4.1. Layout of the Tank Cascade Systems
The four tank cascade systems that underwent analysis are associated with the type of linear cascades in which all tanks are aligned in a row [10
]. The number of tanks per cascade varies between one and eleven. The investigated tank cascade systems are situated in first-order, shallow saucer-shaped valleys, which are regarded as not being prone to erosion. Consequently, the geomorphological conditions of the study area seem to have favored the construction and sustainable function of these systems. All systems provide irrigation water for paddy cultivation west of the floodplain of the Malwathu Oya River (Figure 1
The tanks are embedded in a cultural landscape setting, which is regarded as typical for village tank cascade systems. In an idealized model, this setting is composed of four concentrically arranged zones, namely the tank, the paddy fields, the settlement areas and the surrounding shrub lands and forests [34
]. Each of these zones fulfills specific functions and is traditionally used for different purposes. All four zones are represented in the investigated tank cascade systems. In the following, local particularities are briefly discussed.
During both field campaigns in April 2010 and September 2011, the majority of tanks, except the single tank in system 1, were desiccated, reflecting weather conditions during previous months, with an unusually dry rainy season in spring 2010 and the end of the dry season in September 2011.
Various observations like soil erosion and damage to the dams and sluices indicated that the maintenance measures were neglected. Cultivation of rice or other plants with high temporal water consumption is known to be a local form of subsequent use of desiccated tank bodies [9
]. However, due to the prohibition of the cultivation of tank bodies by the government, this phenomenon could not be observed in the study area [7
]. In contrast, the traditional use of desiccated tank beds as pasture during the dry season is still practiced in the four systems investigated.
A typical feature of village tank ecosystems is a wetland area, called thaulla in Sinhala, which can be found upstream of the tank body at the root of the discharging channel. Runoff and channeled irrigation water from upstream tanks pass this swampy area in which aquatic plants force the deposition of fines [36
]. Such thaulla areas could not be identified for the investigated tank cascade systems. It is assumed that the intensive irrigation agriculture causes relatively quick consumption of the stored water, which in turn prevents the development of an upstream wetland. In addition, due to increased population pressure, thaulla areas have been increasingly removed from all over Sri Lanka’s dryland zone during the last decades [35
Desiltation of tanks is practiced regularly by local villagers as a maintenance measure to sustain the storage capacity of the tanks [16
]. The extracted sediments are utilized as raw material for bricks or as fertilizer for the fields [13
]. Revisiting system 2 in 2014 and interviews with local villagers showed that sediments had in the recent past also been illegally exploited using heavy machinery and utilized as construction material (Figure 5
) (personal interview with a local villager, 23 September 2015) [37
]. This practice affects the functionality of the tank cascade system as the deepening of the tank bed lead to an exposure of the surface near regolith aquifer. Consequently, this tank bed does not fall dry anymore and cannot be used as pasture during the dry season.
The ratio of tank command area (Acoa) to tank water-spread area (Awa) is a measure for the efficiency of tank cascade systems. Ideally, this ratio should be ≥1.0 in order to provide enough storage capacity for irrigation [10
]. With a ratio of 1.0, this optimal condition is only present in system 2.
The north-central dry zone of Sri Lanka is characterized by a dispersed settlement pattern. This pattern is also reflected in the distribution of settled areas in the investigated tank cascade systems, in which various small settled areas occur. In all systems, settled areas are situated in close vicinity to tanks and the associated paddy fields. Archaeological findings in system 2 indicate a (pre)historic utilization of this area [33
]. Chena cultivation, a slash-and-burn practice, is applied in all systems. It is widely assumed that the beginning of shifting cultivation dates prior to the onset of paddy cultivation. To intersperse chena cultivation with irrigation-based rice cropping is typical for the dry zone in Sri Lanka and is practiced until today ([38
] and literature cited herein). Remarkable is the chena cultivation in the command area of tank 2 in system four, as traditionally this slash-and-burn practice is not based on irrigation.
