The three main agrogenic greenhouse gases are Carbon Dioxide, Methane, and Nitrous Oxide. I'll identify a farming practice which contributes to each of these: 

Intensive animal production - (CO2)
A recent IFOAM report notes that GHG emissions resulting from agriculture in Europe have been in steady decline since the 90s. This maps onto a number of factors: one is a reduction in livestock kept in this period, another is the 'exporting' of our carbon emissions. IFOAM suggests that the amount of carbon unlocked through deforestation to grow animal feed abroad, currently omitted from carbon counting due to the EU's accounting rules, is equal to 5% of Europe's total carbon emissions. This is more than is produced through enteric fermentation, and most of this carbon is an externality of the intensive pork and poultry industry, as these animals consume a high energy diet of imported grain. 

The Millenium Ecosystem Assessment predicts a 50% increase in food production by 2050 to keep up with population growth. IFOAM predicts that, if the cost of converting forest and grassland into pasture is accounted for, this growth will push the UK's agricultural carbon output to triple by in the same period. This is due to a global shift towards eating more meat, and due to intensive meats taking us far beyond our ecosystem's carrying capacity. (There is an inverse Kutznet relation between animal products and carbon that becomes exponential as we outstrip the carrying capacity of pasture.)

A full third of the world's global arable land is used for livestock feed, and there is a pronounced economic pressure to increase this amount. The amount of carbon locked in woody plants and grassland soil is so great that it dwarfs most agricultural mitigation measures. Preventing these changes in land use is the most significant way to reduce the carbon emissions in the agriculture industry. 

Overuse of Nitrogen fertilizer (N2O)

Nitrogen fertilizer reduces to Nitrous Oxide in the atmosphere. This is a long-lived greenhouse gas, which has a warming effect more than 200x greater than CO2. The UK produces more N2O than CO2 and CH4 combined, a disproportionate amount which is the greatest in Europe. [Eurostat]

Nitrogen balance is essential in agricultural systems: Nitrogen is the most important growth-limiting nutrient in soils, and it is very common in intensive systems to apply nitrogen fertilizer to increase crop yield. 

While there are times where the N carrying capacity of soil is greater: e.g. when soil is warm, when rain is light, and when crops are at a stage in their lifecycle where they are uptalking a lot of nitrogen, there is always an unprocessed excess. Studies show a direct linear relationship between Nitrogen fertilizer application and nitrous oxide emissions. Therefore reduction in nitrogen fertilizer applied will be the most effective route to reducing nitrogen emissions. 

There are also carbon costs association with the production of nitrogen fertilizers, as the Haber-Bosch process is very energy intensive. This cost is uncounted, but estimated to be around 4-5% of Europe's gross agrigenic GHG. [IFOAM] It also has a pronounced effect on the bioavailability of soil carbon (as it helps stable carbons decompose).

Misapplication of ruminant diet (CH4)

There has been much discussion around the importance of CH4 production on farm ever since the FAO's 'Livestock's long shadow' report, in which it was estimated that 7% of all global GHG emissions come from livestock. A large proportion of this is methane (CH4), which is produced in the gut of ruminants during enteric fermentation, the process which allows energy extraction from roughage. 

There is a lot of misinformation about the relevance of methane, much of which comes from the meat industry. The facts are these: ACE claims that Corporate campaign goals cannot be too strongly at odds with the profit incentives of the companies (otherwise making changes is much harder). (source)  -- which means it is broken down much faster than both CO2 and NH4. However, the hydroxyl radical in the upper atmosphere breaks CH4 down into CO2. So while its pronounced GHG effect is short lived, methane persists in the upper atmosphere as CO2 with a half-life of 100 years. 

The issue here is that we are producing ruminant livestock far beyond the carrying capacity of their ecosystem. When stocked correctly, and grazed carefully, ruminants are part of a carbon cycle that sequesters stable carbon deep into soil, mostly as root matter. There is an enlivened debate about whether it's better to reduce the methane produced by hastening cattle to slaughter by finishing them on high-energy grains, or it's better to reduce the carbon cost of grain production by feeding cattle exclusively on grass. The yield gap between pasture-fed and non-organic systems mean that while pasture fed cattle produce less CO2 per hectare, they produce more CO2 per kg of milk or beef. 

