Search

Data supporting the manuscript Automatic flower detection and phenology monitoring using time-lapse cameras and deep learning. The dataset contains images and annotations used for training and testing an automatic flower detection model (Mask RCNN) as well as two trained models., Advancement of spring is a widespread biological response to climate change observed across taxa and biomes. However, the species level responses to warming are complex and the underlying mechanisms difficult to disentangle. This is partly due to lack of data, which is typically collected by repeated direct observations, and thus very time-consuming to obtain. Data deficiency is especially pronounced for the Arctic where the warming is particularly severe. We present a method for automatized monitoring of flowering phenology of specific plant species at very high temporal resolution through full growing seasons and across geographical regions. The method consists of image-based monitoring of field plots using near-surface time-lapse cameras and subsequent automatized detection and counting of flowers in the images using a convolutional neural network. We demonstrate the feasibility of collecting flower phenology data using automatic time-lapse cameras and show that the temporal resolution of the results surpasses what can be collected by traditional observation methods. We focus on two Arctic species, the mountain avens Dryas octopetala and Dryas integrifolia in 20 image series from four sites. Our flower detection model proved capable of detecting flowers of the two species with a remarkable precision of 0.918 (adjusted to 0.966) and a recall of 0.907. Thus, the method can automatically quantify the seasonal dynamics of flower abundance at fine-scale and return reliable estimates of traditional phenological variables such as onset, peak, and end of flowering. We describe the system and compare manual and automatic extraction of flowering phenology data from the images. Our method can be directly applied on sites containing mountain avens using our trained model, or the model could be fine-tuned to other species. We discuss the potential of automatic image-based monitoring of flower phenology and how the method can be improved and expanded for future studies.

This is chapter 2 of the State of Environmental Science in Svalbard (SESS) report 2021. Predicted temperature increases in the Arctic are expected to fundamentally alter tundra ecosystem dynamics. The Arctic’s extreme year-to-year and place-to-place variability make long-term monitoring challenging, yet essential for environmental conservation, management and policy making. COAT has developed a framework that addresses these complex issues using a holistic, ecosystem-based adaptive approach. This is achieved by integrating data on the state of various characteristics of the ecosystem measured at relevant sites and relevant times to reach clearly defined goals and targets for monitoring the terrestrial food web. For this reason, COAT Svalbard is an essential component of the Svalbard Integrated Arctic Earth Observing System (SIOS). COAT Svalbard contains six monitoring modules, with study sites in two contrasting regions in Svalbard, Nordenskiöld Land (inland) and Brøggerhalvøya (coastal). Five of the modules focus on the Svalbard food web — vegetation, Arctic fox, geese, ptarmigan and reindeer. The sixth module, a climate-monitoring network with full-scale operational weather stations and associated infrastructure, has now been fully implemented. Svalbard’s tundra ecosystems have undergone rapid and substantial changes in climatic conditions — manifested particularly as rising surface temperatures, longer and warmer growing seasons, shortening of the snow-covered season and rising permafrost temperatures. Currently, monitored vertebrate populations appear to be stable or increasing in these regions. Long-term monitoring of vegetation communities is being implemented and will enhance understanding of bottom-up processes in the terrestrial food web.

