiem i kulturą w Polsce. Tom 1.

Środowisko-Człowiek-Cywilizacja_ (eds Makohonienko, M. _et al._) 57–70 (Bogucki Wyd Naukowe, 2007). Google Scholar  * Łyskowski, M. & Wardas-Lasoń, M. Georadar investigations and

geochemical analysis in contemporary archaeological studies. _Geol. Geophys. Environ._ 38(3), 307–315 (2012). Article  Google Scholar  * Korhola, A. & Rautio, M. Cladocera and other

branchiopod crustaceans. In _Tracking Environmental Change Using Lake Sediments, Vol. 4: Zoological Indicators_ (eds Smol, J. P. _et al._) 5–41 (Kluwer Academic Publishers, 2001). Chapter 

Google Scholar  * Birks, H. H. Plant macrofossils. In _Tracking Environmental Change Using Lake Sediments, 3: Terrestrial, Algal, and Siliceous Indicators_ (eds Smol, J. P. _et al._) 49–74

(Kluwer Academic Publishers, 2001). Google Scholar  * Battarbee, R. W. Diatom analysis. In _Handbook of Holocene Palaeoecology and Palaeohydrology_ (ed. Berglund, B. E.) 527–570 (Wiley,

1986). Google Scholar  * Luoto, T. P., Nevalainen, L., Kultti, S. & Sarmaja-Korjonen, K. An evaluation of the influence of water depth and river inflow on quantitative Cladocera-based

temperature and lake level inferences in a shallow boreal lake. _Hydrobiologia_ 676, 143–154 (2011). Article  CAS  Google Scholar  * Luoto, T. P. Intra-lake patterns of aquatic insect and

mite remains. _J. Paleolimnol._ 47, 141–157 (2012). Article  Google Scholar  * Hann, B. J. Methods in Quaternary ecology. Cladocera. _Geosci. Canada_ 16, 17–26 (1989). Google Scholar  *

Dimbleby, G. W. _The Palynology of Archaeological Sites_ (Academic Press. Inc., 1985). Google Scholar  * Edwards, K. J. Using space in cultural palynology: The value of the off-site pollen

record. In _Modelling Ecological Change: Perspectives from Neoecology, Palaeoecology and Environmental Archaeology_ (eds Harris, D. R. & Thomas, K. D.) 61–74 (Routledge Taylor &

Francis Group, 2016). Google Scholar  * Kittel, P., Sikora, J. & Wroniecki, P. A Late Medieval motte-and-bailey settlement in a lowland river valley landscape of central Poland.

_Geoarchaeology_ 33(5), 558–578 (2018). Article  Google Scholar  * Antczak-Orlewska, O. _et al._ The environmental history of the oxbow in the Luciąża River valley—Study on the specific

microclimate during Allerød and Younger Dryas in central Poland. _Quat. Int._ https://doi.org/10.1016/j.quaint.2021.08.011 (2021). Article  Google Scholar  * Dearing, J. A. Core correlation

and total sediment influx. In _Handbook of Holocene Palaeoecology and Palaeohydrology_ (ed. Berglund, B. E.) 247–270 (Wiley, 1986). Google Scholar  * O’Brien, C. _et al._ A sediment-based

multiproxy palaeoecological approach to the environmental archaeology of lake dwellings (crannogs), central Ireland. _Holocene_ 15, 707–719 (2005). Article  Google Scholar  * Ruiz, Z.,

Brown, A. G. & Langdon, P. G. The potential of chironomid (Insecta: Diptera) larvae in archaeological investigations of floodplain and lake settlements. _J. Archaeol. Sci._ 33, 14–33

(2006). Article  Google Scholar  * Kittel, P. _et al._ A multi-proxy reconstruction from Lutomiersk-Koziówki, Central Poland, in the context of early modern hemp and flax processing. _J.

Archaeol. Sci._ 50, 318–337 (2014). Article  Google Scholar  * Kittel, P. _et al._ On the border between land and water: the environmental conditions of the Neolithic occupation from 4.3

until 1.6 ka BC at Serteya, Western Russia. _Geoarchaeology_ 36, 173–202 (2021). Article  Google Scholar  * Makohonienko, M. _et al._ Environmental changes during Mesolithic-Neolithic

transition in Kuyavia Lakeland, Central Poland. _Quat. Int._ https://doi.org/10.1016/j.quaint.2021.11.020 (2021). Article  Google Scholar  * Porinchu, D. F. & MacDonald, G. M. The use

and application of freshwater midges (Chironomidae: Insecta: Diptera) in geographical research. _Prog. Phys. Geogr._ 27, 378–422 (2003). Article  Google Scholar  * Brooks, S. J., Langdon, P.

