NORTHERN FOREST: A multi-driver framework for near-term iterative forecasting of ecosystem states

The state of an ecosystem is determined by an interplay between biological, physical and human factors. Due to anthropogenic climate change, this interplay is changing in many systems. Since these changes are rapid and the interplay is complex, it is difficult to predict how ecosystem state will change in the long term. Frameworks for near-term forecasting of changes in ecosystem state are therefore all the more valuable. Such frameworks facilitate planning and adaptation and can thus be of great help to managers and other stakeholders. The aim of NORTHERN FOREST has been to develop a model-based routine for near-term forecasting of insect outbreaks and forest damage in northern-boreal mountain birch forest. Outbreaks by geometrid moths have been an increasing problem in this ecosystem during the last 20 years because climate warming has allowed southern moth species to expand their ranges northwards and eastwards. In NORTHERN FOREST, we have combined existing long-term time series with newly developed data products to build statistical models that can forecast the risk of outbreaks and forest damage 1-3 years into the future. The project has focused on the mountain birch forest in Finnmark and was developed in collaboration with the County governor of Troms and Finnmark, the Director of Forestry in Finnmark and the Finnmark Estate. The project has sought to facilitate dialogue between researchers and users about the structure an interpretation of forecasting models, and how near-term forecasts can lead to management action. This has been achieved through regular electronic meetings with the users, and annual workshops that gather researchers and users for presentation and interpretation of project results. By the end of the project in October 2023, NORTHERN FOREST has developed climate data on temperature and precipitation for 1980-2022 with a spatial resolution of 100 m and 250 m, covering Troms and Finnmark as well as Nordland (only 250 m). A forest productivity map for Finnmark has also been developed. These products are openly available on the webpages of Norwegian Meteorological Institute and Norwegian Institute for Bioeconomy research, respectively. These are stand-alone products from the project, but also enter the forecasting models for moth outbreaks and forest mortality. The models for outbreaks are based on long-term time series of moth larval abundance collected by the Climate-Ecological Observatory for Arctic Tundra (COAT). The downscaled climate data provides explanatory variables that represent weather and climate throughout the moth life cycle, and machine learning is used to identify the most relevant of these variables based on explained variance in the larval time series. The models are evaluated by cross-validation techniques developed specifically for time series, where the model is first fitted to the earliest years in the series, and we assess how well the future datapoints are predicted when iteratively adding one year at a time. The models for forest mortality are based on COATs regional surveys of forest health in Finnmark, including a survey that was conducted under NORTHERN FOREST in Aaugust 2020. This survey covered large parts of Finnmark and included measurements on about 13 000 birch stems. Yearly variation in satellite-based loss of canopy cover, caused by moth outbreaks, is the main explanatory variable in the forest mortality models, but we also model local site characteristics that are expected to affect forest resilience, including forest cover and terrain wetness, as well as forest productivity and long-term climate extracted from the maps developed in the project. All explanatory variables are extracted from data products that have coverage for all of Finnmark, so that predictions of forest mortality can be generated for the whole region. As satellite measurements of canopy loss are available from year 2000 until present, this allows us to make maps of expected forest mortality both for next year and in the past. This also it possible to estimate the total area experiencing forest mortality during the period. Validation against aerial photographs and independent forest health data shows that the models provide good predictions of variation in forest mortality on the landscape scale, while local predictions for specific sites have greater uncertainty. By the end of the project, the statistical developmental work for the forecasting models is almost complete, and we have started work on a web portal that makes forecasts and area estimates available for managers and the public. We expect this portal to be established on the web pages of COAT (https://www.coat.no/) by the end of 2024. The portal will partially be a practical tool, that allows users to plan based on expected changes in the risk of outbreak and forest mortality, but will also be a channel for outreach, where yearly updates of model predictions can be explored by the public and other researchers.

