eDNA has thus far been mainly used in the early detection and monitoring of invasive species, contributing to the increase of IAS presence records. The use of eDNA techniques could facilitate a more effective method for recording IAS absence than do regular monitoring surveys or possibly may aid in the compilation of independent data similar to the approach used for proving (non)successful eradications (Dejean et al., 2012). Currently, eDNA research is focusing its effort on the species detection efficiencies based on the competence of sampling, amplification and sequencing techniques. A detailed review has been conducted based on the potential for the future application of eDNA tool by identifying the proportion of positive detections of IAS within individual research (Table 1). The review proves how useful the tool can be dealing with IAS detection. A recent increase in presented eDNA research conducted on invasive species is only the tip of the iceberg of what can be achieved for conservation and IAS management. There is however a number of limitations that should be remembered before applying eDNA data to retrieve distribution data for SDMs.

Freshwater ecosystems, lentic, and lotic, provide excellent study area for defining the wide range of detection possibilities of eDNA techniques as well as the limitations. Small-scale freshwater lentic bodies provide an excellent opportunity to study eDNA characteristics related to degradation, which can affect successful detectability of species. Recent studies have tried to underline degradation rates in correlation to abiotic factors, such as, (i) most effective water stratum for eDNA detection (Moyer et al., 2014), (ii) pH, UV-B (Strickler et al., 2015), (iii) effects of temperature on eDNA degradation (Strickler et al., 2015; Eichmiller et al., 2016), and (iv) temporal effects (Dejean et al., 2011). Freshwater lotic bodies can provide important information due to their longitudinal downstream dynamics, such as, (i) eDNA persistence in the environment (Jerde et al., 2016; Wilcox et al., 2016), (ii) residence time of eDNA (Jerde et al., 2016), and (iii) the ecology of eDNA (Barnes and Turner, 2016). In case of newly introduced IAS, measures of low abundances present another limitation (Jerde et al., 2011) which is highly important when discerning between presence and absence records. Some of the reported examples are applied to non-invasive species, but the reason why we focus on IAS is that time, i.e., rapid response, is key to management, so that an identified IAS can be eradicated/controlled before any negative ecosystem impact occurs. Since eDNA can assist in more rapid detection and early response to IAS invasions than traditional sampling, this technology most greatly benefits identification of invasive species.

All the limitations of eDNA that are currently being studied are crucial for IAS assessment. When monitoring, especially in a new environment, it is fundamental to detect it at extremely low abundances and report negative or positive presence. False positives and negatives are essentially relevant for their use within SDM and cannot be misjudged, whether they are products of sampling bias or metabarcoding bioinformatics pipeline. The distribution patterns and biology of the eDNA is another important factor influencing the accuracy of information which is relevant for the distribution of IAS within the models. The accuracy that we can obtain through eDNA highly depends on the strategies followed during the fieldwork and through laboratory protocols. In order to more accurately state the proportion of the positive (or negative) detections, independent observations (Steel et al., 2013) would need to become an essential part of eDNA studies to overcome the bias of false positives or negatives. An increased eDNA sampling effort based on a temporary scale would provide a more accurate proportion of positive (negative) detections and should be replaced by research proposed on a single sampling events (Simmons et al., 2015; Fujiwara et al., 2016; Hänfling et al., 2016). Independent observations would need to become a necessary procedure especially when dealing with estimations of newly introduced species (Jerde et al., 2011) or dealing with the estimations of successful eradication measures (Dunker et al., 2016).

