Resilience for Just Transitions of Agroecosystems Under Climate Change: Northern Midlands and Mountains, Vietnam

Abstract

The aim of this research is to identify policy and practice interventions that support a just transition towards resilient practices for resource-dependent communities. We focus on Thai Nguyen and Phu Tho, two provinces in the Northern Midlands and Mountains of Vietnam. The region is reliant on agriculture but is assessed as highly vulnerable to climate change. We surveyed 105 farming households. A Likert-type questionnaire asked respondents to self-assess their experiences of weather extremes and of changes they had made to their farming practices. Our results show that for both Thai Nguyen and Phu Tho, farmers see the effects of climate change on their crops. Respondents in Thai Nguyen were more likely to report technically driven adaptation and engagement with extension services. Respondents in Pho Tho were more likely to continue traditional practices. For both, use of traditional knowledge and practices was related to taking measures to adapt to climate change. Our main conclusion is that at least three actions could support a just transition to resilient livelihoods. First is incorporating natural science and traditional knowledge into decision-making for just transitions. Second is considering long-term implications of interventions that appear to support livelihoods in the short term. Third is tailoring messaging and engagement strategies to the requirements of the most vulnerable people. The main message of this study is that a just transition for resource-dependent communities will inevitably be context-specific. Even in centralized and authoritarian contexts, flexibility to adapt top-down policies to locals’ own experiences of changing climates is needed.

1. Introduction

The role of nature and natural resources in supporting a just transition is gathering international policy attention. A just transition can be understood as a response to climate change, biodiversity, and sustainability issues that leaves nobody behind, and in particular, it does not unfairly disadvantage people and places who rely on natural resources for jobs and livelihoods. The International Labour Organisation, UN Environment Programme, and International Union for the Conservation of Nature estimate that 75 million people globally work in nature-based jobs. Most of these are concentrated in Asia and the Pacific and in low-to-middle-income countries, and much of this work is in the agricultural sector [1].

However, it is also increasingly recognized that the biodiversity protection, carbon sequestration, and climate adaptation benefits of nature-based jobs will not automatically translate into fair and decent work and livelihoods for the least well-off. A structured just transition process is required to ensure that nature-based activities such as sustainable agriculture designed to bring global benefits also support the least well-off. This is especially so in low-emitting countries and contexts, where livelihoods in sectors such as agriculture and fisheries depend on continued access to healthy ecosystems [2].

2. Research Context

2.1. Just Transitions and Resource-Dependent Livelihoods

The consideration of just transitions has broadened beyond its earlier focus on energy and labour to encompass nature-based livelihoods and resource-dependent communities (e.g., [2,3,4]). A core concern here is ensuring that people and places who depend on natural resources for livelihoods—especially in the Global South—do not face disadvantage in creating fair and decent work and livelihoods as a result of top-down biodiversity conservation, environmental protection, or carbon sequestration initiatives [5,6].

A just transition framing adds nuance to more traditional ‘sustainable livelihoods’ approaches, which have faced calls for more attention to structural drivers of poverty and to impacts of global climate change, which can exceed or overwhelm the ability of rural households and communities to cope [7,8]. Singh et al. [9] nonetheless argue that livelihoods remain a useful analytical tool for understanding how sustainable and just a transition actually is. The reason for this, Singh et al. argue, is that livelihoods ground analysis of just transitions in people’s everyday engagements with social and environmental systems. To put these two perspectives together, we see livelihoods as a heuristic to make sense of the different ways in which people experience environmental change and the strategies that are proposed in response, and just transitions as a way of turning these experiences into a set of principles that support fairer and more effective decision-making for nature-based livelihoods under environmental change [10]. In the context of nature-based livelihoods, just transitions may involve the development of norms, integration of multiple knowledge systems in evidence bases, fairer decision-making, and upholding rights and fostering empowerment (e.g., [1,11,12]). All of these speak to addressing structural processes that reproduce poverty and inequality. For all of these, a rich evidence base is also required to inform decision-making in the face of the pace and extent of climate and biodiversity change.

Research has been conducted on just transitions for resource-dependent communities, especially agricultural ones, across different geographical contexts (e.g., [3,10,13]). However, Lo [5] calls for further enquiry into what just transitions mean for resource-based economies in authoritarian contexts. In such settings, limited opportunities for participation and bottom-up input in environmental protection policymaking mean a different approach may be required to embed social justice and just transition concerns in environmental policy [5]. Lo proposes three contributions researchers in authoritarian contexts can make to just transitions: examining socio-economic impacts of environmental policies; developing alternative forms of participation for affected communities; and supporting an evidence-based policy process that allows the effectiveness of different interventions to be assessed.

We respond to Lo’s call by examining impacts of climate change, and perceived effectiveness of countermeasures, for two agriculture-dependent provinces in Northern Vietnam. We focus, in particular, on support for resilience as a key component of a just transition for agriculture and as a vehicle that may allow social justice concerns for the least well-off to be embedded into local and regional environmental policy. Resilience in this context can be defined as the ability of socio-ecological systems to be able to fulfil their function in changing conditions, recognizing that a resilient system can and should be able to transform whilst delivering on its main goal [13]. Kohmonen and Siltaohja [13] argue that promoting the resilience of actors such as farmers in the transition process, and enabling them to build capacities, can act as a form of restorative justice.

2.2. Climate Change, Livelihoods, and Resilience in Rural Vietnam

It is also important to acknowledge recent research on climate change, livelihoods and resilience in rural Vietnam. Perhaps unsurprisingly, existing research in a rural Vietnamese context indicates the least well-off are most negatively affected by climate-related extremes [14]. In response, recent work indicates famers in rural parts of Vietnam may be more likely to take adaptive action in response to a changing climate if they have access to knowledge and information about weather and climate [15,16]. Previous adaptation behaviours and experience of what has worked in the past can also influence whether and how farmers adapt in response to a changing climate [15,17]. However, it is also argued that the provision of ‘more and better’ technical information about new practices alone by government divisions or international aid agencies is unlikely to be effective in guiding farmers towards climate adaptive practices [17,18]. This is especially so if top-down engagement efforts emphasize technological innovations to the detriment of traditional and indigenous knowledges plus social and cultural contexts [19]. Much recent work on resilience and climate adaptation in rural Vietnamese contexts has argued that household- or community-level responses alone are inadequate in the face of the pace and scale of climate change, and therefore that government policies across multiple sectors have a fundamental role in supporting climate-resilient livelihoods (e.g., [14,16]). However, it is also argued that social capital and support networks have a vital role in enabling community members to share knowledge and best practices [16,17,18]. Our work builds on this existing research by exploring how strategies to support climate adaptation and resilience reported in existing research on rural agriculture in Vietnam may be tied explicitly into a just transition for resource-dependent communities.