G. & Heiri, O. _The Identification and Use of Palaearctic Chironomidae Larvae in Palaeoecology._ QRA Technical guide no. 10 (Quaternary Research Association, 2007). * Heiri, O., Birks,

H. J. B., Brooks, S. J., Velle, G. & Willassen, E. Effects of within-lake variability of fossil assemblages on quantitative chironomid-inferred temperature reconstruction. _Palaeogeogr.

Palaeoclimatol. Palaeoecol._ 199, 95–106 (2003). Article  Google Scholar  * Kittel, P., Sikora, J. & Wroniecki, P. The morphology of the Luciąża River valley floor in the vicinity of the

Rozprza medieval ring-fort in light of geophysical survey. _Bull. Geogr. Phys. Geogr. Ser._ 8, 95–106 (2015). Article  Google Scholar  * Hingham, R. & Barker, P. _Timber Castles_

(University of Exeter Press, 2002). Google Scholar  * Marciniak-Kajzer, A. _Archaeology on Medieval Knights’ Manor Houses in Poland_ (Wyd. Uniwersytetu Łódzkiego, Wyd. Uniwersytetu

Jagiellońskiego, 2016). Book  Google Scholar  * Moller Pillot, H. K. M. _Chironomidae Larvae of the Netherlands and Adjacent Lowlands. Biology and Ecology of the Aquatic Orthocladiinae,

Prodiamesinae, Diamesinae, Buchonomyiinae, Podonominae, Telmatogetoninae_ (KNNV Publishing, 2013). Book  Google Scholar  * Luoto, T. P. An assessment of lentic ceratopogonids,

ephemeropterans, trichopterans and oribatid mites as indicators of past environmental change in Finland. _Ann. Zool. Fenn._ 46, 259–270 (2009). Article  Google Scholar  * Cierniewski, J.

Spatial complexity of the Cybina river valley organic soils against the background of physiographic conditions. _Soil Sci. Annu._ 32(4), 3–51 (1981). CAS  Google Scholar  * Rydelek, P.

Origin and composition of mineral constituents of fen peats from Eastern Poland. _J. Plant Nutr._ 36(6), 911–928 (2013). Article  CAS  Google Scholar  * Wachecka-Kotkowska, L. _Rozwój rzeźby

obszaru między Piotrkowem Trybunalskim, Radomskiem a Przedborzem w czwartorzędzie_ (Wyd. Uniwersytetu Łódzkiego, 2015). Book  Google Scholar  * Kittel, P. _et al._ Lacustrine, fluvial and

slope deposits in the wetland shore area in Serteya, Western Russia. _Acta Geogr. Lodz_ 110, 103–124 (2020). Google Scholar  * Ciszewski, D. Pollution of Mała Panew River sediments by heavy

metals: Part I. Effect of changes in river bed morphology. _Pol. J. Environ. Stud._ 13(6), 589–595 (2004). CAS  Google Scholar  * Borówka, R. Late Vistulian and Holocene denudation magnitude

in morainic plateaux: Case studies in the zone of maximum extent of the last ice sheet. _Quat. Stud. Pol._ 9, 5–31 (1990). Google Scholar  * Prusinkiewicz, Z., Bednarek, R., Kośko, A. &

Szmyt, M. Palaeopedological studies of the age and properties of illuvial bands at an archaeological site. _Quat. Int._ 51(52), 195–201 (1998). Article  Google Scholar  * Kühtreiber, T. The

medieval castle Lanzenkirchen in Lower Austria: reconstruction of economical and ecological development of an average-sized manor (12th–15th century). _Archaeol. Pol._ 37, 135–144 (1999).