NORTHERN FOREST har anvendt en protokoll for brukerinvolvering i forskning (Strategic foresight), hvor sentrale brukere av forskningen (her grunneier og skogforvaltningen i Finnmark), involveres allerede før prosjektet omsøkes, og bidrar til å identifisere sentrale kunnskapsbehov som prosjektet skal dekke. Gjennom prosjektet har brukergruppen vært jevnlig involvert i planlegging av aktiviteter og diskusjon av tolkning og tilgjengeliggjøring av forskningsresultater. Brukergruppens innspill har bidratt tydelig til økt relevans av prosjektets aktiviteter for lokal forvaltning, bl.a. gjennom utvikling av romlige produkter som forvaltningen kan ta i direkte bruk (se resultatrapport). Ved avslutningen av prosjektet anses det at brukergruppen bør utvides for å få bredere perspektiver fra forvaltning og praktisk skogbruk i regionen, og videreføres som en fast brukergruppe tilknyttet den langsiktige overvåkningen til Klimaøkologisk Observasjonssystem for Arktisk Tundra (COAT) i nordlig skog og tregrense (https://www.coat.no/Skog-og-tregrense). Prosjektet har bidratt til et taktskifte i COATs forskning/overvåking av bjørkeskogsøkosystemet, fra et tradisjonelt spørsmålsdrevet perspektiv til en mer adaptiv rutine, der overvåkingsdata driver modeller for nærtidsvarsling av sentrale tilstandsvariabler (skogdød, insektutbrudd). Slike modeller gir mulighet for å forutsi utviklingen i en gitt variabel 1-3 år frem i tid, og kan dermed tillate en iterativ prosess, der prediksjoner følges av validering, modellforbedring og nye prediksjoner. Vi har utviklet årlige kart og arealestimater over predikert skogdød i hele Finnmark samt tidsseriemodeller for bjørkemålerutbrudd, som forteller forvaltingen hvor i utbruddssyklus man vil befinne seg neste vekstsesong. Et premiss for utviklingen av disse modellene har vært utviklingen av nye nedskalerte klimadata i prosjektet (se resultatrapport). Disse har nytte langt ut over prosjektets problemstillinger, da klima er en sentral driver av økosystemprosesser i nordlige områder. Prosjektets forskning har blitt satt i internasjonal kontekst via et nytt samarbeid med Wageningen University & Research. Dette samarbeidet har fokusert på strukturelle endringer i skogøkosystemer både i Finnmark og på tvers av hele det boreale biomet (med Nord Amerika som case). Dette samarbeidet vil fortsette i 2024 med ytterligere publisering av prosjektresultater og søkes videreført på lengere sikt med andre finansieringskilder. Gjennom prosjektet har brukere og forskere erkjent at det er nødvendig med en bedre forståelse av drivere av og mekanismer for rekruttering hos både bjørk og alternative treslag (furu) for å utvikle langsiktige scenarier for bjørkeskogsøkosystemets utvikling under et endret klima. I tråd med en Strategic foresight protokoll har vi derfor utviklet et nytt prosjekt (Arctic Forest Futures) som bygger på NORTHERN FOREST og COATs langsiktige overvåking i Øst-Finnmark. Prosjektet er innvilget støtte fra forskningsrådet fra 2024.

This proposal targets ecological impacts of multiple climatically determined stressors on a unique high-latitude ecosystem, the northern-boreal mountain birch forest of northern Fennoscandia. As a first step in a Strategic Foresight protocol, we have collaborated with stakeholders to identify the key stressors of this ecosystem, namely outbreaks by geometrid moths, browsing by ungulates, and climate, the latter acting both directly and indirectly via its impact on the other stressors. Targeting knowledge gaps and needs identified as priorities by stakeholders, our multidisciplinary team takes an integrative approach to understanding the effects of these stressors on the target ecosystem. To address climate impacts, we will utilize new methodology to develop high-resolution spatial climate data for developing models of climatically determined growing conditions, applicable both to producing spatial maps of current forest productivity and to predict future productivity under climate warming. Based on satellite maps and aerial imagery of moth defoliation, we will conduct regional-scale field surveys of forest mortality and regeneration to reveal forest resilience to outbreaks, conditional on site-specific factors, including climate, growing conditions and ungulate browsing. By integrating these research products with existing time series, the project will culminate in the development of a Near-Term (1-3 year) Iterative Forecasting system for moth abundance, defoliation and forest mortality. The system will produce forecasts to aid management of the focal ecosystem, and provide a more general case for how stakeholder involvement and ecological forecasting can aid the stewardship of severely stressed and rapidly changing ecosystems. Annual workshops will secure stakeholders involvement in data analysis, synthesis, and the development of field protocols for stakeholders’ participation in data collection. Annual forecasts will be made publicly available on a web portal.