INTEGRATION OF SPATIAL BIG DATA FOR IMPROVED LONG TERM FOREST RESOURCE PLANNING

Dylan Walshe1,2, Daniel McInerney1, and Kenneth A. Byrne2

1Coillte Forest, Castletroy Business Park, Plassey Road, Limerick.2Department of Biological Sciences, University of Limerick.

Study Area

Figure 1: Age class distribution of Coillte forests in theSlieve Blooms Study Area

Figure 2: Sentinel 2 RGB image of the Slieve Blooms (red)taken on 10/07/2018

Methods

A variety of models will be created and tested e.g.multinomial, RandomForest, k-nn, and SVM.

SoftwareThis project will involve processingspatial Big Data and so will becompleted using a GNU/Linuxdistribution operating system (e.g.Ubuntu) as it facilitates parallelprocessing and offers efficiencies indata processing. All of the images usedwill be processed by free and opensource software such as GDAL, QGIS,Orfeo Toolbox, R, pktools, ARCSI, SNAP,FUSION. The scripts written as part ofthis project will be made availableonline (Github) and allow for replicationof results.

Remsoft is an optimised operationaland scheduling software that is used inCoillte to plan harvesting in tactical (3-5years) and strategic (20-30 years) timeframes. The results from the modellingwill feed into this software and be usedto alter harvesting plans.

ConclusionUsing remote sensing, LiDAR, and fieldinventory datasets, a model forestimating forest parameters will beproduced in order to update stand levelinventory in the Slieve Blooms. Threescenarios pertaining to the impact ofstorms, biomass removal, andencroachment of invasive species willthen be modelled in relation to timberproduction, sustained yield, andecosystem services for state forestsmanaged by Coillte. All three scenarioswill be modelled on two time scales,tactical and then strategic.

AcknowledgementsThis research is funded by the IrishResearch Council’s Employment BasedPostgraduate Programme inpartnership with Coillte 2017 – 2021.

Work Package 1Data

• Acquire• Consolidate• Validate

Modelling• Height• DBH• Volume• Stocking

Uncertainty• Accuracy• Precision• Bias

Work Package 2Storm

Frequency• Decisions• Production• Profile

Biomass Removal

• Nutrientavailability

• Bio-economy

Invasive Species• Birch• Rhododendron

Objectives:1. To develop and optimise statistical models

for the spatial estimation of forestparameters at individual forest stand levelusing multi-temporal airborne laserscanning (LiDAR) data, optical satelliteimagery and field inventory data.

2. Investigate the impact of future modellingscenarios on timber production, sustainedyield, and ecosystem services. The specificscenarios will relate to the impact ofincreased storm frequency, biomassremoval, and encroachment of invasivespecies.

AbstractIreland’s forestry sector delivers keyeconomic, environmental, and social servicesto the public. The sector is expected to play akey role in the transition to a carbon freesociety and aid in mitigating the effects ofclimate change. Achieving this requires thedevelopment of scientifically robust tools tounderpin forest management decisions acrossshort to long term times scales.

This project will use remote sensing datasetsto estimate forest parameters such as treeheight, stem diameter, standing volume, andstocking. The outputs from the models willthen inform the second objective which is toinvestigate the impact of increased stormfrequency, the impact of biomass removal,and changes in silviculture due toencroachment of invasive species onsustainability.

The growing importance of the forestry sector,particularly the increased demand for woodas biomass, and the impacts of increasedstorm frequency makes this project timely andessential as the results of the evaluation ofmodels will be crucial in analysing timbersupply, management decisions, and futureplanning as well as providing novelcontributions to Coillte’s operational forestresource planning work flows.

Study AreaThe Slieve Blooms has been chosen as theproject site because two LiDAR campaignshave been flown (2013 and 2018). The studyarea also has a historical archive of cloud-freesatellite imagery from Copernicus’ Sentinel 2and NASA’s Landsat 8. Sentinel 2 will be usedas a primary source of imagery due to itsspectral resolution and revisit time of ~5 days.It is used for intensive timber production, andprovides a range of eco-system services e.g.amenity, landscape, and biodiversity. The totalforest area is 11,240 ha and is made up ofSitka Spruce (8,620 ha), Lodge Pole Pine(1,234 ha), Norway Spruce (323 ha) and theremainder is other broad leaved species.

Expected Results• A model capable of estimating forest

parameters at stand level using fieldinventory report measurements.

• A multi-temporal LiDAR datasetwhich estimates change at the standlevel.

• A profile of site characteristics forstorm susceptible forests which willbe used to identify at future riskareas.

• An impact assessment report onnutrient availability due to theremoval of biomass on subsequentforest rotations.

• An impact assessment on the effectsof encroachment of invasive specieshas on timber production andsustained yield.