Artificial Intelligence spotlights the importance of forest communities in afforestation

In the scenic Western Himalayan foothills, an artificial intelligence (AI) application has sniffed out how natural resource management policies are doing on the ground.

The AI application used a machine learning approach and detected that the long-term success of future community-based plantation programs in the Indian Himalayas depends on the provision of alternative grazing options to forest communities and facilitating secure rights over forests.

To arrive at the findings, machine learning (ML) algorithms simultaneously parsed through 200 forest management regions (FMR) in Himachal Pradesh’s Kangra valley.

“Generally, our findings reconfirm research showing that forestry policies that support strong local institutions and tenure are more likely to drive positive outcomes,” said Pushpendra Rana, lead author of the study that documents the use of ML algorithms in natural resources policy and governance.

The computational method integrated with social-ecological-systems theory has helped Rana and co-author Dan Miller at the University of Illinois unravel the efficacy of Joint Forest Management (JFM) and Community Forest Management (CFM) policies in the valley across a 14-year period (2002 to 2016).

Machine learning harnesses modern computing power to explore patterns in large datasets, an advantage over traditional policy impact evaluations.

“Machine learning aided us in detecting where the policies are working and what impacts they have had in the long-term and not just after the project ends. This long term evaluation is important because tree cover growth takes a long time, especially in high altitudes,” according to Rana of the university’s Department of Natural Resources and Environmental Sciences.

“This method also tells us how we can use funds efficiently in forest management. Most importantly, you can tailor afforestation policies in a way that it does no social harm,” Rana emphasised, cautioning that ML should be considered a complement to rather than a replacement for theory-based econometric approaches.

Impact of forest management policies

Community-based forest management relies on the involvement of local communities in resource governance to improve social and ecological outcomes.

Underscoring the importance of local participation in forest improvement, the analysis shows that if grazing lands are snatched away from farming communities for afforestation, forest protection is unlikely. Land stewardship is the key.

There is context to this, as Rana explained, in case of JFM where the state is in control, parcels of land where trees are planted are fenced in, restricting access to grazing grounds for cattle.

JFM (a Forest Development Agency program funded by the national government) was the flagship community participatory initiative for forests in India starting in the 1990s. In the early 2000s, JFM interventions were begun in several of the study FMRs to involve communities in forest regeneration and protection.

This program supported local participation in afforestation programs to enhance tree cover and boost local livelihoods. Communities were organized into JFM committees to raise and protect plantations on public lands.

But unlike CFM initiatives that are led and facilitated by communities, JFM did not allocate forest property rights to communities or the right to derive revenue from forest resources claimed by the state.

Tushar Dash, an Odisha-based scholar and forest rights activist, said CFM is far more effective in forest regeneration and ecological protection.

“One, because they are actually led by user groups (people who are directly depending on forests) and because CFM practices are adaptive as they have evolved over time. They continuously reflect the local needs and adapt accordingly. Also, the communities have an intimate people-nature relationship,” Dash who was not associated with the study, told Mongabay-India.

But if you look at forest department plans for JFMs, they are not adaptive as they are prepared for 10 years, said Dash. “They are mostly oriented towards extraction,” he said.

Decoding forest management one area at a time

Using satellite images from NASA, Rana’s machine learning algorithm detected that neither CFM nor JFM appears to have had large overall impacts on the long-term increase in tree cover.

Explaining the reason for this observation, Rana said there are several factors (overall policies, socio-economic status, market influences, local incentives, soil quality variations) that come into play in revegetation practices and such policies (CFM, JFM) may not have a great impact in changing the long-term vegetation growth.

“The other reason is lack of seriousness in allocating real ownership rights to people which make them unwilling to invest their time and resources on saving trees once funds for maintenance of these areas dry up,” said Rana who has spent a decade in the Indian Forest Service.

Read our story on an assessment of the implementation of the Forest Rights Act of India.

The findings show CFM bore fruit in the medium to long term (5-9 years after intervention) when local existing livelihoods were secured and not threatened (like grazing based), especially when members of the forest cooperatives had greater access to grazing areas.

“These grazing areas provide members of community organisations with a regular source of income through grass auctions and grazing resources to support local livestock-based livelihoods,” the study notes.

Further, CFM also worked in forest parcels with a higher number of marginal people, possibly indicating inclusion of their needs in plantation-making and resource management.

This is because marginalised populations are expected to have higher forest dependence due to their heavy reliance on local forest resources for subsistence needs such as grazing and collection of fuelwood, fodder, and small timber.

For state-led JFM, biophysical factors such as temperature and rainfall mattered most.

For instance, JFM did not perform well in FMRs that experienced low temperatures even when these regions were large. This finding suggests “failure of JFM strategies to trigger vegetation growth in colder mountainous areas where local and migratory grazing is common—often resulting in loss of vegetation.”

JFM activities mainly targeted low hill areas and plains in the study region, which have higher average temperatures. The reason for this prioritisation was to maximise the likelihood of success for desired forest plantation species such as teak (Tectona grandis), bamboo (Dendrocalamus strictus), khair (Acacia catechu) and shisham (Dalbergia spp.) in order to achieve high tree cover.

In India, the Forest Rights Act recognises and secures community rights or rights over community forest resources, in addition to individual rights.

“This provision clearly recognises the right of local communities and gram sabhas (village councils) for protection and management of community forest resources. This provision has been very effectively used in Odisha and Maharashtra by a large number of gram sabhas to strengthen gram sabhas and community forest management systems,” elaborated Tushar Dash.

“The initial years of implementation of this provision has demonstrated that CFM along with secured tenure rights over forest resources are very effective in strengthening conservation,” he said.

CITATION

Rana, P., & Miller, D. C. (2018). Machine learning to analyze the social-ecological impacts of natural resource policy: insights from community forest management in the Indian Himalaya. Environmental Research Letters.

Banner image: An aerial view of Bir, Kangra valley in Himachal Pradesh India. Photo by Fredi Bach from Switzerland/Wikimedia Commons.