2.3. Study Aims and Outline

The aim of this paper is therefore to use the case of two agricultural provinces in Vietnam with differing levels of wealth and productivity—Thai Nguyen and Phu Tho—to identify policy and practice interventions that may support a just transition towards more resilient practices in a resource-based context. Having outlined critical challenges in just transitions and sustainable livelihoods, we now introduce the two provinces. We then use findings from a survey with farmers across the two provinces to identify actions that can support resilience towards improving livelihoods. Finally, we propose a series of actions that can support a just transition through agriculture-based jobs in contexts like Thai Nguyen and Phu Tho provinces.

3. Vietnam’s Northern Midlands and Mountains: Thai Nguyen and Phu Tho Provinces

The Intergovernmental Panel on Climate Change’s Sixth Assessment Report summarizes for South-East Asia that future warming will be slightly less than the global average. The assessment projects that rainfall is likely to increase in the northern part of the region (including Vietnam), but that observed mean rainfall trends are not spatially coherent or consistent across datasets or seasons. The IPCC also predicts fewer but more extreme tropical cyclones will affect the region. For the East Asia climate zone, within which the Northern Midlands and Mountains lies, the IPCC projects an increase in frequency and intensity of heavy precipitation, leading to more landslides, more frequent droughts, and an increase in intensity and number of tropical cyclones [20].

More specifically, Vietnam’s National Adaptation Plan summarizes past and future climate change impacts on Vietnam. The plan notes that Vietnam’s average temperature has increased by 0.98 °C between 1958 and 2018, with 0.74 °C of this change since 1986 and a further 1.2–1.7 °C of average temperature rise across Vietnam predicted by 2050 under a high-emission scenario. Rainfall has increased by 2.1% nationwide since 1958, with a projected increase of 10–15% by 2050 in the south of the country, but a projected decrease in the north. Storms and tropical depressions have increased, are following unusual paths in some cases, and are projected to become more frequent and intense. Finally, the duration of the dry season has been increasing in the north of Vietnam, and central and norther delta regions are projected to be at even higher risk of drought, scarcity, and desertification under the projected climate impacts Vietnam is facing [21].

3.1. Northern Midlands and Mountains

Vietnam’s Northern Midlands and Mountains region includes 14 provinces (Figure 1). The region has borders with Guangxi and Yunnan in China and the Lao People’s Democratic Republic. To the east and south are the Red River Delta and North Central and Central Coast regions, respectively. The total area of Northern Midlands and Mountains is 95,000 km² (accounting for 28% of Vietnam’s area), with a total population of 14,688,190 people (accounting for 15.2% of Vietnam’s population) [22]. The region has a diverse population with more than 30 ethnic groups living together, of which the Kinh people account for more than 40% and the remaining ethnic groups account for nearly 60%.

The Northern Midlands and Mountains is a region where commercial agriculture is vital to regional economic sustainability and also livelihoods and jobs. The region has special strengths for developing industrial crops of subtropical and temperate origin, and is the largest tea region in the country, accounting for 74.3% of the Vietnam’s tea growing area in 2018. The crop structure has also shifted towards forming concentrated, large-scale commodity production areas. From 2017 to 2020, the region converted about 54,000 ha of ineffective rice-growing land to aquaculture or to annual and perennial crops with higher economic value [23].

The proportion of labour in agriculture, forestry, and fishery in the Northern Midlands and Mountains accounts for over 60% (compared to 26.9% for Vietnam as a whole), including about 35% labour belonging to ethnic groups. Due to a focus on training for agriculture, the proportion of trained workers in the whole region has nearly doubled, from 13.9% in 2011 to 25.9% in 2021, accounting for 30% of the number of rural workers with jobs after training in the whole country [24]. For the Northern Midlands and Mountains as a whole, the development of a skilled agricultural workforce is hence a key part of supporting livelihoods in the region, alongside drives for the commercialisation and profitability of agriculture. Such opportunities for upskilling and training are considered a core part of a just transition, especially for nature-based jobs [1]. It is therefore vital that improvements made to date and outstanding challenges can be sustained under a changing climate. In practice, this means (a) understanding what the impacts of a changing climate are and how they are affecting people in the region and (b) developing interventions and transitions that ensure the continued sustainability of livelihoods under a changing climate.

The climate of the Northern Midlands and Mountains is characterized by tropical humid monsoons and cold winters. However, flash floods, landslides, and severe cold spells have intensified in recent years, with implications for crops and livestock and therefore livelihoods. Over 2010–2020, the total damage caused by climate change and associated disasters in the Northern Midlands and Mountains region amounted to more than VND 16 trillion [25]. Vietnam’s National Adaptation Plan highlights the higher risk of flash floods and landslides in the region. The plan emphasizes that although exposure is not high in the Northern Midlands and Mountains compared to the rest of Vietnam, the region’s high poverty rate makes it vulnerable to climate impacts [21]. Cuesta et al. [26] assess the Northern Midlands and Mountains region as having the highest exposure to climatic risks and weakest social sustainability (which they define as a combination of inclusion, resilience, and participation) in Vietnam. Cuesta et al. also note that there are wide variations between the provinces and districts of the region [26]. Vo and Tran [27] likewise assess the Northern Midlands and Mountains as the most climate-vulnerable region in Vietnam from the perspective of socio-demographic profile, due to low levels of adult education and high numbers of dependents per household. Vo and Tran also argue that the region is highly vulnerable from a livelihood perspective, due to the reliance on agricultural activities as the main source of income [27].

3.2. Thai Nguyen and Phu Tho Provinces

Within the Northern Midlands and Mountains, we focus on Thai Nguyen and Phu Tho provinces. Note that as this research was conducted prior to the first step of the implementation of Vietnam’s Plan to Arrange and Merge Administrative Units in March 2025; the data in this paper relates to the administrative boundaries of Thai Nguyen and Phu Tho provinces as of December 2024. Nonetheless, the overarching points about climate impacts, income levels, and the agricultural economy still hold. Thai Nguyen and Phu Tho are two provinces strongly affected by climate change with manifestations such as increased temperature, rainfall, and extreme weather events.

Table 1. Demographic and socio-economic characteristics of Thai Nguyen and Phu Tho provinces, in relation to the Northern Midlands and Mountains and Vietnam.