Google Scholar  * Kočár, P., Čech, P., Kozáková, R. & Kočárová, R. Environment and economy of the early medieval settlement in Žatec. _Interdiscip. Archaeol._ 1, 45–60 (2010). Google

Scholar  * Brown, A. D. & Pluskowski, A. G. Detecting the environmental impact of the Baltic Crusades on a late medieval (13th-15th century) frontier landscape: Palynological analysis

from Malbork Castle and hinterland, Northern Poland. _J. Archaeol. Sci._ 38, 1957–1966 (2011). Article  Google Scholar  * Beneš, J. _et al._ Archaeobotany of the Old Prague Town defence

system, Czech Republic: Archaeology, macro-remains, pollen, and diatoms. _Veg. Hist. Archaeobot._ 11(1/2), 107–119 (2002). Article  Google Scholar  * Badura, M. & Latałowa, M. Szczątki

makroskopowe roślin z obiektów archeologicznych Zespołu Przedbramia w Gdańsku. In _Zespół Przedbramia ul. Długiej w Gdańsku. Studium archeologiczne_ (ed. Pudło, A.) 231–247 (Muzeum Historii

Miasta Gdańska, 2016). Google Scholar  * Dobrowolski, R. _et al._ Environmental conditions of settlement in the vicinity of the mediaeval capital of the Cherven Towns (Czermno site,

Hrubieszów Basin, Eastern Poland). _Quat. Int._ 493, 258–273 (2018). Article  Google Scholar  * Makohonienko, M. Środowisko przyrodnicze i gospodarka w otoczeniu średniowiecznego grodu w

Łęczycy w świetle analizy palinologicznej. In _Początki Łęczycy. Tom I—Archeologia środowiskowa średniowiecznej Łęczycy. Przyroda–Gospodarka–Społeczeństwo_ (eds Grygiel, R. & Jurek, T.)

95–190 (MAiE w Łodzi, 2014). Google Scholar  * Koszałka, J. Źródła archeobotaniczne do rekonstrukcji uwarunkowań przyrodniczych oraz gospodarczych grodu w Łęczycy. In _Początki Łęczycy. Tom

I – Archeologia środowiskowa średniowiecznej Łęczycy. Przyroda–Gospodarka–Społeczeństwo_ (eds Grygiel, R. & Jurek, T.) 191–241 (MAiE w Łodzi, 2014). Google Scholar  * Digerfeldt, G.

Studies on past lake-level fluctuations. In _Handbook of Holocene Palaeoecology and Palaeohydrology_ (ed. Berglund, B. E.) 127–143 (Wiley, 1986). Google Scholar  * Magny, M.

Palaeoclimatology and archaeology in the wetlands. In _The Oxford Handbook of Wetland Archaeology_ (eds Menotti, F. & O’Sullivan, A.) 585–597 (Oxford University Press, 2013). Google

Scholar  * Płóciennik, M. _et al._ Summer temperature drives the lake ecosystem during the Late Weichselian and Holocene in Eastern Europe: A case study from East European Plain. _CATENA_

214, 106206 (2022). Article  Google Scholar  * Święta-Musznicka, J., Badura, M., Pędziszewska, A. & Latałowa, M. Environmental changes and plant use during the 5th–14th centuries in

medieval Gdańsk, northern Poland. _Veget. Hist. Archaeobot._ 30, 363–381 (2021). Article  Google Scholar  * Rackham, J. & Sidell, J. London’s landscapes: The changing environment. In

_The Archaeology of Greater London. An Assessment of Archaeological Evidence for Human Presence in the Area Now Covered by Greater London_ (ed. Kendall, M.) 12–27 (Museum of London, 2000).

Google Scholar  * Ledger, P., Edwards, K. & Schofield, J. A multiple profile approach to the palynological reconstruction of Norse landscapes in Greenland’s Eastern Settlement. _Quat.

Res._ 82(1), 22–37 (2014). Article  Google Scholar  * Albert, B. & Innes, J. Multi-profile fine-resolution palynological and micro-charcoal analyses at Esklets, North York Moors, UK,

with special reference to the Mesolithic-Neolithic transition. _Veget. Hist. Archaeobot._ 24, 357–375 (2015). Article  Google Scholar  * Sikora, J., Kittel, P. & Wroniecki, P. From a

point on the map to a shape in the landscape. Non-invasive verification of medieval ring-forts in Central Poland: Rozprza case study. _Archaeol. Pol._ 53, 510–514 (2015). Google Scholar  *

Sikora, J. _et al._ A palaeoenvironmental reconstruction of the rampart construction of the medieval ring-fort in Rozprza, Central Poland. _Archaeol. Anthropol. Sci._ 11(8), 4187–4219