Kleiven, E.F., Framstad, E., Bakkestuen, V., Böhner, H., Cretois, B., Frassinelli, F., Ims, R.A., Jepsen, J.U., Soininen, E.M. & Eide, N.E. 2022. Ny nasjonal smågnagerovervåking i fjell basert på kamerafeller. Forslag til innsamlingsdesign og dataprosessering. NINA Rapport 2170. Norsk institutt for naturforskning. Denne rapporten presenterer et forslag til et nytt nasjonalt overvåkingsprogram for smågnagere i fjell basert på kamerafeller. Utredningen er bestilt av Miljødirektoratet. Overordnet målsetting for overvåkingsprogrammet er å bidra med data til smågnagerindikatoren i fagsystemet for økologisk tilstand i fjell. Programmet skal så langt mulig, dekke variasjonen i regionale og lokale klimaforhold, samt andre naturgitte forhold, for å kunne vurdere effekter av påvirkningsfaktorer dersom det oppstår endringer i smågnagerbestandene over tid. Resultater fra programmet skal kunne brukes i vurderingen av bevaringstiltak for fjellrev, snøugle og dverggås. Ny nasjonal smågnagerovervåking foreslås basert på vel 50 lokaliteter som dekker alle sentrale fjellområder og et utvalg av mer perifere fjellområder i øst, vest og langs kysten. Disse lokalitetene vil gi bestandsdata for smågnagere som tilfredsstiller alle formålene med overvåkingsprogrammet. Lokalitetene er videre prioritert ut fra hvor godt de dekker overvåkingsformålene og i hvilken grad det allerede finnes dataserier for lokale smågnagerbestander i nærheten. Tolv lokaliteter med prioritet I dekker i hovedsak sentrale fjellområder fra sør til nord, viktige områder for fjellrev og har eksisterende dataserier for smågnagere. Seksten lokaliteter med prioritet II supplere de foregående, spesielt ved å dekke større deler av klimagradienten fra oseaniske til kontinentale områder. Tjuefem lokaliteter med prioritet III utfyller de foregående ved å øke dekningen av ulike fjellområder. Endelig plassering av lokalitetene må vurderes etter feltbefaring. Innen hver overvåkningslokalitet foreslås et romlig hierarkisk overvåkningsdesign. På lokalitetsnivå bør overvåkningen dekke de sentrale habitatene (lynghei, tuemark, snøleier) for de viktigste smågnagerartene i fjell: gråsidemus, lemen, markmus/fjellmarkmus. Samtidig bør både lav- og mellomalpin sone dekkes. Innen hvert habitat og hver klimasone anbefales seks kamerafeller (romlige replikater), dvs. i alt 36 kamerafeller pr. lokalitet. Dette vil med stor sannsynlighet være tilstrekkelig til å få en god indeks for bestandsstørrelse for alle de viktige smågnagerartene for hvert habitat og klimasone innen hver overvåkningslokalitet, selv med noe datatap som følge av tekniske feil. Den fulle versjonen av overvåkingsprogrammet bør omfatte de aller fleste av de foreslåtte lokalitetene. Noen få lokaliteter som anses å dekke marginale fjellområder kan ev. utelates. For en nedskalert versjon av overvåkingen anbefales å bruke alle lokaliteter med prioritet I eller II. En minimumsversjon av overvåkingen bør omfatte lokalitetene med prioritet I, ev. supplert med noen med prioritet II som er mest relevante for arbeidet med å bevare fjellrev. I utgangspunktet foreslås ingen reduksjon av overvåkingsdesignet pr. lokalitet, men det anbefales en evaluering av programmet etter fem års drift, for å vurdere om det er forsvarlig å redusere antall kamerafeller pr. lokalitet. Andre initiativer som kan gi supplerende informasjon om utbredelsen av smågnagertopper, er gjennomgått og vurdert. Det vil kreve nærmere vurdering av disse initiativenes observasjonsinnsats og metoder for å kunne vurdere nytten av slik informasjon for det foreslåtte overvåkingsprogrammet. I rapporten beskrives fullstendig arbeidsflyt fra innhenting av minnekort til et ferdig kvalitetssikret og tilgjengeliggjort datasettsett med bestandsindeks for ulike smågnagere. Denne arbeidsflyten kombinerer maskinlæring for automatisk bildeklassifisering med manuell kontroll av utvalgte bilder. Det vil i oppstarten være behov for tilrettelegging av treningsdata til utvikling av en ny maskinlæringsmodell, samt kvalitetssikring av modellklassifiseringene. Overvåkingsprogrammet bør koordineres av en faginstitusjon som støttes av et bredere faglig råd. Driften av programmet i felt bør i størst mulig grad koordineres med eksisterende feltapparat (f.eks. SNO) for å oppnå både logistiske og faglige synergier med andre overvåkingsprogrammer. Det er videre estimert en budsjettramme for de tre alternative overvåkingsoppleggene, full versjon, nedskalert versjon og minimumsversjon. Noe utviklingsarbeid vil kreves i oppstarten av programmet. Det er fremfor alt store investeringskostnader knyttet til kamerainnkjøp, deretter kostnader til årlig drift varierende fra litt under 2,5 millioner i en minimumsversjon til nesten 11 millioner i full versjon. Kleiven, E.F., Framstad, E., Bakkestuen, V., Böhner, H., Cretois, B., Frassinelli, F., Ims, R.A., Jepsen, J.U., Soininen, E.M. & Eide, N.E. 2022. New national monitoring