	Thai Nguyen	Phu Tho	Northern Midlands and Mountains Average	Vietnam Average
Life expectancy at birth	74.5	74.4	72.6	74.5
Literacy rate (15+)	98.8	98.6	91.9	96.6
Percentage trained labour force	36.6	34.1	26.8	27.2
Labour productivity (2022) (Mill. dongs/employed)	243.9	128.0	136.4	188.7
Informal employment rate	54.8	63.7	71.5	65.1
Land use: total area (thousand hectare)	352.2	353.5	9518.4	33,134.5
Land use: agricultural land (thousand hectare)	110.4	117.7	2281.2	11,673.4
%age agricultural land (2022)	31.3	33.3	24.0	35.2
Planted area of rice (thousand hectare)	67.5	58.4	653.1	7119.3
Yield of rice (Quintal/ha)	55.7	59.2	52.3	61.1
Production of paddy (thousand tonnes)	376.3	346.1	3414.4	43,497.7
Production of tea (thousand tonnes)	264.4	178.1	N/A	1125.1
Number of acting enterprises per 1000 inhabitants	4.3	4.3	3.2	9.2
Monthly average income per capita at current prices (thousand dong)	4908	4330	3438	4962
Monthly average income per capita at current prices for lowest quintile (thousand dong)	1437	1361	869	1451
Average income per capita at current prices for agriculture, forestry and fishery (thousand dong)	607	282	575	502
Human Development Index (2022)	0.770	0.724	N/A	0.737
Cost of living index (where Hanoi = 100)	92.37	88.74	N/A	N/A
Poverty rate (%)	1.7	3.7	10.7	3.4

(Source: National Statistics Office of Vietnam [28]. All data for 2023 unless otherwise stated.)

outlines some of the demographic and socio-economic characteristics of the two provinces. In general, both provinces have a life expectancy, literacy rate, and proportion of trained labour force that are above the average for the Northern Midlands and Mountains and for Vietnam as a whole. Both Thai Nguyen and Phu Tho provinces also have a proportion of agricultural land and rice yields in line with the Vietnam average and together produce a notable proportion of all tea in Vietnam.

However, although average incomes in Thai Nguyen are in line with the Vietnam average, those in Phu Tho are notably below the average. Similarly, in Thai Nguyen, average incomes for agricultural workers, the Human Development Index, and the poverty rate are all better than the average for Vietnam. Conversely, for Phu Tho, the average incomes for agricultural workers, Human Development Index, and poverty rate are all well below the average for Vietnam.

Comparing Thai Nguyen and Phu Tho provinces allows us to compare two provinces that are geographically near each other but have different socio-economic characteristics. Thai Nguyen is a relatively well-off province, whereas Phu Tho appears much less wealthy (albeit still better off than some others in the Northern Midlands and Mountains). This in turn allows us to explore what resilience strategies for a just transition might be required across different socio-economic contexts in a region that has been assessed as being one of the most climate-vulnerable and with some of the biggest adaptive capacity challenges within Vietnam [26,27].

4. Methodology

A questionnaire survey was conducted on households and household farms, representatives of cooperatives, cooperative groups, and agricultural enterprises in Thai Nguyen and Phu Tho provinces. A total of 45 households were selected in Thai Nguyen and 60 in Phu Tho. The sample covered households across rice, tea, and fruit tree agroecosystems. The number of specific ecosystems sampled in each province is listed in

Table 2. Agroecosystems sampled in each province.

Agroecosystems	Thai Nguyen	Phu Tho	Total
Rice	33	42	75
Tea	39	42	81
Fruit trees	8	10	18

. Cluster random sampling was used, from an overall population of districts with tea, rice, and fruit production in Thai Nguyen and Phu Tho provinces. For each province, two districts were selected, and for each district, one commune was selected. A random survey of agricultural households in the commune was then conducted based on a list of agricultural households provided by the local government.

An adapted version of the climate-resilient agriculture framework developed by Rao et al. [29] was used to collate physical environmental data for each location, as well as respondents’ own assessment of the impacts of climate change and their experiences with countermeasures. The value of using this approach is that it provides a framework that allows priority areas for action to be identified and interventions to be proposed that can enhance the resilience of agroecosystems for specific contexts [29]. Given our intention to identify actions that can be taken to support resilience towards a just transition, this framework thus provides a useful heuristic. In places, additional questions were added to reflect the local context in Thai Nguyen and Phu Tho provinces (for instance, assessing impacts of climate change on tea versus rice, rather than crops in general, given the importance of tea to the provinces).

Table 3. Survey questions, based on the framework of Rao et al. [29].

Aspect	Original Indicators Selected for Inclusion	How Adapted for Survey	How Measured in Survey
Ecological	Crop biodiversity	Surrounding biodiversity	Likert self-assessment of surrounding biodiversity
	Livestock biodiversity
	Flood frequency	Floods	Likert self-assessment of impact on rice and tea crops
		Flash floods	Likert self-assessment of impact on rice and tea crops
		Landslides	Likert self-assessment of impact on rice and tea crops
	Water productivity	Water quality	Likert self-assessment of water quality
	Drought frequency	Drought	Likert self-assessment of impact on rice and tea crops
	Frost frequency	Cold	Likert self-assessment of impact on rice and tea crops
	Fertilizer usage	Use of organic fertilizers	Kg/ha for rice and tea
		Use of NPK synthetic chemical fertilizers	Kg/ha for rice and tea
		Nitrogen fertilizers	Kg/ha for rice and tea
		Phosphate fertilizers	Kg/ha for rice and tea
		Potassium fertilizers	Kg/ha for rice and tea
	Forest cover	Forest cover ratio	%age forest cover
	Pesticide use	Use of pesticides	Kg/ha for rice and tea
		Biological pesticides	Kg/ha for rice and tea
		Growth stimulants	Kg/ha for rice and tea
		Herbicides	Kg/ha for rice and tea
	Soil organic carbon	Soil quality	Likert self-assessment of soil quality
	Soil drainage
	Soil depth
Economic	Per capita food supply	Crops per household	Number of different crops grown
		Tea tree yield stability	Likert self-assessment of stability
		Rice yield stability	Likert self-assessment of stability
	GVA from crops GVA from livestock	Income generated from agricultural production	Self-reported percentage generated from agricultural production
		Level of access to tea market	Likert-self assessment of accessibility to tea market
		Level of access to rice market	Likert-self assessment of accessibility to rice market
	Non-farm income Poverty rate	Access to credit	Likert-self assessment of access to credit
		Credit loans	Likert-self assessment of access to loans
	Marketable surplus	Number of commercial crops	Number of different crops grown
	Agricultural employment	Acreage of agricultural land/household	Average farm size per household (ha)
		Acreage of agricultural land/labour	Average farm size per household (ha) divided by number of labourers
Social	Adoption of improved practices	Applying traditional knowledge and experience to production	Likert self-assessment of how often action is performed
		Regularly improve the soil	Likert self-assessment of how often action is performed
		Practice/measures to mitigate and adapt to climate change	Likert self-assessment of how often action is performed
		Develop annual production plans	Likert self-assessment of how often action is performed
		Adjust crop calendars in a timely manner	Likert self-assessment of how often action is performed
		Manage pests and diseases	Likert self-assessment of how often action is performed
		Access weather information	Likert self-assessment of how often action is performed
		Irrigate water sparingly	Likert self-assessment of how often action is performed
		Switch to drought-resistant crop varieties	Likert self-assessment of how often action is performed
		Cover cultivated areas to retain soil moisture	Likert self-assessment of how often action is performed
		Use ecological measures to control pests and diseases	Likert self-assessment of how often action is performed
Institutional	Access to extension services	Access to extension services	Likert self-assessment of how often action is performed
		Participating in training	Likert self-assessment of how often action is performed
	Disaster preparedness	Preparing for disasters Support during disasters	Likert self-assessment of how often action is performed
	Agriculture insurance

outlines how the framework was adapted into survey questions for the purposes of the study.