Either way, we produce way too much beef, and without radical changes to their feedstyle, this is having a destructive impact on the environment (albeit one that is dwarfed by pigs and poultry).

I think it's very hard to divide these issues, and identify 'one practice' that is more important than the others. We need to radically decarbonize our agricultural economy, and to do that we need to address the way that we eat, as well as the way that we farm. The highest-impact thing we could do, right now, is to stop eating pigs and poultry. 

However, I think that very careful production of organic and low-emission ruminant meat is a more culturally sensitive solution, with a greater chance of follow-through. There are several ways we can reduce ruminant emissions: 

	1. Improved feed. 
		a. Certain types of seaweed from the Asparagopsis family reduce the CH4 produced from enteric fermentation by introducing bromoformes which interrupt the gut microbes which produce methane as a byproduct of enteric fermentation. When fed at 0.5% ration, this reduces methane at a rate of 80%. This practice is increasingly common on organic farms, despite being an external input, because Asparagopsis's sporophyte variety can be produced in organic aquaculture. It also grows at a rate of 5cm per hour, and is an incredible global carbon sink. 
		b. It's possible to feed cattle a grass mixture which is much easier for them to digest, producing less methane. Pure ryegrass produces as little methane as a diet of maize, while producing much less nitrogen due to being lower protein. There is also a nitrogen factor: Urine from cattle creates spots that are far too nitrogen-rich for plants to absorb, so creates nitrogen pollution in run-off, or travels down liquid carbon channels to aid decomposition of deep soil carbon. So reducing nitrogen in cattle's urine is a smart idea. 
		
	2. Improved manure composting 
		a. IFOAM report suggests that it's possible to reduce N20 by 50% and methane by 70% by properly composting manure. This creates a carbon-rich compost which can be repurposed as part of a holistic system. It's important to processing livestock bedding separately, as liquid manure creates greater emissions, and aerate the manure regularly to prevent the production of methane (an anaerobic process).
		b. The manure can also be allowed to compost anaerobically, allowing the CH4 to be captured and used as biogas. This has the potential to offset the sourcing of gas from fossil fuels. 
		
	3. Improved grazing
		a. The potential for holistic planned grazing for carbon sequestration appears to have been overstated. While much of the world's soil is now depleted, and it's chance to sequester carbon is pronounced, the evidence suggests a hard biophysical limit, after which soil carbon is saturated. Whether HPG systems can create new soil has yet to be proven, but this would greatly increase the carbon sequestration potential of the cattle system. 
		b. I think that HPG is a culturally relevant solution: There is competition for land use, there is a demand for good food, combining carbon sequestration with a profitable land use seems more like to succeed in the long term than using all available land for pure carbon sequestration crops (like trees)
		c. The first 1m of global soil contains 3X global atmospheric carbon. The Grazed and Confused report concedes that in UK soils 60% of carbon is found beneath the first 30cm in a 1m bore. This means that there is a great potential for deep-rooted grass growth and dieback to have a significant impact on carbon sequestration, at least until soils saturate. 

	4. Agroforestry
		a. Trees have multiple benefits here: Woody plants are a significant carbon sink. High-lead producing trees (e.g. Aspen, poplar) deposit large amounts of leaf biomass, which is a atmospherically neutral way to fix some nitrogen into the soil. They also assist in denitrification of surface water by obstructing runoff, and abate NH4 pollution due to root filtration. 
		b. Trees also provide high-welfare shelter for animals, and reduce the amount of feed required to finish them, by providing shade (energy is required to cool the animal, requiring more feed). 
		

So in short: I think we should increase the ethical meat industry, increasing the global seaweed industry, and develop the organic market. IFOAM recommends that if all farms went organic today, we would see a 20% reduction in GHG emissions from the agricultural sector, because we could no longer use nitrogen in excess, or stock animals so far beyond the carrying capacity of their environment. However: changing diets, and developing new technologies to offset emissions, are still essential to meet climate targets.