of small rodents in Norwegian Arctic and Alpine tundrabased on camera traps. Proposed sampling design and data processing. NINA Report 2170. Norwegian Institute for Nature Research. This report presents a proposal for a new national monitoring program for small rodents in mountains, based on camera traps. The report was commissioned by the Norwegian Environment Agency. The overall objective of the monitoring programme is to contribute data to the small rodent indicator for the assessment of ecological condition in tundra ecosystems. The programme shall, as far as possible, cover variations in regional and local climate conditions, as well as other natural conditions, to assess the effects of impact factors on possible changes in small rodent populations over time. Results from the programme can also be used in the assessment of conservation measures for Arctic foxes, snowy owls and lesser white-fronted geese. The new national small rodent monitoring is proposed to be based on over 50 sites covering all central mountain areas and a selection of more peripheral alpine areas in the east, west and along the coast. These sites will provide population data for small rodents that satisfy all the objectives of the monitoring programme. The sites are also prioritised based on how well they cover the monitoring objectives and whether data series for small rodent populations are available in the vicinity. Twelve priority I sites mainly cover central alpine areas from the south to alpine and Arctic areas in the north, important areas for Arctic foxes and have existing data series for small rodents. Sixteen priority II sites supplement the previous ones, particularly by covering larger parts of the climate gradient from oceanic to continental areas. The twenty-five priority III sites further increase the coverage of various mountainous areas. The final location of the sites must be considered after field inspection. Within each site, a spatial hierarchical monitoring design is proposed. At the site level, monitoring should cover the central habitats (heathlands, tussocks, snow beds) for the main small rodent species in Norwegian tundra ecosystems (grey-sided vole Myodes rufocanus, Norwegian lemming Lemmus lemmus, field vole Microtus agrestis and tundra vole Microtus oeconomus). Both the low and middle alpine zones should be covered. Within each habitat and climate zone, six camera traps (spatial replicas) are recommended, i.e., a total of 36 camera traps per site. This will most likely be sufficient to obtain a good population size index for all important small rodent species for each habitat and climate zone within each monitoring site, even with some data loss due to technical malfunction. The full version of the monitoring programme should include most of the proposed sites. A few sites considered to cover marginal mountain areas may be omitted. A scaled-down version of the monitoring should include all sites of priority I or II. A minimum version should include sites of priority I, possibly supplemented by some of priority II that are most relevant for conservation of Arctic foxes. Initially, no reduction in the monitoring design per site is proposed, but an evaluation of the program after five years of operation is recommended to assess whether it is prudent to reduce the number of camera traps per site. Other initiatives that may provide supplementary information on the distribution of small rodent peaks have been reviewed and assessed. Further assessment of these initiatives' observation efforts and methods will be required to assess the benefits of such information for the proposed monitoring programme. The report describes the complete workflow from collecting memory cards to a ready-made quality-assured and accessible dataset set with abundance indexes for various small rodent species. This workflow combines machine learning for automatic image classification with manual control of selected images. Initially, adaptation of training data for the development of a new machine learning model will be needed, as well as quality assurance of the model classifications. The monitoring programme should be coordinated by a research institution supported by a broader scientific council. The field operation of the programme should as far as possible be coordinated with existing field activities (e.g., by the Norwegian Nature Inspectorate), to achieve both logistical and scientific synergies with other monitoring programmes. An overall budget has been estimated for the three alternative monitoring schemes, the full version, the scaled-down version, and the minimum version. Some development work will be required at the start of the programme. Initial costs associated with camera procurement are particularly large, followed by annual operation costs varying from just under 2.5 million NOK in a minimum version to almost 11 million NOK in full version.