Table 4. Characteristics of farming households in the study area.

	Provinces	Demographic Characteristics	Number of Information Providers (People)	Percent of Information Providers (%)
Gender	Thai Nguyen	Male	28	62.2
		Female	17	37.8
		Total	45	100
	Phu Tho	Male	36	60
		Female	24	40
		Total	60	100
Education level	Thai Nguyen	Elementary school	7	15.6
		Middle school	13	28.9
		High school	14	31.1
		Intermediate school	7	15.6
		College or higher	4	8.9
		Total	45	100.0
	Phu Tho	Elementary school	6	10.0
		Middle school	20	33.3
		High school	28	46.7
		Intermediate school	3	5.0
		College or higher	3	5.0
		Total	60	100.0
Age	Thai Nguyen	Below 30	1	2.2
		30–39	6	13.3
		40–49	11	24.4
		50–59	18	40
		Over 60	9	20
		Total	45	100.0
	Phu Tho	30–39	5	8.3
		40–49	19	31.7
		50–59	21	45
		Over 60	15	25
		Total	60	100.0
Household size (people)	Thai Nguyen	2–3	12	26.7
		4–5	27	60.0
		Over 6	6	13.3
		Total	45	100.0
	Phu Tho	1	1	1.7
		2–3	15	25
		4–5	34	56.7
		Over 6	10	16.8
		Total	60	100.0
Agricultural workers/household	Thai Nguyen	1	7	15.6
		2–3	36	80.0
		Over 4	2	4.4
		Total	45	100.0
	Phu Tho	1	12	20.0
		2–3	45	75.0
		Over 4	3	5.0
		Total	60	100.0

summarizes basic characteristics of the survey respondents.

The values of the indicators reported in the Results were calculated based on simple averages across all respondents and the One-Way ANOVA method to compare the average values between Thai Nguyen and Phu Tho, without applying weights to control for unbalanced structure by crop type or region. The reason for this is that only agricultural households in tea and rice-producing communes were determined at the initial sampling stage.

5. Results

5.1. Ecological Indicators

The first points to establish are ecological factors relating to resilience and adaptation, namely, how the climate and weather has changed in the two provinces under study; whether farmers and residents believe they are experiencing these changes; and what changes are being made to the farming environment in response.

Firstly, according to hydro-meteorological data [30,31], air temperature in Thai Nguyen and Phu Tho has increased by 0.5–1.10 °C. Average annual rainfall increased in the period 1961–2018 by 10–20%, while rainfall in autumn and winter decreased gradually. In Thai Nguyen, average annual temperature increased by 0.90–1.20 °C. Average annual rainfall at all stations tended to decrease. However, the rate of annual rainfall decrease at the stations was very different. For example, at Thai Nguyen station, annual rainfall decreased by 22.3%, while Dinh Hoa station decreased by 4.2%. In Phu Tho, average annual temperature increased by 0.50 °C. Average annual rainfall increased slightly (about 1.2%); however, heavy rainfall was concentrated in the second half of August and September. For example, in 2023, rainfall in the second half of the month was generally 10–20% higher than the average of many years in the same period. From June to August 2020, heat waves occurred with higher intensity. Climate change increased extreme weather phenomena such as prolonged heat, drought, severe cold, flash floods, and landslides. In 2020, Phu Tho experienced 20 events classified as disasters, mainly heavy rains, thunderstorms, hail, floods, etc., increasing in frequency and severity, making prevention very difficult, and causing great damage to agricultural production. As such, there is indication that climate and weather in both provinces is changing, in a way that may impact agricultural activity.

Secondly, the majority of households (73.4%) report being affected by drought. Conversely, only 13% of surveyed households report being affected by landslides, and 21% by flash floods (

Table 5. Impact of climate change on agricultural production of farming households.

Type of Extreme Event	Number of Households Self-Reporting as Affected	Number Participating in Interview (People)	Percent of Participating in Interview (%)
Drought	Yes	78	74.3
	No	27	25.7
Landslides	Yes	13	12.4
	No	92	87.6
Severe cold, harmful cold	Yes	46	43.8
	No	59	56.2
Pipe floods, flash floods	Yes	21	20
	No	84	80

). Respondents also reported that whilst they were affected by changes in climate and weather, these changes did not necessarily translate into perceived impacts on the agroecosystems within which they worked. Perceived impacts were largely reported as ‘low’ or ‘very low’ (

Table 6. Impact of climate change on farming households in the study area.

Type of Disaster	Thai Nguyen Mean	Phu Tho Mean	Mean	Sig
Drought (Overall)	0.62	0.83	0.74	0.014 *
Rice	1.32	1.83	1.62	0.179
Tea tree	1.91	2.27	2.11	0.352
Cold, severe cold (Overall)	0.67	0.27	0.44	0.000 *
Rice	1.13	0.67	0.87	0.107
Tea tree	1.53	0.37	0.87	0.000 *
Flash floods (Overall)	0.24	0.17	0.20	0.329
Rice	0.62	0.57	0.59	0.844
Tea tree	0.29	0.75	0.55	0.108
Floods (Overall)	0.36	0.13	0.23	0.007 *
Rice	0.47	0.27	0.35	0.261
Tea tree	0.51	0.10	0.28	0.013 *
Landslides (Overall)	0.07	0.17	0.12	0.126
Rice	0.16	0.57	0.39	0.110
Tea tree	0.11	0.67	0.43	0.030 *

* Statistical significance with p < 0.05.

). However, reported impact levels are different between rice and tea tree, with the average value among households affected by drought being 1.62 for rice and 2.11 for tea (Table 6). Meanwhile, reported impact levels for flash floods, floods, and landslides are all less than 0.5. This result shows that these types of disasters are perceived as having a relatively small impact on the agroecosystem in the study area.

Thirdly, in response to changing climate and weather patterns,

Table 7. Environmental dimensions of ecosystems.

