The Case of the Null Protea SRS
At first sight this seems a subject for some hilarity, but on second thought it appears to deserve several appropriate responses. As Blikbrain, within certain parameters, accepts the situation quite amiably, the uninformed might suggest the Null Protea SRS as a good way to boost their SRS tally. However, it is evident that this approach would be self-defeating in many respects. Taken to its extremes, Blikbrain is sure to have hiccups when digesting a Site Record from the Central Kalahari. During data checking I have encountered only a few Null SRS and was therefore surprised to find that 1 700 Null SRS have been submitted. Unfortunately, I became aware too late of the facility to be of use.
A problem would be the size of the area covered by a Null SRS. A positive record could be an area from 1 m to 500 m in diameter. So, let us take the hypothetical case of an isolated stand of proteas surrounded by barren potential-protea country. The lone population could probably claim for itself a 500 m diameter terrain. In order to emphasize its splendid isolation, school geometry teaches that it can be surrounded by six circular areas also of 500 m diameter, i.e. six Null SRS. The purist might insist on another larger circle of Null SRS around the first and so on, but it is evident that the situation could get out of hand.
Another hypothetical case is the long schlep up a steep hill only to find no proteas on top. An Atlasser could vent his frustration by immortalizing the hill-top with a Null SRS. Traverses in rough country often takes place along tracks with site records at regular intervals. When a barren site pops up among the regular spacing, an Atlasser might be tempted by a Null SRS: this would only really be warranted if supported by lateral Null sites at a right angle to the traverse. It is evident that the size of the area, as well as other factors, should feature quite well in the Special Remarks box, but I have no knowledge if this appears in all Null SRS.
The project includes numerous parameters such as geographical position, climate, topography, geography and alien plant infestation that influence protea distribution, but underlying interactions of parameters, sometimes not easily discernable, may cause interesting variations on normal distribution of plants. One case is Ld cfrm that normally occurs quite extensively in the Cape of Good Hope Nature Area on wind-blown marine lime-rich sands. Along the east coast of the area there is consolidated calcified dunes of which the derived soil, as can be expected, supports a population of Ld cfrm for a short distance inland. There is a rather abrupt change to sandstone soil, supporting Pr lepi, so that the change can be used as a geological mapping tool.
However, similar calcified dunes occur in the south above Dias Strand, with some wind-blown marine sands further inland. But in this case there are no proteas at all. It is possible that until recently this area had been heavily infested with Rooikrans. On the west coast between Platboom and Gifkommetjie, there are also calcified dunes which, although well vegetated and without alien plants, supports no Ld cfrm. (A few plants were seen closer to the sea, on marine sands.)
Another geological example is shown by the granite hilltops near Darling, which, during geological excursions in the area, yielded no proteas. This is in stark contrast to similar granite areas at Paardeberg and Paarl.
In some places the habitat of small protea species atlassed some years ago, appear to have been taken over by more robust indigenous species, resulting in the local (temporary?) absence of proteas. Even robust proteas appear to have been the usurping plants in places.
A consideration for the project would be the depiction of Null SRS on maps. Normally vacant spaces on maps could mean either no proteas at the time of observation, or a total lack of observation.
It would be interesting to hear the views of experienced Atlassers on the topic of absence of proteas in likely places, and the eventual map recording. It would seem that 1 700 Null SRS, which appears at first sight unduly large, may in fact be insufficient for the whole project.
Chris van Vuuren, Simonstown
It is not strictly true to say that a Null Sight Record Sheet is not mentioned in the Protea Atlas Manual. On page 59 (OK, the very last page) we discuss what to do if you go to a site where you knew there were proteas present historically, but you could not locate them after an extensive search. There lies the rub. Null Sight Record Sheets must be extensively searched. A single plant means that a Null SRS is incorrect. The default size of a plot is 500 m. Generally smaller plots can be detected by having "Fragmented" or "Linear" in the Habitat Box Islands Code, or if confined by mans activities, the codes "Island", "Verge", "Corridor" or "Powerline".
However, your concern is valid. We do need to know where the proteas arent. And we will be using habitat information to do this. But we will not be using habitat data from the SRS to "fill in the blanks", because atlassers have not collected habitat data from the blanks. And we have a problem such habitat data do not exist. Our data are far too fine we cannot find comparable data with our habitat information to do such extrapolations. Any modelling of distributions will have to be done at a courser scale. The minute by minute modeling we are using for Bayesian modeling is the finest data available for the country.
Early on we realized that we had a problem. Any modelling exercise would have Ld salignum growing all the way to the central Kalahari (data from which Blikbrein can cope with easily, if proteas did occur there). However, for other species, plots in which they do not occur act as Null plots for those species. This assumes that the majority of atlassers have atlassed all species of proteas on their plots. So it is only for a few species that we have such large-scale problems. In reality, using rainfall and geology (as just two of many environmental variables available electronically on GIS) we can note that proteas never occur on Bokkeveld shales, unless rainfall exceeds 800 mm. We can therefore switch off a large area of the map as "protea-unsuitable." However, a few choice Null SRS are needed to confirm and "fix" this prediction. That is what our 1 500 Null SRS do. These Null SRS do not have to be every 500 m - a few strategically located Nulls can fill in large areas of map.
Far more of a problem is interpolation. Within Fynbos there are habitats that dont have proteas. Fortunately these are few and far between, but they do exist. The largest of these areas are the high altitude alpine restio- and grass-lands and the lower, dry, northern, sandstone slopes. For these we do need better Null SRS, but fortunately, Ld salignum obliges and occurs in low numbers on the borders of these areas. We had enough problems getting atlassers motivated to atlas these odd Conebushes. Imagine the problems we would have had with Null SRS!
So our strategy was to get the majority of atlassers to atlas the proteas, and choose a few, most meticulous, atlassers to go and atlas Nulls. In reality, only David Osborne and official Protea Atlas trips regularly mapped the Null Sight Record Sheets. Atlassers who we sent to visit unatlassed areas and reported unsuitable habitat and no proteas, also filled in Null SRS to prevent future fruitless trips to these areas.
This does mean that we do have problems. See the map below for our Null SRS. We are especially short of Null SRS in the Kouga Mountains. We still need to place a few strategic Null SRS in some Karoo and Thicket vegetation types, but on the whole our problem is really that of interpolation. Most of these problems are dry, lower slopes and Grassy Fynbos.
Thus while it is true that we should have more Null SRS than Protea SRS, in reality our strategic approach means that we have painted most of the map. The big surprise was just how widespread and pervasive proteas are in Fynbos!
The strategic placement of additional Null SRS is a crucial part of the Bayesian modelling project.
Back PAN 53