(2019). Article  Google Scholar  * Tolksdorf, J. F., Turner, F., Nelle, O., Peters, S. & Bruckner, H. Environmental development and local human impact in the Jeetzel valley (N Germany)

since 10 ka BP as detected by geoarchaeological analyses in a coupled aeolian and lacustrine sediment archive at Soven. _E&G Quat. Sci. J._ 64, 95–110 (2015). Article  Google Scholar  *

Oonk S., Slomp C. P. & Huisman D. J. Geochemistry as an aid in archaeological prospection and site interpretation: Current issues and research directions. _Archaeol. Prospect._ 16, 35–51

(2009). Article  Google Scholar  * Zieliński, T. & Pisarska-Jamroży, M. Which features of deposits should be included in a code and which not? _Przegl. Geol._ 60, 387–397 (2012). Google

Scholar  * Clift, P. D. _et al._ Grain-size variability within a mega-scale point-bar system, False River, Louisiana. _Sedimentology_ 66, 408–434 (2019). Article  Google Scholar  * Blott,

S. J. & Pye, K. GRADISTAT: A grain size distribution and statistics package for the analysis of unconsolidated sediments. _Earth Surf. Process. Landf._ 26, 1237–1248 (2001). Article 

Google Scholar  * Rolland, N. & Larocque, I. The efficiency of kerosene flotation for extraction of chironomid head capsules from lake sediments samples. _J. Paleolimnol._ 37, 565–572

(2007). Article  Google Scholar  * Schmid, P. E. _A Key to the Chironomidae and Their Instars from Austrian Danube Region Streams and Rivers. Part I. Diamesinae Prodiamesinae and

Orthocladiinae_ (Federal Institute for Water Quality of the Ministry of Agriculture and Forestry, 1993). Google Scholar  * Andersen, T., Cranston, P. S. & Epler, J. H. _Chironomidae of

the Holarctic Region: Keys and Diagnoses. Part 1. Larvae. Insect Systematics and Evolution. Supplement 66_ (Scandinavian Entomology, 2013). Google Scholar  * Walker, I. R. Midges:

Chironomidae and related Diptera. In _Tracking Environmental Change Using Lake Sediments, Volume 4: Zoological Indicators_ (eds Smol, J. P. _et al._) 43–66 (Kluwer Academic Press, 2001).

Chapter  Google Scholar  * Vallenduuk, H. J. & Moller Pillot, H. K. M. _Chironomidae Larvae of the Netherlands and Adjacent Lowlands. General Ecology and Tanypodinae_ (KNNV Publishing,

2007). Google Scholar  * Moller Pillot, H. K. M. _Chironomidae Larvae Biology and Ecology of the Chironomini_ (KNNV Publishing, 2009). Book  Google Scholar  * Juggins, S. _C2 Version 1.5

User Guide. Software for Ecological and Palaeoecological Data Analysis and Visualisation_ (Newcastle University, 2007). Google Scholar  * Schweingruber, F. H. _Tree Rings. Basics and

Applications of Dendrochronology_ (Kluwer Academic Publishers, 1988). Google Scholar  * Skripkin, V. V. & Kovaliukh, N. N. Recent developments in the procedures used at the SSCER

Laboratory for the routine preparation of lithium carbide. _Radiocarbon_ 40(1), 211–214 (1998). Article  CAS  Google Scholar  * Krąpiec, M., Rakowski, A. Z., Huels, M., Wiktorowski, D. &

Hamann, C. A new graphitization system for radiocarbon dating with AMS on the dendrochronological laboratory at AGH-UST Kraków. _Radiocarbon_ 60(4), 1091–1100 (2018). Article  Google

Scholar  * Zoppi, U., Crye, J., Song, Q. & Arjomand, A. Performance evaluation of the new AMS system at Accium BioSciences. _Radiocarbon_ 49, 173–182 (2007). Article  CAS  Google Scholar

  * Reimer, P. _et al._ The IntCal20 Northern Hemisphere radiocarbon age calibration curve (0–55 cal kBP). _Radiocarbon_ 62(4), 725–757 (2020). Article  CAS  Google Scholar  * Bronk Ramsey,

C. _OxCal Version 4.4.2_. Available at: https://c14.arch.ox.ac.uk (2020). * Bronk Ramsey, C. Bayesian analysis of radiocarbon dates. _Radiocarbon_ 51(1), 337–360 (2009). Article  Google