Sub-Themes	Thai Nguyen		Phu Tho		Rice		Tea Tree
	Rice	Tea Tree	Rice	Tea Tree	Mean	Sig	Mean	Sig
Use of organic fertilizers	4.67	63.0	0.83	0.83	2.48	0.1	27.48	0.01 *
Use of NPK synthetic chemical fertilizers	12.73	71.33	18.78	70.07	16.19	0.2	70.63	0.9
Nitrogen fertilizers	1.78	11.02	1.42	9.72	1.57	0.6	10.28	0.8
Phosphate fertilizers	7.16	33.82	1.22	1.33	3.76	0.00 *	15.26	0.02 *
Potassium fertilizers	1.76	6.45	1.05	10.92	1.35	0.4	9.03	0.6
Manure	14.67	169.33	55.67	669.15	38.10	0.3	452.88	0.1
Use of pesticides	1.42	2.6	1.42	2.18	2.35	0.9	2.35	0.26
Biological pesticides	0.13	0.33	0.37	0.37	0.27	0.1	0.35	0.8
Growth stimulants	0.9	0.49	0.2	0.75	0.5	0.08	0.64	0.3
Herbicides	0.33	0.16	0.5	0.31	0.43	0.2	0.24	0.2
Soil quality	3.4	3.4	3	3	3.2	0.00 *	3.2	0.00 *
Water quality	3	3	3	3	3	0.00 *	3	0.00 *
Forest cover ratio	52.88		50.25	50.25	51.38	0.002 *
Surrounding biodiversity	3		2.53	2.53	0.00	0.00 *

* Statistical significance with p < 0.05.

shows that chemical fertilizers, pesticides, and growth stimulants were commonly used in agroecosystems, and that there are differences between the strategies adopted for rice and tea plants. For instance, use of synthetic fertilizers is higher in Phu Tho, whereas use of organic fertilizers is higher in Thai Nguyen, especially for tea plants. Moreover, pesticides and herbicides were still widely used for rice and tea tree. These results indicate higher use of organic fertilizers in Thai Nguyen compared to Phu Tho but also a continued reliance on chemical fertilizers, pesticides, and growth stimulants across both provinces and multiple agroecosystem types.

In summary, what all of the above suggests is that climate and weather patterns are changing, and that respondents in the two provinces experience these changes in different ways. However, the perceived impacts on agroecosystems in both provinces are, for now, low. Chemical and synthetic treatment of crops remains a common strategy for respondents in both provinces, with a slightly higher interest in organic and traditional approaches in Thai Nguyen compared to Phu Tho.

5.2. Economic Indicators

The second point to establish is the economic status of farmers in the two provinces. On the whole, across both provinces, the reported stability of tea tree yield is higher than that of rice yield (

Table 8. Economic aspects of agroecosystems at the study site.

Themes	Sub-Themes	Thai Nguyen Mean	Phu Tho Mean	Mean	Sig
Economic	Tea tree yield stability	2	2	2	0.91
	Rice yield stability	1.02	0.5	0.723	0.00 *
	Income generated from agricultural production	87.29	78.90	82.49	0.15
	Access to credit	0.33	1.83	1.19	0.00 *
	Credit loans	1.71	1.77	1.74	0.52
	Level of access to tea market	3.91	3.07	3.43	0.05 *
	Level of access to rice market	0.78	1.22	1.03	0.09
	Acreage of agricultural land/household	5847.11	13,686.1	10,326.53	0.00 *
	Acreage of agricultural land/labour	3010.74	7238.08	5426.36	0.00 *
	Number of crops/household	5.91	5.68	5.78	0.41
	Number of commercial crops	1.688	1.583	1.6286	0.55

* Statistical significance with p < 0.05.

). However, the reported yield stability of these two crops only fluctuates between 1 and 2 (corresponding to very low level to low level). Table 8 also shows that agricultural production is the reported main source of income for 87.29% of surveyed households in Thai Nguyen and 78.90% of surveyed households in Phu Tho. In other words, many households in the surveyed regions are very vulnerable to fluctuations in revenue from agriculture, and hence could be negatively affected if climate change impacts agricultural yields. Respondents across both provinces report relatively low access to credit, indicating a low investment in agricultural production, but with access to credit being higher in Phu Tho. Respondents in both provinces reported perceived good access to tea markets (but significantly higher in Thai Nguyen), whereas respondents in both reported low access to rice markets (but significantly higher in Phu Tho). Average farm sizes were also significantly higher in Phu Tho compared to Thai Nguyen, albeit with respondents in Thai Nguyen reporting more income from agricultural production and higher access to tea markets.

This points to a situation where despite not yet reporting impacts on agroecosystems, many surveyed households in the two provinces are highly reliant on agricultural production and hence potentially vulnerable if future climate and weather changes negatively impact farming. Access to credit (and hence investment) is higher in Phu Tho, albeit with less variation in crops and less access to tea markets than Thai Nguyen. Farms in Thai Nguyen are smaller, but more intensive and profitable with higher access to tea markets.

5.3. Social Indicators

The third area to explore is the social environment within which farming in the two surveyed provinces happens.

Table 9. Social dimensions of agroecosystems in the study area.

Themes	Sub-Themes	Mean		Mean	Sig
		Thai Nguyen	Phu Tho
Social	Applying traditional knowledge and experience to production	2.71	3.87	3.37	0.00 *
	Regularly improve the soil	2.13	2.97	2.61	0.010 *
	Practice/measures to mitigate and adapt to climate change	1.04	1.80	1.48	0.051
	Develop annual production plans	0.76	0.48	0.60	0.351
	Adjust crop calendars in a timely manner	0.84	0.37	0.57	0.097
	Manage pests and diseases	1.02	0.63	0.8	0.227
	Access weather information	1.13	0.67	0.87	0.166
	Irrigate water sparingly	1.31	0.84	1.05	0.138
	Switch to drought-resistant crop varieties	2.09	1.42	1.71	0.088
	Cover cultivated areas to retain soil moisture	1.11	1.02	1.06	0.771
	Use ecological measures to control pests and diseases	1.50	0.47	0.91	0.00 *

* Statistical significance with p < 0.05.

shows that the average level of application of traditional experience and knowledge in production across the two provinces is 3.37 (which is to above average and close to regular application), with the use of traditional experience and knowledge being significantly higher in Phu Tho. Although self-reported soil improvement practices only reach 2.13–2.79 (corresponding to low application), this is again significantly higher in Phu Tho. Other measures surveyed to mitigate and adapt to climate change, such as developing annual production plans, adjustment of crop schedules, management of pests and diseases, accessing weather information, saving water, and using ecological measures, are all at very low or low application levels in production activities. In summary, these results show significant differences between Thai Nguyen and Phu Tho in the application of traditional knowledge and experience plus soil improvement (which are higher in Phu Tho) and in the use of ecological measures to improve soil (which are higher in Thai Nguyen). More technical measures to adapt to changes in weather and climate, such as developing production plans and accessing weather information, are reported more frequently in Thai Nguyen. The adoption of other behavioural or practical measures to maintain climate resilience otherwise remain low across respondents in both provinces.

5.4. Institutional Indicators

The fourth and final area to explore is the institutional factors that support climate-resilient agricultural practices. In the studied provinces, authorities at all levels (province, district, commune) have developed plans to respond to climate change.