Jepsen, J.U., Speed, J.D.M., Austrheim, G., Rusch, G., Petersen, T.K., Asplund, J., Bjerke, J.W., Bjune, A.E., Eide, N.E., Herfindal, I., Ims, R.A., Israelsen, M.F., Kapfer, J., Kolstad, A.L., Nordén, J., Sandercock, B., Stien, J., Tveito, O.E., Yoccoz, N.G. 2022. Panel-based Assessment of Ecosystem Condition – a methodological pilot for four terrestrial ecosystems in Trøndelag. NINA Report 2094. Norwegian Institute for Nature Research. The panel-based assessment of ecosystem condition (PAEC) is an evidence-based approach to assess the condition of Norwegian ecosystems. The assessment is carried out by an expert panel with broad expertise in the ecosystems to be assessed and is inspired by approaches used in international assessments such as IPCC and IPBES. The assessment follows an earlier developed protocol. In this report, PAEC is piloted for major terrestrial ecosystems in the county of Trøndelag; forest, alpine, open-lowlands, and wetlands. For each ecosystem, a list of indicators of change in ecosystem condition in response to anthropogenic drivers is developed. The indicators fall within seven main ecosystem characteristics: primary production, biomass distribution among trophic levels, functional groups within trophic levels, functionally important species, biological diversity, landscape ecological patterns, and abiotic factors. The expected change in indicators in response to anthropogenic drivers are termed phenomena, and their selection is based on published literature, including reference to the confidence of a change being observed in response to anthropogenic drivers and the mechanism leading to a deterioration in ecosystem state. Datasets to quantify each indicator are identified and collated and the quality of each dataset is assessed in terms of its spatial and temporal appropriateness. In the first assessment step, the validity (VP) of each phenomenon is scored and used to infer confidence in the causal relationship between changes in the indicator and anthropogenic drivers. The next step is an evaluation of the biological and statistical significance of the evidence for the occurrence of each phenomenon, termed evidence (EP) of the phenomenon. The third step is a consolidated assessment of the ecological state based on the associated indicators and phenomena, first for each ecosystem characteristic, and subsequently for the ecosystem as a whole. The assessment is based on the validity, the quality of the evidence, and the data quality for each phenomenon. This provides a qualitative assessment of deviation from the reference condition of “no deviation”, “limited deviation” or “substantial deviation”. The assessments are each supported by narrative accounts. The pilot assessment involved analysis of 24 datasets documenting 41 indicators. Several indicators were included in multiple ecosystems. In total there were 27 indicators used for forest ecosystems, 24 for alpine ecosystems, and 16 in each of wetlands and open lowlands. In the forest ecosystems, substantial deviation from the reference condition was identified for five of the ecosystem characteristics. The two exceptions were primary productivity where there was a limited deviation from the reference condition, and biological diversity where there was no deviation from the reference condition (but the latter was based on a single indicator and hence an entirely inadequate indicator coverage). The deviations were found primarily in climatic variables, cervids and their forage and predators, and dead wood. Overall, the forest ecosystem was assessed as having a substantial deviation from the reference condition. In the alpine ecosystems, substantial deviation from the reference condition was identified for the abiotic ecosystem characteristic, largely attributed to indicators associated with temperature, seasonality, and snow. Limited deviation from the reference condition was assessed for functionally important species and primary productivity (both based on a partially adequate indicator coverage) and for biological diversity, functional groups within trophic levels and landscape ecological patterns (but these were based on an inadequate indicator coverage). For the ecosystem characteristic biomass distribution among trophic levels, the quality of evidence was insufficient to conclude regarding the condition of the single indicator involved, and no overall assessment of this ecosystem characteristic could be undertaken. Overall, the alpine ecosystem was assessed as having limited deviation from the reference condition. For both open lowland and wetland ecosystems, several ecosystem characteristics were not assessed due to a lack of relevant indicator datasets. For this reason, no overall assessment of the ecosystems as a whole could be undertaken. However, for both ecosystems, there was a