All investigated tank systems, with the associated agricultural fields and settlement areas, are embedded in open deciduous forests or scrubland, which are utilized as pasture for cattle and for logging. A distinction between parkland and forested areas as proposed by Tennakoon could not be observed [34
4.2. Sediment Characteristics
Sediment texture and macroscopic sediment character are used as proxies to differentiate between allochtonous sediments and the autochthonous weathered bedrock (saprolite) [16
]. The weathering grade of the parent metamorphic bedrock is locally highly heterogeneous, resulting from the petrographic character and small scale changes in environmental conditions [39
]. The transition between saprolite and superimposing allochtonous deposits is characterized in profiles NA01 and NA03 by a diffuse change from gravels in a matrix of sandy loam to alternating layers of sandy and loamy material at a depth of 170 cm b. s. (profile NA01) and 200 cm b. s. (profile NA03). In profile NA02, the saprolite was identified at a depth of 140 cm b. s. based on a diffuse change from sandy loam to coarse sandy loam. A distinct boundary as described in a previous study about tank sediments in the Anuradhapura district was not observed in the extracted sediments [42
Two different types of sediment facies can be expected to overlie the in-situ weathered saprolite: Facies (a) is associated with fluvial deposits in an alluvial plain dating to a phase prior to tank construction, and facies (b) is expected to be characterized by siltation in the still water body of the tanks [42
]. Tank sediments can be assumed to be of varying origin due to human impact, such as soil erosion in consequence of tillage or cattle watering [43
]; furthermore, this material may also result from erosional processes at the slopes. In a previous study, the sand/(silt + clay) ratio was successfully applied as a proxy to characterize the introduced facies for tanks in the Anuradhapura district ranging in size between 23 ha and 15 ha [42
]. Contrastingly, in the present study it was not possible to distinguish between the two facies by this proxy. Another study on tank sediments c. 30 km east of Anuradhapura showed that consideration of bulk chemical and physical sediment parameters like electric conductivity, pH-value, total organic and inorganic carbon, chemical parameters and magnetic susceptibility also do not enable a differentiation between autochthonous saprolite and allochtonous sediments to be made [16
]. This might be attributed to the relatively small size of tanks investigated. Tanks 5 and 6 have an area <1 ha; unlike larger tanks they desiccate throughout the dry season and are affected by tropical soil formation processes [44
]. A resulting textural degradation of layering as well as the mixture of chemical composition, especially within the uppermost meter, is primarily caused by bioturbation by plant roots, micro-organisms and soil fauna as well as pediturbation due to the shrinking and swelling of clays and their translocation in desiccation cracks [45
The precipitation of carbonates and the frequent occurrence of oxidation and reduction marks reflect regular changes of hydromorphous conditions associated with the periodical desiccation of the tanks [46
]. The occurrence of calcite concretions is a common feature in the soils of the dry zone [47
]. Dissolved calcium originating from the weathering of feldspars is transported laterally in the subsoil, gets enriched and precipitates as calcareous concretion in the poorly drained soils of the valley bottom, receiving its carbonates from conversion of CO2
-containing soil water [48
]. Reduction and oxidation marks in profiles NA01 and NA03 indicate seasonal variations of the groundwater table [44
]. Precipitation of sesquioxide is an indicator for oxidizing processes linked with the desiccation of the tanks [49
]. Other factors affecting sedimentation conditions and post-sedimentary alteration of the tank deposits are related to the above-mentioned traditional subsequent use of the desiccated tanks during the dry season as paddy fields, livestock pasture or a source for construction material. Furthermore, the desiltation of tanks leads to a considerable removal of material out of the reservoirs, which is also reflected in the hiatus of younger sediments.
Carbon from bulk sediment samples originating from tank sediments was used for radiocarbon dating, as extracted sediments did not contain macro plant remains or charcoal. Here the different origins of the dated carbon, e.g., soil organic matter and—in the context of lake sediments—old-carbon reservoirs (e.g., shales, bedrock), need to be taken into account [50
]. Soil organic matter is regarded as a product of continuous processes of accumulation and decomposition [51
]. Resulting AMS ages are therefore frequently interpreted as minimum ages for soil formation [52
]. Dating of carbon from old reservoirs is associated with the “hardwater effect” leading to an overestimation of resulting ages [50
]. As carbonaceous bedrock is absent in the study area, and carbonates precipitation corresponds to a secondary precipitation of carbonates, it is concluded that the calcites found are authigenic or correspond to remnants of organisms like shells or bones [55
]. Additional sources for errors, resulting in an underestimation of ages, may include a vertical relocation of humic acids, an incomplete removal of roots or rootlets and bioturbation [56
]. Ages measured are in stratigraphic order and match the pattern presented for a comparable system 30 km east of Anuradhapura [16
]. Taking the uncertainties mentioned into account, the ages provide a rough Holocene chronology for the introduced profiles.
The tank cascade systems are deeply interwoven in the socio-economy and landscape household in the central dry zone of Sri Lanka. In this paper, the layout and function of four traditional village tank cascade systems is presented. The results contribute to a documentation of this traditional water-harvesting technique, which built the foundation for irrigation agriculture not only in the dry zone of Sri Lanka, but also in South India since several hundred years. Furthermore, an analysis of tank sediments allows conclusions on the buffering capacity of tank cascade systems against past socio-economic developments (e.g., the abandonment of the ancient capital of Anuradhapura) and climate changes.
Abandoned land management systems have a tendency to shift from depositional into erosional environments after the neglect or cessation of maintenance measures [58
]. Such a transition is not reflected in the sedimentological record from tank cascade system 2. The radiocarbon sample Poz-37935 dating to 1110–1210 CE (profile NA03) matches with the period after the abandonment of Anuradhapura in 1017 CE (Table 6
). Although the city lost its status as capital of the Anuradhapura kingdom, archaeological evidence indicates a continuation of the occupation and use of its hinterland [60
]. It is reported for this period that reservoirs and canals were affected by siltation and fell partly into disuse [61
]. Climatically, this phase corresponded to the Medieval Warm Period, characterized in South Asia by alternating intervals of stronger and weaker monsoonal intensities [15
]. This increase in climatic dynamics is also not reflected in the sedimentological record of profile NA03, as there is a lack of coarser detritus corresponding to more dynamic depositional environments. In summary, the sediment facies dated to 1110–1210 CE clearly indicates siltation processes in a depositional environment. This could be tentatively interpreted as evidence, that these decentral managed tanks were not affected by severe erosion during a period, which was characterized by socio-economic instability and increased climatic fluctuations. The presented results underline the significance of small-scale tank cascades to buffer effects related to climatic fluctuations. Despite being exposed to fluctuations of monsoonal intensities and socio-economic change and facing periods of decay, e.g., in the 14th century CE [3
], they have been in use for some 2000 years and continue to meet a variety of water demands in the north-central dry zone of Sri Lanka. The integration of traditional techniques and knowledge in planning their future sustainable development has great potential to anchor these systems as a cornerstone in coping with future climate change scenarios.