DAS Farm Benchmarks compare the long-term production of farms against

surrounding farms of the same type. Three benchmarks are available: Overall

Production, Resilience and Consistency

Overall Production - is the long-term mean production of a farm compared to its

peers and provides insights into questions such as “which is the most productive

farm in this area?”

Resilience - is the mean production of a farm compared to its peers in the worst

five years of the last 22 and provides insights into questions such as “in the bad

years, which farms perform best or struggle most in this area?”

Consistency - is the long-term production variation of a farm compared to its

peers and provides insights into questions such as “which is the consistent or

reliable farm in this area?”

DAS Farm Benchmarks seek to answer these questions by controlling for

differences between farms, such as climate and production system - honing in on

the long-term management performance of that farm business.

Farm production is measured here using Net Primary Productivity (NPP), which is

an estimate of the rate of carbon assimilation by plants (plant growth). NPP is

used because it strongly correlates with both crop yield and stock-carrying

capacity. Fundamentally, farm production is related to the amount of plant

matter grown. NPP is the most targeted measure of this.

DAS has NPP data for all of Australia from 2001 to the present at a 16-day cadence.

Three “raw” NPP metrics are calculated through different temporal reductions of

the NPP time series, with cropping land calculated differently from grazing land.

Grazing land considers all NPP values across all years because stock requires to

feed all year. The annual maximum NPP value is used for cropping land as this

most strongly correlates with crop production.

Overall Production is the mean NPP of all images in the dataset (or the mean of

annual maximum NPP for all years for cropping).

Resilience is the mean NPP (or mean annual max for cropping land) of the five

years where NPP was lowest.

Consistency is the standard deviation of NPP of all images in the dataset (or

the standard deviation of annual max for cropping).

These “raw” NPP metrics are useful in themselves as an absolute measure of

productivity between farms and across regions. But in these raw metrics, the

dominant influence of land use and rainfall is clear (Figure 1).

Figure 1. Raw Overall Production values across SE Australia. Note how values

change according to both land use and rainfall.

To control for the fundamental differences in production between regions, DAS

Farm Benchmarks are a peer-to-peer comparison that aims to assess the

management performance of a farm business. We do this by only comparing the

areas of a farm with areas of other farms with the same land use within 30

kilometres.

Each pixel within a farm is compared to pixels of the same land use within a 30

km radius. This distance was chosen because long-term rainfall patterns do not

vary significantly at this distance (water availability being a primary constraint of

plant growth). This radius also provides ~283,000 hectares to find pixels of the

same land use to compare - providing a large sample size for the comparison.

Each pixel is compared to its peers and scored using the z-score. This standardises

the comparison between pixels and their peers and gives a consistent range of

scores across all areas.

Z-scores are in units of standard deviations. To improve interpretability z-scores

are converted to probability deciles. This means that a farm with a Benchmark of

decile 9 performs better than 90% of its peers for that Benchmark. A decile 5 farm

is a median farm - better than half of its peers but worse than the other half.

Each farm is compared to its peers, and each farm has its own set of peers. So in

the image below, we can see the benchmark value of all farms in a 30 km radius,

but these are the peers of only the centre farm. All the other farms each have their

own peers, from which their benchmark is calculated. This means that a different

benchmark value for a farm on the far west of the image compared to one on the

far east cannot be interpreted as one having higher/lower productivity than the

other - it means that one performs better than the other when compared to their

respective peers.

Figure 2. Resilience Benchmark values for farms within a 30 km radius area in

Western Australia.

Pixels within farms are compared to pixels of the same broad land use types

within 30 km to account for different products from different land use types (e.g.

irrigated vs dryland cropping). This is done using the Australian Land Use and

Management Classification (ALUM) dataset produced by ABARES. While this

dataset is the best available land use mapping, it is not perfect.

Where a farm’s ALUM-mapped land use is incorrect, the benchmarks are likely,

not valid.

Our method uses a 30 km radius to control for climatic variability. Where

properties are very large, such as pastoral stations, the majority of peer pixels may,

in fact, fall within the same property. This invalidates the assumption that peer

pixels are under different management control. DAS Farm Benchmarks are not

appropriate for comparing very large properties.

Australian soil types are highly heterogeneous by global standards and coarsely

mapped. Soil type is not controlled for in the calculation of DAS Farm

Benchmarks because in many places, it would result in too few peers to provide

statistically appropriate comparisons. Consequently, the farms with better soils

may have higher benchmark values than their peers with poorer soils.

Many farms have areas that are not used for annual crops or grazing or are

unproductive. Examples include remnant vegetation, plantations, saline land and

buildings. Where these areas are mapped as such by ALUM, they are masked in

the calculation of Benchmarks. Where they are not and are a significant

proportion of the farm, they may affect Benchmark values.

The NPP metric used by DAS is designed specifically for annual plant-based

systems. When applied over plantations and remnant vegetation (woody

vegetation), it tends to return lower values than it does for annual plants. This

means that farms with large areas of woody vegetation that are not mapped by

ACLUM, or unproductive areas, will tend to compare poorly to their peers.