Table 10. Institutional dimensions of agroecosystems in the study area.

Themes	Sub-Themes	Thai Nguyen Mean	Phu Tho Mean	Mean	Sig
Institutional	Access to extension services	0.53	0.30	0.40	0.01 *
	Preparing for disasters	0.40	0.43	0.42	0.74
	Supporting disasters	0.80	0.93	0.88	0.04 *
	Participating in training	0.64	0.70	0.68	0.55

* Statistical significance with p < 0.05.

shows that the indicator for using agricultural extension services ranges from 0.3 to 0.53, indicating that the number of households using agricultural extension services provided by the locality is low (accounting for about 40% of surveyed households) but higher in Thai Nguyen. Anecdotal information collected from respondents indicates agricultural extension services chosen by households are focused on fertilizers and seeds.

Moreover, the self-assessed disaster preparedness indicator reported is low for both provinces, at 0.40 (Thai Nguyen) and 0.43 (Phu Tho). Anecdotal information collected from the study area indicates that few if any households buy agricultural insurance. The number of surveyed households reported as participating in training sessions is very limited (Thai Nguyen 0.64; Phu Tho, 0.70), with significantly higher (but still low) perception of support for disasters in Phu Tho. These findings point to engagement with extension services, and self-assessed preparedness and support for response to weather-related extremes, as being low across both surveyed provinces.

5.5. Relationships Between Variables

Given our interest in policy and practice interventions that can support a transition to more resilient livelihoods for resource-dependent communities, we assessed correlations between a sub-set of questions. We took three variables in our dataset that represent the existing productivity of livelihood (agricultural area per household), potential to access resources for adaptation (access to credit), and understanding of traditional and natural processes (application of traditional knowledge and experience) and assessed these against different resilience-strengthening actions in our data.

Access to credit was assessed to have no relationship with practices/measures to adapt to climate change; access to agricultural extension services; disaster preparedness; disaster recovery support; or participation in training. No relationship was present among Thai Nguyen, Phu Tho, or the overall dataset.

For Thai Nguyen, there was a significant positive relationship between agricultural area per household and disaster preparedness (χ² = 45.00; p ≤ 0.05), albeit this relationship was not repeated for Phu Tho or for the overall dataset. Size of agricultural area per household was assessed to have no relationship with practices/measures to adapt to climate change; access to agricultural extension services; disaster recovery support; or participation in training for Thai Nguyen, Phu Tho, or the overall dataset.

The application of traditional knowledge and experience was assessed to have a significant positive relationship with practices/measures to adapt to climate change for Thai Nguyen (χ² = 19.07; p ≤ 0.01), Phu Tho (χ² = 13.39; p ≤ 0.05), and the overall dataset (χ² = 18.15; p ≤ 0.01). The application of traditional knowledge and experience was also assessed to have a significant positive relationship with preparing for disasters for Thai Nguyen (χ² = 15.37; p ≤ 0.01) and for the whole dataset (χ² = 20.19; p ≤ 0.01), as well as with participating in training for Thai Nguyen (χ² = 19.03; p ≤ 0.01) and for the whole dataset (χ² = 19.64; p ≤ 0.01). Neither of these relationships were significant for Phu Tho. The application of traditional knowledge and experience was assessed to have no relationship with access to agricultural extension services or to disaster recovery support among Thai Nguyen, Phu Tho, or the overall dataset.

5.6. Synthesis

Before proceeding to the discussion of the results, it is worth summarizing the comparative strengths and challenges of each province, when it comes to enabling a just transition to resilience for resource-dependent workers. This is intended to highlight the relative preparedness in each province across different aspects and is not intended to suggest one province is more or less prepared than the other.

From our data, Thai Nguyen appears to be adopting more technical approaches that may sustain and enhance livelihoods under a changing climate, and engaging with extension services to a greater degree. However, there is a danger that this intensification may lead to over-reliance on tea crops, which in turn could put livelihoods at risk if the effects of flash floods and severe cold on tea crops reported by our respondents intensify. Indeed, increases in instances of flash flooding and severe cold are raised as a risk for the Northern Midlands and Mountains in Vietnam’s National Adaptation Plan [21].

Conversely, Phu Tho appears to have more farmers who continue to apply traditional knowledge and persist with traditional techniques in response to a changing climate. We observed a positive relationship between the application of traditional knowledge and taking measures to adapt to climate change in Phu Tho, as we saw in Thai Nguyen too. However, the higher reported access to credit in Phu Tho is not statistically linked to changes in farming practice, and reported access to extension services is lower. This indicates that policy measures aimed at promoting resilience and supporting a just transition may not be reaching or benefitting some of the less well-off farmers in Phu Tho.

6. Discussion

We structure the discussion of our results around the three contributions to just transitions in authoritarian contexts identified by Lo [5]. For each, we reflect on how our results relate to both existing research findings from Vietnam plus the literature on just transitions and livelihoods in a resource-dependent context. We also use the three contributions of Lo [5] to reflect on our overall research aim, which is to explore the policy and practice interventions that may support a just transition towards more resilient practices in a resource-based context. Across the three areas below, the main message of our study is that a just transition for resource-dependent communities will inevitably be context-specific. What ‘works’ in Thai Nguyen will not necessary work in Phu Tho, and vice versa. Even in centralized and authoritarian contexts, flexibility to adapt top-down policies to locals’ own experiences of changing climates is needed.

6.1. Supporting Evidence-Based Policy Processes

One point raised by Lo [5] is that to enable just transitions in authoritarian contexts, researchers can support evidence-based policy processes by generating policy knowledge that facilitates transitions towards sustainability. In particular, this may involve assessing the advantages and disadvantages of different policy options [5]. Our findings support this, and, in particular, highlight the importance of different knowledge systems feeding into policy decisions. This includes natural and physical science knowledge as well as local and traditional knowledge.

Section 5.1 showed that respondents experienced the impacts of climate change on their crops in more complicated ways than regional-scale weather and climate data alone might indicate. Those in Phu Tho reported being affected more by drought and landslides, and those Thai Nguyen reported being affected more by flooding. Despite observed regional-level changes in climate and weather patterns, respondents did not self-assess the impact of weather events on their crops as being particularly high, although they did acknowledge they were seeing impacts. This indicates that climate impacts can vary over very short geographical distances (i.e., even within or between provinces), and that the extent to which those working the land experience impacts may differ from what might be expected from looking at regional-scale data alone. Conversely, Section 5.5 showed a significant positive correlation—especially in Thai Nguyen but also across the dataset as a whole—between applying traditional knowledge and experience and taking measures to adapt to climate change or participate in training.

In short, our findings show a need for finer-scale understandings of climate change impacts that can help to identify priority areas for intervention and propose appropriate actions to support livelihoods under a changing climate. However, our findings also indicate that it may be farmers’ traditional knowledge and experience that drives them to adapt their practices or engage with training (see also analogous research in Vietnam, e.g., [15,19]). More broadly, reflecting the call of Lo [5] for a fuller evidence base to support decision-making for just transitions, our findings show how natural and physical science evidence too can support policy for a just transition, by guiding interventions and actions through a finer-scale and localized understanding of what climate impacts will look like ‘on the ground’. However, the existing just transition literature also calls for recognition justice for local and indigenous knowledge (e.g., [4,11]), arguing that local and indigenous knowledge can be an important part of equitable and effective adaptation. Because our results show that use of traditional knowledge appears linked to farmers adapting their practices, the evidence base for decision-making ought to integrate physical science data with farmers’ and residents’ own experiences of climate change.

6.2. Examining Policy Impacts

A second and related contribution from Lo [5] towards just transitions in authoritarian contexts is the importance of researchers continuing to champion the necessity of a just transition process. In practice, this means researchers examining the socio-economic impacts of environmental policies, especially on vulnerable communities [5]. Our findings provide some insight on which policies and strategies farmers are engaging with that can strengthen resilience and a just transition for livelihoods, and what the consequences of this might be.

Our first insight concerns the kinds of crops and farming strategies that may be promoted. Section 5.3 showed that respondents in both provinces reported better access to tea markets than rice markets, with this access being significantly higher in Thai Nguyen. Yet, Section 5.1 showed a significantly higher use of organic and phosphate fertilizers in Thai Nguyen compared to Phu Tho, and also a greater reported impact on tea crops from flooding and cold spells in Thai Nguyen. Bearing in mind that overall agricultural land is lower but tea production much higher in Thai Nguyen (Table 1), this indicates a more intensified model of production in Thai Nguyen than Phu Tho, with greater use in Thai Nguyen of technical approaches to boost productivity (Section 5.3). Given the vulnerability of tea crops to flooding and cold spells reported by respondents in Thai Nguyen (Section 5.1), these findings suggest policies and interventions that promote the intensification of production and greater market access for a single crop might improve incomes in the short-term for less wealthy places such as Thai Nguyen and Phu Tho, but may not necessarily lead to a just transition in the longer term. The reason for this is that intensifying the production of a single crop, such as tea in Thai Nguyen, could enhance the vulnerability that farmers face when extremes do come (e.g., reliance on a single crop), which enhance resilience locally but could have negative ecological impacts elsewhere (e.g., use of pesticides and fertilizers). This finding is in line with that of Halbherr et al. [17] in a Vietnamese context, who advocate for considering broader issues such as market access and the ability to switch crops when making assessments as to the resilience of rural farming communities. It also reflects that of Ha et al. [32], who in analogous research in Thai Nguyen caution against a focus on hard solutions grounded in infrastructure projects and in top-down planning and management policies for agriculture. Our results also support Lima [10], who in a cross-country comparative context advocates a just bio-economy transition that moves away from monocultures and towards more diverse strategies.

Our second insight relates to policies and strategies that may encourage farmers to adapt existing practices or engage in training and upskilling. Section 5.5 indicated that access to credit does not appear related to respondents taking measures to adapt to climate change or engaging with training. We also found that farm size was only significantly related to disaster preparedness in Thai Nguyen, and not in Phu Tho or across the wider dataset. Access to extension services was also significantly higher in Thai Nguyen, an already comparatively wealthier province (Section 5.4), alongside greater use of technical approaches to adaptation (Section 5.3). As discussed in greater depth in Section 5.5 and Section 6.1, the application of traditional knowledge and practices appears to be strongly linked with changing practices or engaging in training initiatives. If we consider that the application of traditional knowledge could relate to heightened perception of changes in climate and weather, then our findings arguably support those of Tuan et al. [18] in a Vietnam-wide survey, who found that heightened climate change perception among respondents led to income diversification and raising financial capacity to adapt. These findings indicate that policies and interventions that aim to enhance access to credit or build farmers’ capacity may not be effective in reaching the most vulnerable farmers. Conversely, we saw (Section 5.5) that traditional knowledge and practices did appear to have a relationship with taking adaptation measures across both Thai Nguyen and Phu Tho regardless of income. Yet, as we now discuss, this is not a reason to over-celebrate traditional practices, which need to be updated and may struggle in the face of the pace and extent of climate change.

6.3. Alternative Forms of Participation

The third contribution from Lo [5] to just transitions in authoritarian contexts is in designing alternative forms of participation. Writing in the Chinese context, Lo argues that even in authoritarian settings, the participatory and collaborative approach required for a successful just transition must involve both the central government, and the stakeholders with local knowledge on how to develop regionally appropriate visions and approaches.

Our findings focus on how institutional factors—especially access to extension services, support during disasters, and participating in training—could be made more effective as forms of participation that may support farmers to transition to resilient livelihoods. Section 5.4 shows that self-reported access to extension services, preparation for disasters and participation in training is low in both Thai Nguyen and Phu Tho. Nonetheless, access to extension services is significantly higher in Thai Nguyen (Section 5.3). Although not statistically significant, respondents in Thai Nguyen also reported making more attempts to engage with new practices that would enable them to respond to climate-related extremes and slower-onset changes (Section 5.3). The correlations in Section 5.5 show that statistically significant relationships between farm size and preparing for disasters, and between the application of traditional knowledge and participation in training, exist in Thai Nguyen but not in Phu Tho. These findings reinforce what has been observed in other research in Vietnam, which sees a limited effect of access to extension services on farmers’ adaptation decisions [18]. This also reflects existing research findings that argue that the top-down provision of information and data alone may not lead to changes in practice if it does not take into account social and cultural factors, or farmers’ own lived experiences (e.g., [19,32]).

A critical challenge is therefore designing forms of participation that can reach all types of farmers, especially those who are less well-off (and potentially more vulnerable) like those in Phu Tho. This is where the engagement of stakeholders with local knowledge on how to develop regionally appropriate visions and approaches [5], such as local government staff or researchers at local institutions, may be valuable in characterizing the different types of farmers that exist in a locality. Acknowledging the differences that can exist between farmers can help to tailor engagement and participation strategies to make extension services and training sessions more appealing to a wider constituency of farmers. It may even be possible to recognize that a ‘just transition’ might mean something different to different types of farmers in the same region, and to craft different yet locally appropriate visions of a just transition that are possible within existing overarching regimes. For example, although we did not adopt this typology explicitly in our own study, one could argue that respondents in Thai Nguyen fit the Kuhmonen and Siltaoja [13] characterization of ‘adaptive’ farmers who aim at continuous development while fitting within the current regime, whereas those in Phu Tho are closer to the Kuhmonen and Siltaoja [13] characterization of ‘persistent’ farmers who continue with practices that have been used previously and may receive less of their income from farming. If one were to follow this typology, then participatory approaches and extension services towards a just transition for farmers in Thai Nguyen could be tailored to focus on the gradual introduction of new technologies and new crops in response to shifting weather and climate patterns, whereas those in Phu Tho could be tailored towards upskilling to diversify income streams and sustain livelihoods, whilst also discussing which farming practices can continue versus which may no longer be viable under a changing climate.