Scholar  * Bronk Ramsey, C. Deposition models for chronological records. _Quat. Sci. Rev._ 27(1–2), 42–60 (2008). Article  Google Scholar  * Kohonen, T. Self-organized formation of

topologically correct feature maps. _Biol. Cybern._ 43, 59–69 (1982). Article  MathSciNet  MATH  Google Scholar  * Kohonen, T. _Self-Organizing Maps_ (Springer, 2001). Book  MATH  Google

Scholar  * Park, Y.-S. _et al._ Application of a self-organizing map to select representative species in multivariate analysis: A case study determining diatom distribution patterns across

France. _Ecol. Inform._ 1, 247–257 (2006). Article  Google Scholar  * Zhang, Q. _et al._ Self-organizing feature map classification and ordination of _Larix principis-rupprechtii_ forest in

Pangquangou Nature Reserve. _Acta Ecol. Sin._ 31, 2990–2998 (2011). Google Scholar  * Ney, J. J. Practical use of biological statistics. In _Inland Fisheries Management in North America_

(eds Kohler, C. C. _et al._) 137–158 (American Fisheries Society, 1993). Google Scholar  * Płóciennik, M. _et al._ Fen ecosystem responses to water-level fluctuations during the early and

middle Holocene in central Europe: A case study from Wilczków, Poland. _Boreas_ 44(4), 721–740 (2015). Article  Google Scholar  * Brosse, S., Giraudel, J. L. & Lek, S. Utilisation of

non-supervised neural networks and principal component analysis to study fish assemblages. _Ecol. Model._ 146(1), 159–166 (2001). Article  Google Scholar  * Lek, S., Scardi, M., Verdonschot,

P. F. M., Descy, J. P. & Park, Y. S. _Modelling Community Structure in Freshwater Ecosystems_ (Springer, 2005). Book  Google Scholar  * Quinn, G. P. & Keough, M. _Experimental

Design and Data Analysis for Biologists_ (University of Cambridge, 2002). Book  Google Scholar  * Płóciennik, M., Kruk, A., Michczyńska, D. J. & Birks, H. J. B. Kohonen artificial neural

networks and the IndVal index as supplementary tools for the quantitative analysis of palaeoecological data. _Geochronometria_ 42, 189–201 (2015). Article  Google Scholar  * Vesanto, J.

& Alhoniemi, E. Clustering of the self-organizing map. _IEEE Trans. Neural Netw._ 11, 586–600 (2000). Article  CAS  PubMed  Google Scholar  * Ward, J. H. Hierarchical grouping to

optimize an objective function. _J. Am. Stat. Assoc._ 58, 236–244 (1963). Article  MathSciNet  Google Scholar  * Alhoniemi, E., Hollmén, J., Simula, O. & Vesanto, J. Process monitoring

and modeling using the self-organizing map. _Integr. Comput. Aided Eng._ 6(1), 3–14 (1999). Article  Google Scholar  * Dufrêne, M. & Legendre, P. Species assemblages and indicator

species: The need for a flexible asymmetrical approach. _Ecol. Monogr._ 67, 345–366 (1997). Google Scholar  * McCune, B. & Mefford, M. S. _PcOrd Multivariate Analysis of Ecological Data.

Version 6.06_ (MjM Software, 2011). Google Scholar  * Hadfield, J. D., Krasnov, B. R., Poulin, R. & Nakagawa, S. A tale of two phylogenies: Comparative analyses of ecological

interactions. _Am. Nat._ 183(2), 174–187 (2014). Article  PubMed  Google Scholar  * R Core Team. _R: A Language and Environment for Statistical Computing_ (R Foundation for Statistical

Computing, 2021). Google Scholar  * Brooks, M. E. _et al._ glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. _R Journal_ 9(2),

378–400 (2017). Article  Google Scholar  * Hartig, F. _DHARMa: Residual Diagnostics for Hierarchical (Multi-level/Mixed) Regression Models. R Package Version 0.4.5_.

https://CRAN.R-project.org/package=DHARMa (2022). * Bartoń, K. _MuMIn: Multi-model Inference. R Package Version 1.43.17_. https://CRAN.R-project.org/package=MuMIn (2020). * de