substantial deviation from the reference condition for abiotic factors (temperature, seasonality, and snow). In open lowlands, there was a substantial deviation in functionally important species (ungulates) and limited deviation in primary productivity and biological diversity. In wetlands, there was a limited deviation from the reference condition in primary productivity, biological diversity, and landscape ecological patterns. Most challenges encountered during this pilot assessment related to the inadequacy of the datasets for assessing ecosystem condition. Reasons behind this include that ecosystem extents are not adequately mapped, particularly those characterised by small and fragmented patches, and a taxonomic or geographical limitation of datasets and environmental monitoring. The report suggests further development of indicators for operational application. Finally, the report also suggests knowledge needs and prioritisation for further research to support the future implementation of ecosystem assessments in Norway. Norwegian Environment Agency: M-2190 | 2021

This is chapter 2 of the State of Environmental Science in Svalbard (SESS) report 2019 (https://sios-svalbard.org/SESS_Issue2). COAT is a response to the urgent international calls for establishment of observation systems that make it possible to gain insight into how climate impacts Arctic tundra. COAT Svalbard is an essential component of SIOS and serves to optimise and integrate the ecosystem-based terrestrial monitoring. Variations from year to year and differences from place to place make long-term monitoring essential to support the complex decisions involved in conservation, management and policymaking. The COAT approach is holistic, covering entire ecosystems by integrating information about living and non-living factors over time and space, with clearly defined monitoring goals. Monitoring modules within COAT Svalbard track five key climate-sensitive food web pathways as well as climate parameters that determine how those pathways function (www.coat.no). Each module is described in terms of expected direct and indirect relationships between organisms in the food web pathways, and how climate and management interventions might influence these interactions. The programme is implemented according to a peer-reviewed Science Plan with a solid foundation in the scientific literature. COAT Svalbard focuses on two contrasting Arctic regions: Nordenskiöld Land and Brøggerhalvøya and surrounding areas. The vertebrate populations being monitored in these areas currently appear to be stable or growing. However, the lack of long-term monitoring of the vegetation communities on which these animals rely hinders understanding of bottom-up processes within the food web. COAT aims to fill these gaps and provide new insight into how climate change impacts High Arctic tundra ecosystems.

This is the peer reviewed version of the following article: Metcalfe, D.B., Cherif, M., Jepsen, J.U., Vindstad, O.P.L., Kristensen, J.Å. & Belsing, U. (2018). Ecological stoichiometry and nutrient partitioning in two insect herbivores responsible for large-scale forest disturbance in the Fennoscandian subarctic. Ecological Entomology, 44(1), 118-128, which has been published in final form at https://doi.org/10.1111/een.12679. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Use of Self-Archived Versions. 1. Outbreaks of herbivorous insects can have large impacts on regional soil carbon (C) storage and nutrient cycling. In northernmost Europe, population outbreaks of several geometrid moth species regularly cause large‐scale defoliation in subarctic birch forests. An improved understanding is required of how leaf C and nutrients are processed after ingestion by herbivores and what this means for the quantity and quality of different materials produced (frass, bodies). 2. In this study, larvae of two geometrid species responsible for major outbreaks (Epirrita autumnata and Operophtera brumata) were raised on exclusive diets of var. czerepanovii (N. I. Orlova) Hämet Ahti and two other abundant understorey species (Betula nana, Vaccinium myrtillus). The quantities of C, nitrogen (N) and phosphorus (P) ingested and allocated to frass, bodies and (in the case of C) respired were recorded. 3. Overall, 23%, 70% and 48% of ingested C, N and P were allocated to bodies, respectively, rather than frass and (in the case of C) respiration. Operophtera brumata consistently maintained more constant body stoichiometric ratios of C, N and P than did E. autumnata, across the wide variation in physico‐chemical properties of plant diet supplied. 4. These observed differences and similarities on C and nutrient processing may improve researchers' ability to predict the amount and stoichiometry of frass and bodies generated after geometrid outbreaks.