6.4. Limitations

Our analysis focused on a just transition in terms of identifying interventions that could build and sustain resilience for agriculture and its workers in two resource-dependent local economies. One of these was relatively poor (Phu Tho) and one was slightly better-off (Thai Nguyen). Both were located within a less wealthy region in Vietnam. We acknowledge that our findings are based on a relatively small sample of two tea- and rice-producing districts within each province. We focused on collating the direct experiences of farmers and sought to encompass diversity in gender, age, and education level in our sample, to nuance and build on what has been observed elsewhere in Vietnam and for resource-dependent just transition research in other contexts. Further research may, however, wish to test our insights against a larger sample covering more districts in each province to support specific policy and practice recommendations that enable a local just transition for resource-dependent workers. Building on our results, a subsequent sample could attempt to capture and characterize different ‘types’ of farmers [13] in order to more fully explore differences in adaptation practice and engagement in training depending on the characteristics of farmers.

Similarly, the data available did not allow us to engage with broader questions of equity in food production systems, or with structural drivers of vulnerability, in a way more transformative perspectives on sustainable livelihoods or just transitions for resource contexts call for (e.g., [8,10,19]). Further research in Thai Nguyen and Phu Tho—or resource-dependent regions like them—may wish to track provincial value and yields of crops over time in comparison to, for example, provincial wages in agriculture, forestry, and fisheries. This would help us understand the extent to which ‘restorative justice’ in terms of training, skills, and opportunity for livelihood diversification [13] is required as part of a just transition in Thai Nguyen and Phu Tho. Further qualitative enquiry into why engagement with extension services is low in both provinces would more fully respond to the Lo [5] challenges of designing more effective participatory approaches—and constructing a richer evidence base—within an authoritarian context.

7. Conclusions

Our aim was to use the case of two agricultural provinces in Vietnam with differing levels of wealth and productivity—Thai Nguyen and Phu Tho—to identify policy and practice interventions that may support a just transition towards more resilient practices in a resource-based context. We defined a resilient system as one that can continue to fulfil its core function under changing conditions, recognizing that one way a system may continue to achieve its main goal is by transforming [13]. To meet our aim, we surveyed 105 farming households across the two provinces.

A synthesis of how our findings respond to this aim is as follows. We found that farmers report experiencing the effects of climate change on their crops (Section 5.1), albeit to a limited extent. This indicates that measures to sustain the resilience of farming practices and support livelihoods may be required before impacts become more pronounced. Our data indicates that farmers in Thai Nguyen are more likely to adopt technical approaches to sustaining crop yields, whereas those in Phu Tho are more likely to continue to use traditional methods and techniques (Section 5.2 and Section 5.3). Although this intensification comes alongside higher incomes in Thai Nguyen, it may also put farmers in the province at greater risk if crops are negatively impacted by extreme events (Section 5.1). We found that applying traditional knowledge and practices to farming was linked to taking measures to adapt to climate change (Section 5.3 and Section 5.5), but also that engagement with extension services or participation in training was low across both provinces (Section 5.4). These findings reflect much existing recent research into climate adaptation and resilience for rural agriculture in Vietnam, especially the limited impact of extension services [18], the dangers of over-reliance on technocratic and top-down approaches to adaptation [19,32], and the important role that experience and traditional knowledge plays in leading farmers to adapt their practices [15,17].

Our first key empirical finding is that farmers experience weather extremes differently depending on the province, and on the crop they grow. Most surveyed farmers across the whole dataset reported experiencing drought, and just under half reported experiencing extreme cold (Section 5.1). This is consistent with some of the impacts predicted in Vietnam’s National Adaptation Plan [21]. However, Section 5.1 also showed that perceived impact of drought on crops was significantly higher in Phu Tho, whereas perceived impact of extreme cold was significantly higher in Thai Nguyen. We also observed (Section 5.1) a significant difference in perceived impacts of weather extremes on tea tree crops, but not on rice crops. Although Vietnam’s National Adaptation Plan projects flash floods, landslides, drought, and extreme cold for the Northern Midlands and Mountains, the actual impacts farmers experience could therefore differ substantially depending on their precise geographical location and on the crops they grow. The first policy intervention we propose based on our findings is therefore to create an interdisciplinary evidence base that can identify priority geographical locations where action is required and understand what the most urgent climate threats are that farmers are facing in their immediate surroundings. This evidence base should include not only social science and policy analysis, but also finer-scale natural and physical science evidence as well as farmers’ own anecdotal experiences and observations of the changes around them.

Our second key empirical finding is that the strategies farmers have been encouraged to pursue could put them at greater risk of climate impacts in the future. Section 3.1 reported policy shifts in the Northern Midlands and Mountains towards intensified commercial production of crops such as tea and emphasized creating a skilled agricultural workforce. We found that farmers in Thai Nguyen were more likely to employ technically driven measures to sustain productivity in the face of weather and climate changes (Section 5.3), and to access extension services (Section 5.4). However, we also found that farmers in Thai Nguyen were reporting greater experience of impacts on tea crops from extreme cold and floods (Section 5.1). In summary, the intensification of tea production in Thai Nguyen, in particular, although in line with recent policy decisions, could actually put farmers’ livelihoods at greater risk if it shifts them towards a single crop that is susceptible to the impacts of climate change. A policy and practice response in support of just transitions that we recommend is, therefore, for national and regional governments, and farmers themselves, to look beyond short-term financial gains to consider longer-term risks and vulnerabilities that may arise from intensification or monoculture. This means crafting policies and interventions that will be suitable for the kind of climate that is projected in coming decades.

Our third key empirical finding is the low, but also varying, use of extension services and participation in training across our dataset. Section 5.4 shows that engagement with extension services and participation in training are generally low. However, we do observe that engagement with extension services is significantly higher in Thai Nguyen. We also note in Section 5.5 that preparation for disasters is related to farm size for Thai Nguyen and that participating in training is related to the use of traditional knowledge in Thai Nguyen and the overall dataset. As Thai Nguyen is the wealthier of the two provinces studied, with Phu Tho having lower agricultural incomes and higher poverty rates, this suggests that engagement and participation opportunities are not reaching the most vulnerable farmers. The third recommendation we make in response to our research aim is therefore for regional and local governments to consider why participation is low, and to explore how to enable fuller participation in training and extension services. This might involve working with local governments and local researchers to characterize the different kinds of farmers within a locality, and tailor messaging, engagement, and training strategies accordingly.