To avoid bias due to inconsistent use of eDNA tools a minimum information based on field and laboratory procedures should always be reported and presented in a consistent manner as presented by (Goldberg et al., 2016). Pioneers in eDNA research (Ficetola et al., 2016) highly recommend following general requirements such as, precautionary approach to avoid contamination, respecting a general practice of obtaining control samples, extraction blanks, as well as incorporating PCR positive and negative controls. In cases of individual species assessment, parallel mesocom experiments are highly recommended in order to be able to estimate the limitations of detectability for each individual primer set. Another method to assess limitations of primer detections is assessing detectability of the species “in time” after its removal from the controlled environment. When working on multiple species assessment using a metabarcoding approach, it is recommended, to sequence the control samples, compare the sequencing control outputs with the actual samples, and if none of the last achieve high quality sequence reads by appropriate clean up steps; removal of singletons, chimeras, as well as including a record of removed sequences (Deiner et al., 2017a). Bias due to universal primer preferential amplifications of species can alter the relative abundance of individual species eDNA (Deiner et al., 2017a). A PCR-free method, namely sequence capturing offers promising solutions in order to avoid amplification bias (Shokralla et al., 2016).

In terms of IAS certainty of existence in a non-native environment, false- positive and false- negative are crucial points for management and environmental policies (Moyer et al., 2014; Lahoz-Monfort et al., 2016). Even low rate false- positives pose a bias toward species specific occupancy (Lahoz-Monfort et al., 2016). Errors produced during PCR and sequencing are main source of bias for false- positives whereas false- negatives normally appear due to bias during sampling. Sampling and PCR replicates are key to avoid obtaining false presence and absence and should be routinely corrected with the appropriate statistical tools referred to “site occupancy-detection modeling” (SODM) (Lahoz-Monfort et al., 2016). The SODM model shows precise estimation of the probability for the site occupancy, including overall probability of detection at sites where the species is present. The model provides unbiased estimation of occupancy when properly applied using large amount of initial data, even with a smaller amount of replications. Researchers (Ficetola et al., 2016) adopting SODM as part of their eDNA pipeline, give advice to avoid referring to single occurrences within one sample as reliable ones. Precautionary measures should be taken up before coming to conclusions that non-detection of species corresponds to species absence, and in converse that detections directly relies to species presence (Roussel et al., 2015) simply due to eDNA characteristics, such as potential longevity. In order to overcome the frontiers of eDNA techniques and to make it generally applicable within the SDM the above consistency is pivotal within the immense growing body of eDNA literature.

The method appears to be highly efficient on bony fish and amphibians with successful spatial representativeness in lotic and lentic systems (Civade et al., 2016). It has been shown that the eDNA samples are able to overcome spatial autocorrelation biases (Deiner et al., 2016) which are normally a result of conventional biodiversity assessments. eDNA seasonal diversity at the ecosystem scales (Bista et al., 2017) are key for more holistic understanding of the successful invasions of species within SDMs.

There are many possibilities of using eDNA for SDMs but currently one of the most important novel uses is a more precise sampling of absences which is sometimes difficult or impossible to obtain (Nezer et al., 2017). As commented, the information regarding species existence in certain system measured through eDNA can be susceptible to certain bias, due to eDNA characteristics. However, there exist approaches within the spatial modeling that might be applied to deal with the uncertainties from eDNA results. For instance, Dudík et al. (2006) presented the di-bias approach, which gives a higher weight in the models to those localities where presences or absences are more reliable. In the same way, those localities where eDNA is less reliable can receive a lower weight in the models, such weighting might correspond with the reported detection rates (Table 1). Therefore, there are possibilities from the SDMs to deal with the potential bias arising from using eDNA as a sampling technique which encourage its use despite current relative limitations. The ability to cope with the limitations and strength of the combination of these distinct research fields will benefit from the collaboration between molecular ecologists and modelers contributing to the evolution of two scientific disciplines (Coccia and Wang, 2016). Other disciplines apart from invasion ecology (e.g., biogeography or spatial ecology) might also benefit from future development of molecular ecology tools as a sampling technique. Thus, we highly recommend involving eDNA analysis into spatial models to predict future invasions and many other ecological processes. For example, targeting IAS hot spots and vectors of introduction, is a perfect starting point for detection of IAS and estimation of their future dispersal within the SDMs. Spatial representativeness of IAS within the SDMs is key to understanding the ecology behind their successful dispersal and the management of invasions.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling Editor declared a past co-authorship with one of the authors ET.