Rosario-Martinez, H. _phia: Post-Hoc Interaction Analysis. R Package Version 0.2-1_. https://CRAN.R-project.org/package=phia (2015). Download references ACKNOWLEDGEMENTS We would like to

thank dr hab. R. Stachowicz-Rybka for macrofossil selection for AMS dating and P. Wroniecki for providing an aerial photograph of the study site. The research project has been financed by a

grant from the National Science Centre, Poland based on the decision No. DEC-2013/11/B/HS3/03785 and by an IDUB UGrants First (decision No. 533-D000-GF09-22). AUTHOR INFORMATION AUTHORS AND

AFFILIATIONS * Laboratory of Palaeoecology and Archaeobotany, Department of Plant Ecology, Faculty of Biology, University of Gdansk, 59 Wita Stwosza St., 80-308, Gdańsk, Poland Olga

Antczak-Orlewska * Department of Invertebrate Zoology and Hydrobiology, Faculty of Biology and Environmental Protection, University of Lodz, 12/16 Banacha St., 90-237, Lodz, Poland Olga

Antczak-Orlewska & Mateusz Płóciennik * Institute of Marine and Environmental Sciences, University of Szczecin, 18 Mickiewicza St., 70-383, Szczecin, Poland Daniel Okupny * Department of

Ecology and Vertebrate Zoology, Faculty of Biology and Environmental Protection, University of Lodz, 12/16 Banacha St., 90-237, Lodz, Poland Andrzej Kruk & Richard Ian Bailey *

Department of Historical Archaeology and Weapon Studies, Institute of Archaeology, University of Lodz, 65 Narutowicza St., 90-131, Lodz, Poland Jerzy Sikora * Faculty of Geology, Geophysics

and Environmental Protection, AGH University of Science and Technology, 30 Mickiewicza St., 30-059, Kraków, Poland Marek Krąpiec * Department of Geology and Geomorphology, Faculty of

Geographical Sciences, University of Lodz, 88 Narutowicza St., 90-139, Lodz, Poland Piotr Kittel Authors * Olga Antczak-Orlewska View author publications You can also search for this author

inPubMed Google Scholar * Daniel Okupny View author publications You can also search for this author inPubMed Google Scholar * Andrzej Kruk View author publications You can also search for

this author inPubMed Google Scholar * Richard Ian Bailey View author publications You can also search for this author inPubMed Google Scholar * Mateusz Płóciennik View author publications

You can also search for this author inPubMed Google Scholar * Jerzy Sikora View author publications You can also search for this author inPubMed Google Scholar * Marek Krąpiec View author

publications You can also search for this author inPubMed Google Scholar * Piotr Kittel View author publications You can also search for this author inPubMed Google Scholar CONTRIBUTIONS

O.A.-O. developed the concept of the paper, conceived the study, and contributed chironomid data. D.O. contributed geochemical data. A.K. and R.I.B. provided statistical analyses. P.K.

supervised the project, analysed lithology of the cores and the geomorphology of the area. P.K. and J.S. conducted the fieldwork at the study site. J.S. analysed the archaeological context.

M.K. provided radiocarbon and dendrological dates. O.A.-O., P.K., D.O. and M.P. interpreted the data and wrote the manuscript with contributions from all authors. O.A.-O., J.S., A.K., D.O.

and P.K. created the figures and tables. All authors reviewed the manuscript. CORRESPONDING AUTHOR Correspondence to Olga Antczak-Orlewska. ETHICS DECLARATIONS COMPETING INTERESTS The

authors declare no competing interests. ADDITIONAL INFORMATION PUBLISHER'S NOTE Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional

affiliations. SUPPLEMENTARY INFORMATION SUPPLEMENTARY INFORMATION. RIGHTS AND PERMISSIONS OPEN ACCESS This article is licensed under a Creative Commons Attribution 4.0 International License,

which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a

link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence,

unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory

regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit

http://creativecommons.org/licenses/by/4.0/. Reprints and permissions ABOUT THIS ARTICLE CITE THIS ARTICLE Antczak-Orlewska, O., Okupny, D., Kruk, A. _et al._ The spatial and temporal

reconstruction of a medieval moat ecosystem. _Sci Rep_ 12, 20679 (2022). https://doi.org/10.1038/s41598-022-24762-w Download citation * Received: 11 April 2022 * Accepted: 21 November 2022 *

Published: 30 November 2022 * DOI: https://doi.org/10.1038/s41598-022-24762-w SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get shareable

link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative